[jira] [Comment Edited] (HDFS-15237) Get checksum of EC file failed, when some block is missing or corrupt
[
https://issues.apache.org/jira/browse/HDFS-15237?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17068263#comment-17068263
]
guojh edited comment on HDFS-15237 at 3/27/20, 4:42 AM:
[~zhengchenyu] Have you try this patch
[HDFS-14754|https://issues.apache.org/jira/browse/HDFS-14754] ? Looks like the
redundant block has not been deleted first.
was (Author: gjhkael):
[~zhengchenyu] Have you try this patch
(https://issues.apache.org/jira/browse/HDFS-14754)
> Get checksum of EC file failed, when some block is missing or corrupt
> -
>
> Key: HDFS-15237
> URL: https://issues.apache.org/jira/browse/HDFS-15237
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: ec, hdfs
>Affects Versions: 3.2.1
>Reporter: zhengchenyu
>Priority: Major
> Fix For: 3.2.2
>
>
> When we distcp from an ec directory to another one, I found some error like
> this.
> {code}
> 2020-03-20 20:18:21,366 WARN [main]
> org.apache.hadoop.hdfs.FileChecksumHelper: src=/EC/6-3//000325_0,
> datanodes[6]=DatanodeInfoWithStorage[10.200.128.40:9866,DS-65ac4407-9d33-4c59-8f72-dd1d80d26d9f,DISK]2020-03-20
> 20:18:21,366 WARN [main] org.apache.hadoop.hdfs.FileChecksumHelper:
> src=/EC/6-3//000325_0,
> datanodes[6]=DatanodeInfoWithStorage[10.200.128.40:9866,DS-65ac4407-9d33-4c59-8f72-dd1d80d26d9f,DISK]java.io.EOFException:
> Unexpected EOF while trying to read response from server at
> org.apache.hadoop.hdfs.protocolPB.PBHelperClient.vintPrefixed(PBHelperClient.java:550)
> at
> org.apache.hadoop.hdfs.FileChecksumHelper$StripedFileNonStripedChecksumComputer.tryDatanode(FileChecksumHelper.java:709)
> at
> org.apache.hadoop.hdfs.FileChecksumHelper$StripedFileNonStripedChecksumComputer.checksumBlockGroup(FileChecksumHelper.java:664)
> at
> org.apache.hadoop.hdfs.FileChecksumHelper$StripedFileNonStripedChecksumComputer.checksumBlocks(FileChecksumHelper.java:638)
> at
> org.apache.hadoop.hdfs.FileChecksumHelper$FileChecksumComputer.compute(FileChecksumHelper.java:252)
> at
> org.apache.hadoop.hdfs.DFSClient.getFileChecksumInternal(DFSClient.java:1790)
> at
> org.apache.hadoop.hdfs.DFSClient.getFileChecksumWithCombineMode(DFSClient.java:1810)
> at
> org.apache.hadoop.hdfs.DistributedFileSystem$33.doCall(DistributedFileSystem.java:1691)
> at
> org.apache.hadoop.hdfs.DistributedFileSystem$33.doCall(DistributedFileSystem.java:1688)
> at
> org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:81)
> at
> org.apache.hadoop.hdfs.DistributedFileSystem.getFileChecksum(DistributedFileSystem.java:1700)
> at
> org.apache.hadoop.tools.mapred.RetriableFileCopyCommand.doCopy(RetriableFileCopyCommand.java:138)
> at
> org.apache.hadoop.tools.mapred.RetriableFileCopyCommand.doExecute(RetriableFileCopyCommand.java:115)
> at
> org.apache.hadoop.tools.util.RetriableCommand.execute(RetriableCommand.java:87)
> at
> org.apache.hadoop.tools.mapred.CopyMapper.copyFileWithRetry(CopyMapper.java:259)
> at org.apache.hadoop.tools.mapred.CopyMapper.map(CopyMapper.java:220) at
> org.apache.hadoop.tools.mapred.CopyMapper.map(CopyMapper.java:48) at
> org.apache.hadoop.mapreduce.Mapper.run(Mapper.java:146) at
> org.apache.hadoop.mapred.MapTask.runNewMapper(MapTask.java:799) at
> org.apache.hadoop.mapred.MapTask.run(MapTask.java:347) at
> org.apache.hadoop.mapred.YarnChild$2.run(YarnChild.java:174) at
> java.security.AccessController.doPrivileged(Native Method) at
> javax.security.auth.Subject.doAs(Subject.java:422) at
> org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1730)
> at org.apache.hadoop.mapred.YarnChild.main(YarnChild.java:168)
> {code}
> And Then I found some error in datanode like this
> {code}
> 2020-03-20 20:54:16,573 INFO
> org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient:
> SASL encryption trust check: localHostTrusted = false, remoteHostTrusted =
> false
> 2020-03-20 20:54:16,577 ERROR
> org.apache.hadoop.hdfs.server.datanode.DataNode:
> bd-hadoop-128050.zeus.lianjia.com:9866:DataXceiver error processing
> BLOCK_GROUP_CHECKSUM operation src: /10.201.1.38:33264 dst:
> /10.200.128.50:9866
> java.lang.UnsupportedOperationException
> at java.nio.ByteBuffer.array(ByteBuffer.java:994)
> at
> org.apache.hadoop.hdfs.server.datanode.erasurecode.StripedBlockChecksumReconstructor.reconstruct(StripedBlockChecksumReconstructor.java:90)
> at
> org.apache.hadoop.hdfs.server.datanode.BlockChecksumHelper$BlockGroupNonStripedChecksumComputer.recalculateChecksum(BlockChecksumHelper.java:711)
> at
> org.apache.hadoop.hdfs.server.datanode.BlockChecksumHelper$BlockGroupNonStripedChecksumComputer.compute(BlockChecksumHelper.java:489)
> at
> org.apache.hadoop.hdfs.server.datanod
[jira] [Commented] (HDFS-15237) Get checksum of EC file failed, when some block is missing or corrupt
[
https://issues.apache.org/jira/browse/HDFS-15237?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17068263#comment-17068263
]
guojh commented on HDFS-15237:
--
[~zhengchenyu] Have you try this patch
(https://issues.apache.org/jira/browse/HDFS-14754)
> Get checksum of EC file failed, when some block is missing or corrupt
> -
>
> Key: HDFS-15237
> URL: https://issues.apache.org/jira/browse/HDFS-15237
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: ec, hdfs
>Affects Versions: 3.2.1
>Reporter: zhengchenyu
>Priority: Major
> Fix For: 3.2.2
>
>
> When we distcp from an ec directory to another one, I found some error like
> this.
> {code}
> 2020-03-20 20:18:21,366 WARN [main]
> org.apache.hadoop.hdfs.FileChecksumHelper: src=/EC/6-3//000325_0,
> datanodes[6]=DatanodeInfoWithStorage[10.200.128.40:9866,DS-65ac4407-9d33-4c59-8f72-dd1d80d26d9f,DISK]2020-03-20
> 20:18:21,366 WARN [main] org.apache.hadoop.hdfs.FileChecksumHelper:
> src=/EC/6-3//000325_0,
> datanodes[6]=DatanodeInfoWithStorage[10.200.128.40:9866,DS-65ac4407-9d33-4c59-8f72-dd1d80d26d9f,DISK]java.io.EOFException:
> Unexpected EOF while trying to read response from server at
> org.apache.hadoop.hdfs.protocolPB.PBHelperClient.vintPrefixed(PBHelperClient.java:550)
> at
> org.apache.hadoop.hdfs.FileChecksumHelper$StripedFileNonStripedChecksumComputer.tryDatanode(FileChecksumHelper.java:709)
> at
> org.apache.hadoop.hdfs.FileChecksumHelper$StripedFileNonStripedChecksumComputer.checksumBlockGroup(FileChecksumHelper.java:664)
> at
> org.apache.hadoop.hdfs.FileChecksumHelper$StripedFileNonStripedChecksumComputer.checksumBlocks(FileChecksumHelper.java:638)
> at
> org.apache.hadoop.hdfs.FileChecksumHelper$FileChecksumComputer.compute(FileChecksumHelper.java:252)
> at
> org.apache.hadoop.hdfs.DFSClient.getFileChecksumInternal(DFSClient.java:1790)
> at
> org.apache.hadoop.hdfs.DFSClient.getFileChecksumWithCombineMode(DFSClient.java:1810)
> at
> org.apache.hadoop.hdfs.DistributedFileSystem$33.doCall(DistributedFileSystem.java:1691)
> at
> org.apache.hadoop.hdfs.DistributedFileSystem$33.doCall(DistributedFileSystem.java:1688)
> at
> org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:81)
> at
> org.apache.hadoop.hdfs.DistributedFileSystem.getFileChecksum(DistributedFileSystem.java:1700)
> at
> org.apache.hadoop.tools.mapred.RetriableFileCopyCommand.doCopy(RetriableFileCopyCommand.java:138)
> at
> org.apache.hadoop.tools.mapred.RetriableFileCopyCommand.doExecute(RetriableFileCopyCommand.java:115)
> at
> org.apache.hadoop.tools.util.RetriableCommand.execute(RetriableCommand.java:87)
> at
> org.apache.hadoop.tools.mapred.CopyMapper.copyFileWithRetry(CopyMapper.java:259)
> at org.apache.hadoop.tools.mapred.CopyMapper.map(CopyMapper.java:220) at
> org.apache.hadoop.tools.mapred.CopyMapper.map(CopyMapper.java:48) at
> org.apache.hadoop.mapreduce.Mapper.run(Mapper.java:146) at
> org.apache.hadoop.mapred.MapTask.runNewMapper(MapTask.java:799) at
> org.apache.hadoop.mapred.MapTask.run(MapTask.java:347) at
> org.apache.hadoop.mapred.YarnChild$2.run(YarnChild.java:174) at
> java.security.AccessController.doPrivileged(Native Method) at
> javax.security.auth.Subject.doAs(Subject.java:422) at
> org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1730)
> at org.apache.hadoop.mapred.YarnChild.main(YarnChild.java:168)
> {code}
> And Then I found some error in datanode like this
> {code}
> 2020-03-20 20:54:16,573 INFO
> org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient:
> SASL encryption trust check: localHostTrusted = false, remoteHostTrusted =
> false
> 2020-03-20 20:54:16,577 ERROR
> org.apache.hadoop.hdfs.server.datanode.DataNode:
> bd-hadoop-128050.zeus.lianjia.com:9866:DataXceiver error processing
> BLOCK_GROUP_CHECKSUM operation src: /10.201.1.38:33264 dst:
> /10.200.128.50:9866
> java.lang.UnsupportedOperationException
> at java.nio.ByteBuffer.array(ByteBuffer.java:994)
> at
> org.apache.hadoop.hdfs.server.datanode.erasurecode.StripedBlockChecksumReconstructor.reconstruct(StripedBlockChecksumReconstructor.java:90)
> at
> org.apache.hadoop.hdfs.server.datanode.BlockChecksumHelper$BlockGroupNonStripedChecksumComputer.recalculateChecksum(BlockChecksumHelper.java:711)
> at
> org.apache.hadoop.hdfs.server.datanode.BlockChecksumHelper$BlockGroupNonStripedChecksumComputer.compute(BlockChecksumHelper.java:489)
> at
> org.apache.hadoop.hdfs.server.datanode.DataXceiver.blockGroupChecksum(DataXceiver.java:1047)
> at
> org.apache.hadoop.hdfs.protocol.datatransfer.Receiver.opStripedBlockChecksum(Receiver.java:327)
> at
> org.apache.hadoop.hdfs.protocol.datatransfer.Receiver.processOp(Receiv
[jira] [Commented] (HDFS-15240) Erasure Coding: dirty buffer causes reconstruction block error
[
https://issues.apache.org/jira/browse/HDFS-15240?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17067559#comment-17067559
]
guojh commented on HDFS-15240:
--
[~surendrasingh] We have the same problem，please take a look.
> Erasure Coding: dirty buffer causes reconstruction block error
> --
>
> Key: HDFS-15240
> URL: https://issues.apache.org/jira/browse/HDFS-15240
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding
>Reporter: HuangTao
>Assignee: HuangTao
>Priority: Major
> Attachments: HDFS-15240.001.patch
>
>
> When read some lzo files we found some blocks were broken.
> I read back all internal blocks(b0-b8) of the block group(RS-6-3-1024k), and
> choose 6(b0-b5) blocks to decode other 3(b6', b7', b8') blocks. And find the
> longest common sequenece(LCS) between b6' and b6(b7'/b7 and b8'/b8).
> After selecting 6 blocks of the block group in combinations one time and
> iterating through all cases, I find one case that the length of LCS is the
> block length - 64KB, 64KB is just the length of ByteBuffer used by
> StripedBlockReader. So the corrupt reconstruction block is made by a dirty
> buffer.
> The following log snippet is my check program output
> {code:java}
> decode from [0, 2, 3, 4, 5, 7] -> [1, 6, 8]
> Check Block(1) first 131072 bytes longest common substring length 4
> Check Block(6) first 131072 bytes longest common substring length 4
> Check Block(8) first 131072 bytes longest common substring length 4
> decode from [0, 2, 3, 4, 5, 6] -> [1, 7, 8]
> Check Block(1) first 131072 bytes longest common substring length 65536
> CHECK AGAIN: Block(1) all 27262976 bytes longest common substring length
> 27197440 # this one
> Check Block(7) first 131072 bytes longest common substring length 4
> Check Block(8) first 131072 bytes longest common substring length 4{code}
> Now I know the dirty buffer causes reconstruction block error, but how does
> the dirty buffer come about?
> After digging into the code and DN log, I found this following DN log is the
> root reason.
> {code:java}
> [INFO] [stripedRead-1017] : Interrupted while waiting for IO on channel
> java.nio.channels.SocketChannel[connected local=/:52586
> remote=/:50010]. 18 millis timeout left.
> [WARN] [StripedBlockReconstruction-199] : Failed to reconstruct striped
> block: BP-714356632--1519726836856:blk_-YY_3472979393
> java.lang.NullPointerException
> at
> org.apache.hadoop.hdfs.util.StripedBlockUtil.getNextCompletedStripedRead(StripedBlockUtil.java:314)
> at
> org.apache.hadoop.hdfs.server.datanode.erasurecode.StripedReader.doReadMinimumSources(StripedReader.java:308)
> at
> org.apache.hadoop.hdfs.server.datanode.erasurecode.StripedReader.readMinimumSources(StripedReader.java:269)
> at
> org.apache.hadoop.hdfs.server.datanode.erasurecode.StripedBlockReconstructor.reconstruct(StripedBlockReconstructor.java:94)
> at
> org.apache.hadoop.hdfs.server.datanode.erasurecode.StripedBlockReconstructor.run(StripedBlockReconstructor.java:60)
> at
> java.base/java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:515)
> at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264)
> at
> java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)
> at
> java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)
> at java.base/java.lang.Thread.run(Thread.java:834) {code}
> Reading from DN may timeout(hold by a future(F)) and output the INFO log, but
> the futures that contains the future(F) is cleared,
> {code:java}
> return new StripingChunkReadResult(futures.remove(future),
> StripingChunkReadResult.CANCELLED); {code}
> futures.remove(future) cause NPE. So the EC reconstruction is failed. In the
> finally phase, the code snippet in *getStripedReader().close()*
> {code:java}
> reconstructor.freeBuffer(reader.getReadBuffer());
> reader.freeReadBuffer();
> reader.closeBlockReader(); {code}
> free buffer firstly, but the StripedBlockReader still holds the buffer and
> write it.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-15186) Erasure Coding: Decommission may generate the parity block's content with all 0 in some case
[ https://issues.apache.org/jira/browse/HDFS-15186?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17043125#comment-17043125 ] guojh commented on HDFS-15186: -- [~ayushtkn] This problem can fixed in namenode side, I have the same problem that fixed by [HDFS-14768|https://issues.apache.org/jira/browse/HDFS-14768] > Erasure Coding: Decommission may generate the parity block's content with all > 0 in some case > > > Key: HDFS-15186 > URL: https://issues.apache.org/jira/browse/HDFS-15186 > Project: Hadoop HDFS > Issue Type: Bug > Components: datanode, erasure-coding >Affects Versions: 3.0.3, 3.2.1, 3.1.3 >Reporter: Yao Guangdong >Assignee: Yao Guangdong >Priority: Critical > Attachments: HDFS-15186.001.patch > > > I can find some parity block's content with all 0 when i decommission some > DataNode(more than 1) from a cluster. And the probability is very big(parts > per thousand).This is a big problem.You can think that if we read data from > the zero parity block or use the zero parity block to recover a block which > can make us use the error data even we don't know it. > There is some case in the below: > B: Busy DataNode, > D:Decommissioning DataNode, > Others is normal. > 1.Group indices is [0, 1, 2, 3, 4, 5, 6(B,D), 7, 8(D)]. > 2.Group indices is [0(B,D), 1, 2, 3, 4, 5, 6(B,D), 7, 8(D)]. > > In the first case when the block group indices is [0, 1, 2, 3, 4, 5, 6(B,D), > 7, 8(D)], the DN may received reconstruct block command and the > liveIndices=[0, 1, 2, 3, 4, 5, 7, 8] and the targets's(the field which in > the class StripedReconstructionInfo) length is 2. > The targets's length is 2 which mean that the DataNode need recover 2 > internal block in current code.But from the liveIndices we only can find 1 > missing block, so the method StripedWriter#initTargetIndices will use 0 as > the default recover block and don't care the indices 0 is in the sources > indices or not. > When they use sources indices [0, 1, 2, 3, 4, 5] to recover indices [6, 0] > use the ec algorithm.We can find that the indices [0] is in the both the > sources indices and the targets indices in this case. The returned target > buffer in the indices [6] is always 0 from the ec algorithm.So I think this > is the ec algorithm's problem. Because it should more fault tolerance.I try > to fixed it .But it is too hard. Because the case is too more. The second is > another case in the example above(use sources indices [1, 2, 3, 4, 5, 7] to > recover indices [0, 6, 0]). So I changed my mind.Invoke the ec algorithm > with a correct parameters. Which mean that remove the duplicate target > indices 0 in this case.Finally, I fixed it in this way. > -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16966312#comment-16966312
]
guojh commented on HDFS-14768:
--
[~surendrasingh] I will add patch for branch3.1 and 3.2 soon or later.
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.009.patch, HDFS-14768.010.patch, HDFS-14768.011.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
[jira] [Comment Edited] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16966312#comment-16966312
]
guojh edited comment on HDFS-14768 at 11/3/19 11:53 PM:
[~surendrasingh] I will add patch for branch3.1 and 3.2 later.
was (Author: gjhkael):
[~surendrasingh] I will add patch for branch3.1 and 3.2 soon or later.
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.009.patch, HDFS-14768.010.patch, HDFS-14768.011.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the nod
[jira] [Commented] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16964767#comment-16964767
]
guojh commented on HDFS-14768:
--
[~surendrasingh] [~ferhui] I'm sorry for I lost my patience. rebase trunk on
the new patch.
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.009.patch, HDFS-14768.010.patch, HDFS-14768.011.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> dec
[jira] [Updated] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.011.patch
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.009.patch, HDFS-14768.010.patch, HDFS-14768.011.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNum
[jira] [Resolved] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh resolved HDFS-14768.
--
Resolution: Not A Problem
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.009.patch, HDFS-14768.010.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.
[jira] [Reopened] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh reopened HDFS-14768:
--
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.009.patch, HDFS-14768.010.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDat
[jira] [Resolved] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh resolved HDFS-14768.
--
Resolution: Abandoned
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.009.patch, HDFS-14768.010.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getD
[jira] [Updated] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Status: Open (was: Patch Available)
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.009.patch, HDFS-14768.010.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes())
[jira] [Commented] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16960850#comment-16960850
]
guojh commented on HDFS-14768:
--
[~surendrasingh] Thanks for you UT, I update the code and fixed the checkstyle
error. Please review it. Thank you very much!
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.009.patch, HDFS-14768.010.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
>
[jira] [Commented] (HDFS-14920) Erasure Coding: Decommission may hang If one or more datanodes are out of service during decommission
[
https://issues.apache.org/jira/browse/HDFS-14920?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16960817#comment-16960817
]
guojh commented on HDFS-14920:
--
Thanks [~ferhui] I am clear now. You just want to correct the decommissioning
counter, If index[0, 1, 2, 3] is decommissioning, and another block with index
0 is replica success, Now decommission counter should be 3 not 4.
> Erasure Coding: Decommission may hang If one or more datanodes are out of
> service during decommission
> ---
>
> Key: HDFS-14920
> URL: https://issues.apache.org/jira/browse/HDFS-14920
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: ec
>Affects Versions: 3.0.3, 3.2.1, 3.1.3
>Reporter: Fei Hui
>Assignee: Fei Hui
>Priority: Major
> Attachments: HDFS-14920.001.patch, HDFS-14920.002.patch
>
>
> Decommission test hangs in our clusters.
> Have seen the messages as follow
> {quote}
> 2019-10-22 15:58:51,514 TRACE
> org.apache.hadoop.hdfs.server.blockmanagement.DatanodeAdminManager: Block
> blk_-9223372035600425840_372987973 numExpected=9, numLive=5
> 2019-10-22 15:58:51,514 INFO BlockStateChange: Block:
> blk_-9223372035600425840_372987973, Expected Replicas: 9, live replicas: 5,
> corrupt replicas: 0, decommissioned replicas: 0, decommissioning replicas: 4,
> maintenance replicas: 0, live entering maintenance replicas: 0, excess
> replicas: 0, Is Open File: false, Datanodes having this block:
> 10.255.43.57:50010 10.255.53.12:50010 10.255.63.12:50010 10.255.62.39:50010
> 10.255.37.36:50010 10.255.33.15:50010 10.255.69.29:50010 10.255.51.13:50010
> 10.255.64.15:50010 , Current Datanode: 10.255.69.29:50010, Is current
> datanode decommissioning: true, Is current datanode entering maintenance:
> false
> 2019-10-22 15:58:51,514 DEBUG
> org.apache.hadoop.hdfs.server.blockmanagement.DatanodeAdminManager: Node
> 10.255.69.29:50010 still has 1 blocks to replicate before it is a candidate
> to finish Decommission In Progress
> {quote}
> After digging the source code and cluster log, guess it happens as follow
> steps.
> # Storage strategy is RS-6-3-1024k.
> # EC block b consists of b0, b1, b2, b3, b4, b5, b6, b7, b8, b0 is from
> datanode dn0, b1 is from datanode dn1, ...etc
> # At the beginning dn0 is in decommission progress, b0 is replicated
> successfully, and dn0 is staill in decommission progress.
> # Later b1, b2, b3 in decommission progress, and dn4 containing b4 is out of
> service, so need to reconstruct, and create ErasureCodingWork to do it, in
> the ErasureCodingWork, additionalReplRequired is 4
> # Because hasAllInternalBlocks is false, Will call
> ErasureCodingWork#addTaskToDatanode ->
> DatanodeDescriptor#addBlockToBeErasureCoded, and send
> BlockECReconstructionInfo task to Datanode
> # DataNode can not reconstruction the block because targets is 4, greater
> than 3( parity number).
> There is a problem as follow, from BlockManager.java#scheduleReconstruction
> {code}
> // should reconstruct all the internal blocks before scheduling
> // replication task for decommissioning node(s).
> if (additionalReplRequired - numReplicas.decommissioning() -
> numReplicas.liveEnteringMaintenanceReplicas() > 0) {
> additionalReplRequired = additionalReplRequired -
> numReplicas.decommissioning() -
> numReplicas.liveEnteringMaintenanceReplicas();
> }
> {code}
> Should reconstruction firstly and then replicate for decommissioning. Because
> numReplicas.decommissioning() is 4, and additionalReplRequired is 4, that's
> wrong,
> numReplicas.decommissioning() should be 3, it should exclude live replica.
> If so, additionalReplRequired will be 1, reconstruction will schedule as
> expected. After that, decommission goes on.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.010.patch
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.009.patch, HDFS-14768.010.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
>
[jira] [Commented] (HDFS-14920) Erasure Coding: Decommission may hang If one or more datanodes are out of service during decommission
[
https://issues.apache.org/jira/browse/HDFS-14920?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16960772#comment-16960772
]
guojh commented on HDFS-14920:
--
[~ferhui] If the block group not has all internal blocks, the
additionalReplRequired should compute by liveBlockIndicies, just according the
block's real total block and live block indicies. Is it more sample? And you
patch seems not solved this issue? Have I missing anything?
> Erasure Coding: Decommission may hang If one or more datanodes are out of
> service during decommission
> ---
>
> Key: HDFS-14920
> URL: https://issues.apache.org/jira/browse/HDFS-14920
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: ec
>Affects Versions: 3.0.3, 3.2.1, 3.1.3
>Reporter: Fei Hui
>Assignee: Fei Hui
>Priority: Major
> Attachments: HDFS-14920.001.patch, HDFS-14920.002.patch
>
>
> Decommission test hangs in our clusters.
> Have seen the messages as follow
> {quote}
> 2019-10-22 15:58:51,514 TRACE
> org.apache.hadoop.hdfs.server.blockmanagement.DatanodeAdminManager: Block
> blk_-9223372035600425840_372987973 numExpected=9, numLive=5
> 2019-10-22 15:58:51,514 INFO BlockStateChange: Block:
> blk_-9223372035600425840_372987973, Expected Replicas: 9, live replicas: 5,
> corrupt replicas: 0, decommissioned replicas: 0, decommissioning replicas: 4,
> maintenance replicas: 0, live entering maintenance replicas: 0, excess
> replicas: 0, Is Open File: false, Datanodes having this block:
> 10.255.43.57:50010 10.255.53.12:50010 10.255.63.12:50010 10.255.62.39:50010
> 10.255.37.36:50010 10.255.33.15:50010 10.255.69.29:50010 10.255.51.13:50010
> 10.255.64.15:50010 , Current Datanode: 10.255.69.29:50010, Is current
> datanode decommissioning: true, Is current datanode entering maintenance:
> false
> 2019-10-22 15:58:51,514 DEBUG
> org.apache.hadoop.hdfs.server.blockmanagement.DatanodeAdminManager: Node
> 10.255.69.29:50010 still has 1 blocks to replicate before it is a candidate
> to finish Decommission In Progress
> {quote}
> After digging the source code and cluster log, guess it happens as follow
> steps.
> # Storage strategy is RS-6-3-1024k.
> # EC block b consists of b0, b1, b2, b3, b4, b5, b6, b7, b8, b0 is from
> datanode dn0, b1 is from datanode dn1, ...etc
> # At the beginning dn0 is in decommission progress, b0 is replicated
> successfully, and dn0 is staill in decommission progress.
> # Later b1, b2, b3 in decommission progress, and dn4 containing b4 is out of
> service, so need to reconstruct, and create ErasureCodingWork to do it, in
> the ErasureCodingWork, additionalReplRequired is 4
> # Because hasAllInternalBlocks is false, Will call
> ErasureCodingWork#addTaskToDatanode ->
> DatanodeDescriptor#addBlockToBeErasureCoded, and send
> BlockECReconstructionInfo task to Datanode
> # DataNode can not reconstruction the block because targets is 4, greater
> than 3( parity number).
> There is a problem as follow, from BlockManager.java#scheduleReconstruction
> {code}
> // should reconstruct all the internal blocks before scheduling
> // replication task for decommissioning node(s).
> if (additionalReplRequired - numReplicas.decommissioning() -
> numReplicas.liveEnteringMaintenanceReplicas() > 0) {
> additionalReplRequired = additionalReplRequired -
> numReplicas.decommissioning() -
> numReplicas.liveEnteringMaintenanceReplicas();
> }
> {code}
> Should reconstruction firstly and then replicate for decommissioning. Because
> numReplicas.decommissioning() is 4, and additionalReplRequired is 4, that's
> wrong,
> numReplicas.decommissioning() should be 3, it should exclude live replica.
> If so, additionalReplRequired will be 1, reconstruction will schedule as
> expected. After that, decommission goes on.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14920) Erasure Coding: Decommission may hang If one or more datanodes are out of service during decommission
[
https://issues.apache.org/jira/browse/HDFS-14920?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16960745#comment-16960745
]
guojh commented on HDFS-14920:
--
[~ferhui] As you description，this is a bug to compute the
additionalReplRequired.
> Erasure Coding: Decommission may hang If one or more datanodes are out of
> service during decommission
> ---
>
> Key: HDFS-14920
> URL: https://issues.apache.org/jira/browse/HDFS-14920
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: ec
>Affects Versions: 3.0.3, 3.2.1, 3.1.3
>Reporter: Fei Hui
>Assignee: Fei Hui
>Priority: Major
> Attachments: HDFS-14920.001.patch, HDFS-14920.002.patch
>
>
> Decommission test hangs in our clusters.
> Have seen the messages as follow
> {quote}
> 2019-10-22 15:58:51,514 TRACE
> org.apache.hadoop.hdfs.server.blockmanagement.DatanodeAdminManager: Block
> blk_-9223372035600425840_372987973 numExpected=9, numLive=5
> 2019-10-22 15:58:51,514 INFO BlockStateChange: Block:
> blk_-9223372035600425840_372987973, Expected Replicas: 9, live replicas: 5,
> corrupt replicas: 0, decommissioned replicas: 0, decommissioning replicas: 4,
> maintenance replicas: 0, live entering maintenance replicas: 0, excess
> replicas: 0, Is Open File: false, Datanodes having this block:
> 10.255.43.57:50010 10.255.53.12:50010 10.255.63.12:50010 10.255.62.39:50010
> 10.255.37.36:50010 10.255.33.15:50010 10.255.69.29:50010 10.255.51.13:50010
> 10.255.64.15:50010 , Current Datanode: 10.255.69.29:50010, Is current
> datanode decommissioning: true, Is current datanode entering maintenance:
> false
> 2019-10-22 15:58:51,514 DEBUG
> org.apache.hadoop.hdfs.server.blockmanagement.DatanodeAdminManager: Node
> 10.255.69.29:50010 still has 1 blocks to replicate before it is a candidate
> to finish Decommission In Progress
> {quote}
> After digging the source code and cluster log, guess it happens as follow
> steps.
> # Storage strategy is RS-6-3-1024k.
> # EC block b consists of b0, b1, b2, b3, b4, b5, b6, b7, b8, b0 is from
> datanode dn0, b1 is from datanode dn1, ...etc
> # At the beginning dn0 is in decommission progress, b0 is replicated
> successfully, and dn0 is staill in decommission progress.
> # Later b1, b2, b3 in decommission progress, and dn4 containing b4 is out of
> service, so need to reconstruct, and create ErasureCodingWork to do it, in
> the ErasureCodingWork, additionalReplRequired is 4
> # Because hasAllInternalBlocks is false, Will call
> ErasureCodingWork#addTaskToDatanode ->
> DatanodeDescriptor#addBlockToBeErasureCoded, and send
> BlockECReconstructionInfo task to Datanode
> # DataNode can not reconstruction the block because targets is 4, greater
> than 3( parity number).
> There is a problem as follow, from BlockManager.java#scheduleReconstruction
> {code}
> // should reconstruct all the internal blocks before scheduling
> // replication task for decommissioning node(s).
> if (additionalReplRequired - numReplicas.decommissioning() -
> numReplicas.liveEnteringMaintenanceReplicas() > 0) {
> additionalReplRequired = additionalReplRequired -
> numReplicas.decommissioning() -
> numReplicas.liveEnteringMaintenanceReplicas();
> }
> {code}
> Should reconstruction firstly and then replicate for decommissioning. Because
> numReplicas.decommissioning() is 4, and additionalReplRequired is 4, that's
> wrong,
> numReplicas.decommissioning() should be 3, it should exclude live replica.
> If so, additionalReplRequired will be 1, reconstruction will schedule as
> expected. After that, decommission goes on.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16960728#comment-16960728
]
guojh commented on HDFS-14768:
--
[~ferhui] If datanode is out of service, the block in it should reconstruction
first. This issue try to solve the problem that busy node should not trigger
reconstruction work if have all live block index.
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNo
[jira] [Commented] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956793#comment-16956793
]
guojh commented on HDFS-14768:
--
[~surendrasingh] Ok
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> b
[jira] [Comment Edited] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956746#comment-16956746
]
guojh edited comment on HDFS-14768 at 10/22/19 8:01 AM:
[~surendrasingh] I add a new UT to check that decommission with busy node
should not generate erasure code construction task, please review it. If you
have other ideas, please tell me.
was (Author: gjhkael):
[~surendrasingh] Do you have any ideas about the UT?
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
>
[jira] [Updated] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.008.patch
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.008.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(dat
[jira] [Commented] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956746#comment-16956746
]
guojh commented on HDFS-14768:
--
[~surendrasingh] Do you have any ideas about the UT?
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNod
[jira] [Commented] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956578#comment-16956578
]
guojh commented on HDFS-14768:
--
[~surendrasingh] Thanks.
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager
[jira] [Comment Edited] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956094#comment-16956094
]
guojh edited comment on HDFS-14768 at 10/21/19 1:44 PM:
[~surendrasingh] You can run the UT more times, the UT passed is because in the
process of decommission, the block in busyNode is correct, another block which
reconstruct is incorrect, NN find the block is redundant then will delete one,
NN may delete the correct one, may delete another one, If NN delete the
incorrect one, the UT is passed, if not, the UT can not passed. I am sorry, I
have no idea to get the incorrect block exactly。
was (Author: gjhkael):
[~surendrasingh] You can run the UT more times, the UT passed is because in the
process of decommission, the block in busyNode is correct another block
reconstruct is incorrect, redundant block will delete one, NN may delete the
correct one, may delete another one, If NN delete the incorrect one, the UT is
passed, if not, the UT can not passed. I am sorry, I have no idea to get the
incorrect block exactly。
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManage
[jira] [Commented] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956094#comment-16956094
]
guojh commented on HDFS-14768:
--
[~surendrasingh] You can run the UT more times, the UT passed is because in the
process of decommission, the block in busyNode is correct another block
reconstruct is incorrect, redundant block will delete one, NN may delete the
correct one, may delete another one, If NN delete the incorrect one, the UT is
passed, if not, the UT can not passed. I am sorry, I have no idea to get the
incorrect block exactly。
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.len
[jira] [Commented] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16955660#comment-16955660
]
guojh commented on HDFS-14768:
--
[~surendrasingh] Please review it.
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLi
[jira] [Commented] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16955446#comment-16955446
]
guojh commented on HDFS-14768:
--
rebase
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDa
[jira] [Updated] (HDFS-14768) EC : Busy DN replica should be consider in live replica check.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.007.patch
> EC : Busy DN replica should be consider in live replica check.
> --
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.007.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatano
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16954296#comment-16954296
]
guojh commented on HDFS-14768:
--
rebase trunk
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.g
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.006.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16954292#comment-16954292
]
guojh commented on HDFS-14768:
--
[~hadoopqa] test
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager(
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16954257#comment-16954257
]
guojh commented on HDFS-14768:
--
[~surendrasingh] Sorry about my thoughtless and thanks for you review.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.g
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.005.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.006.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNu
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.005.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNu
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.006.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.006.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDe
[jira] [Commented] (HDFS-14699) Erasure Coding: Storage not considered in live replica when replication streams hard limit reached to threshold
[
https://issues.apache.org/jira/browse/HDFS-14699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16954245#comment-16954245
]
guojh commented on HDFS-14699:
--
[~surendrasingh] The UT testChooseSrcDatanodesWithDupEC in class
TestBlockManager.java. The code:
{code:java}
bm.chooseSourceDatanodes(
aBlockInfoStriped,
cntNodes,
liveNodes,
numReplicas, liveBlockIndices,
liveBusyBlockIndices,
LowRedundancyBlocks.QUEUE_HIGHEST_PRIORITY);
{code}
change to
{code:java}
bm.chooseSourceDatanodes(
aBlockInfoStriped,
cntNodes,
liveNodes,
numReplicas, liveBlockIndices,
liveBusyBlockIndices,
LowRedundancyBlocks.QUEUE_VERY_LOW_REDUNDANCY);
{code}
> Erasure Coding: Storage not considered in live replica when replication
> streams hard limit reached to threshold
> ---
>
> Key: HDFS-14699
> URL: https://issues.apache.org/jira/browse/HDFS-14699
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: ec
>Affects Versions: 3.2.0, 3.1.1, 3.3.0
>Reporter: Zhao Yi Ming
>Assignee: Zhao Yi Ming
>Priority: Critical
> Labels: patch
> Fix For: 3.3.0, 3.1.4, 3.2.2
>
> Attachments: HDFS-14699.00.patch, HDFS-14699.01.patch,
> HDFS-14699.02.patch, HDFS-14699.03.patch, HDFS-14699.04.patch,
> HDFS-14699.05.patch, image-2019-08-20-19-58-51-872.png,
> image-2019-09-02-17-51-46-742.png
>
>
> We are tried the EC function on 80 node cluster with hadoop 3.1.1, we hit the
> same scenario as you said https://issues.apache.org/jira/browse/HDFS-8881.
> Following are our testing steps, hope it can helpful.(following DNs have the
> testing internal blocks)
> # we customized a new 10-2-1024k policy and use it on a path, now we have 12
> internal block(12 live block)
> # decommission one DN, after the decommission complete. now we have 13
> internal block(12 live block and 1 decommission block)
> # then shutdown one DN which did not have the same block id as 1
> decommission block, now we have 12 internal block(11 live block and 1
> decommission block)
> # after wait for about 600s (before the heart beat come) commission the
> decommissioned DN again, now we have 12 internal block(11 live block and 1
> duplicate block)
> # Then the EC is not reconstruct the missed block
> We think this is a critical issue for using the EC function in a production
> env. Could you help? Thanks a lot!
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.006.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.006.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDe
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16954202#comment-16954202
]
guojh commented on HDFS-14768:
--
[~surendrasingh] I put a new path06, please review it.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(),
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.006.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.006.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode
[jira] [Commented] (HDFS-14699) Erasure Coding: Storage not considered in live replica when replication streams hard limit reached to threshold
[
https://issues.apache.org/jira/browse/HDFS-14699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16954193#comment-16954193
]
guojh commented on HDFS-14699:
--
[~zhaoyim] [~surendrasingh] In you test case, why you set the priority to
QUEUE_HIGHEST_PRIORITY, If the priority is not equals 0, you test case is not
passed.
{code:java}
bm.chooseSourceDatanodes(
aBlockInfoStriped,
cntNodes,
liveNodes,
numReplicas, liveBlockIndices,
liveBusyBlockIndices,
LowRedundancyBlocks.QUEUE_HIGHEST_PRIORITY); // If the priority set to
QUEUE_VERY_LOW_REDUNDANCY, this will not works.
{code}
> Erasure Coding: Storage not considered in live replica when replication
> streams hard limit reached to threshold
> ---
>
> Key: HDFS-14699
> URL: https://issues.apache.org/jira/browse/HDFS-14699
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: ec
>Affects Versions: 3.2.0, 3.1.1, 3.3.0
>Reporter: Zhao Yi Ming
>Assignee: Zhao Yi Ming
>Priority: Critical
> Labels: patch
> Fix For: 3.3.0, 3.1.4, 3.2.2
>
> Attachments: HDFS-14699.00.patch, HDFS-14699.01.patch,
> HDFS-14699.02.patch, HDFS-14699.03.patch, HDFS-14699.04.patch,
> HDFS-14699.05.patch, image-2019-08-20-19-58-51-872.png,
> image-2019-09-02-17-51-46-742.png
>
>
> We are tried the EC function on 80 node cluster with hadoop 3.1.1, we hit the
> same scenario as you said https://issues.apache.org/jira/browse/HDFS-8881.
> Following are our testing steps, hope it can helpful.(following DNs have the
> testing internal blocks)
> # we customized a new 10-2-1024k policy and use it on a path, now we have 12
> internal block(12 live block)
> # decommission one DN, after the decommission complete. now we have 13
> internal block(12 live block and 1 decommission block)
> # then shutdown one DN which did not have the same block id as 1
> decommission block, now we have 12 internal block(11 live block and 1
> decommission block)
> # after wait for about 600s (before the heart beat come) commission the
> decommissioned DN again, now we have 12 internal block(11 live block and 1
> duplicate block)
> # Then the EC is not reconstruct the missed block
> We think this is a critical issue for using the EC function in a production
> env. Could you help? Thanks a lot!
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.005.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.004.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.005.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDe
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.004.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.004.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16954183#comment-16954183
]
guojh commented on HDFS-14768:
--
[~surendrasingh] Thanks for your review. I don't understand why the code that
you point out is not unnecessary? liveBusyBlockIndices should include the
condition below:
{code:java}
if (priority != LowRedundancyBlocks.QUEUE_HIGHEST_PRIORITY
&& (!node.isDecommissionInProgress() && !node.isEnteringMaintenance())
&& node.getNumberOfBlocksToBeReplicated() >= maxReplicationStreams) {
if (isStriped && state == StoredReplicaState.LIVE) {
liveBusyBlockIndices.add(blockIndex);
}
continue; // already reached replication limit
}
{code}
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageI
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16953456#comment-16953456
]
guojh commented on HDFS-14768:
--
[~surendrasingh] I put a new patch. Please review it.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getData
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.004.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.004.patch, HDFS-14768.jpg,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFil
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16953447#comment-16953447
]
guojh commented on HDFS-14768:
--
[~ayushtkn] [~surendrasingh] Could you help review the code? I can drop the
code that try to fixed the over-replicated problem. let it fixed in HDFS-14847.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> ass
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16953443#comment-16953443
]
guojh commented on HDFS-14768:
--
[~ayushtkn] Thanks for you reply, This patch try to fix two problem, one is
replica over-replicated another is block incorrect. The code quote by
[~surendrasingh] is try to fixed the over-replicated issue.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, dec
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16953287#comment-16953287
]
guojh commented on HDFS-14768:
--
[~surendrasingh] If I decommission two datanode that have the block index 6 and
8, After first replica schedule, the index 6 is success replica but index 8 is
not. Then the file with index[0, 1, 2, 3, 4, 5, 6(Decommissioning), 7
,8(Decommissioning), 6], after timeout, The BlockManager will schecule it
again, then findLeavingServiceSources will still choose index 6 to replica.
This is another problem. Should I create a new Issue and write another UT?
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeI
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.003.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, deco
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.003.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFi
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.003.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, deco
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16947422#comment-16947422
]
guojh commented on HDFS-14768:
--
[~surendrasingh] Thanks for you reply, the
incrementPendingReplicationWithoutTargets is not public, can not call it
direct. I rebase the trunk on the newest patch.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.003.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNode
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.003.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.003.patch, HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, deco
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16932196#comment-16932196
]
guojh commented on HDFS-14768:
--
[~zhaoyim] Yes, I encounter the same problem.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
>
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16931985#comment-16931985
]
guojh commented on HDFS-14768:
--
[~zhaoyim] I am confused too. [~surendrasingh] Could you help us?
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(dat
[jira] [Issue Comment Deleted] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Comment: was deleted
(was: [~zhaoyim] I am confused too. If you add break point in the check parity
like the following screenshot，you can see the parity block is exception, but if
you add break point in other place, the UT may passed, may be the block is
still change, But I don't make out. [~surendrasingh] can you help me? Thanks!
!1568771471942.jpg!)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> de
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16931965#comment-16931965
]
guojh commented on HDFS-14768:
--
[~zhaoyim] I am confused too. If you add break point in the check parity like
the following screenshot，you can see the parity block is exception, but if you
add break point in other place, the UT may passed, may be the block is still
change, But I don't make out. [~surendrasingh] can you help me? Thanks!
!1568771471942.jpg!
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[dec
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: 1568771471942.jpg
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg, 1568771471942.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
>
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16931153#comment-16931153
]
guojh commented on HDFS-14768:
--
[~zhaoyim] Try the new patch.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFil
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.002.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, HDFS-14768.002.patch,
> HDFS-14768.jpg, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure dec
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.0001.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not a
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.001.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, HDFS-14768.001.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically
[jira] [Comment Edited] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16931021#comment-16931021
]
guojh edited comment on HDFS-14768 at 9/17/19 2:34 AM:
---
[~surendrasingh] I put a new patch based on the latest code. Could you help to
review it?
was (Author: gjhkael):
[~surendrasingh] I put a new patch based on the latest code.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, HDFS-14768.0001.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(dec
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16931021#comment-16931021
]
guojh commented on HDFS-14768:
--
[~surendrasingh] I put a new patch based on the latest code.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, HDFS-14768.0001.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, dec
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.0001.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, HDFS-14768.0001.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.0001.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, HDFS-14768.0001.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatical
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.0001.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, HDFS-14768.0001.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatical
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16931008#comment-16931008
]
guojh commented on HDFS-14768:
--
[~surendrasingh] [~zhaoyim] Used the latest code in branch trunk still have the
problem. The UT I put in the Description can not passed.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescr
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16930340#comment-16930340
]
guojh commented on HDFS-14768:
--
[~zhaoyim] Before checkData I sleep 3min still have the checkData error.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNod
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16930259#comment-16930259
]
guojh commented on HDFS-14768:
--
[~zhaoyim] The index 6 in the busy DN is correct, another index 6 block is
incorrect, the index 6 block is redundancy，the NN may delete the correct one,
after that, If the DN is down that holds index 2 block, NN will reconstruct the
index 2 block bese on index 6 block, then the index 2 block is incorrect,
finally the file is corruption and is not readable.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16928962#comment-16928962
]
guojh commented on HDFS-14768:
--
[~zhaoyim] Yes, after you sleep 3 min, NN will reconstruct a new correct block
for I stop the DN on the block with index 6. But in product, because exist two
index 6 block, NN may delete the correct one, if data block is missing, then
the incorrect block will encode to generate the data block, after that, the
file is corruption and is not useable.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNo
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16928424#comment-16928424
]
guojh commented on HDFS-14768:
--
[~janick] set more numDNS, for example:
numDNs = dataBlocks + parityBlocks + 10;
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decomm
[jira] [Comment Edited] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16928348#comment-16928348
]
guojh edited comment on HDFS-14768 at 9/12/19 8:19 AM:
---
@[~surendrasingh] I update the UT in Description to reproducing the issue, So
just try it.
!1568275810244.jpg!
!1568276338275.jpg!
was (Author: gjhkael):
@[~surendrasingh] I update the UT in Description to reproducing the issue, So
just try it.
!1568275810244.jpg!
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, 1568276338275.jpg,
> HDFS-14768.000.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommission
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: 1568275810244.jpg
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, HDFS-14768.000.patch,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(c
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16928348#comment-16928348
]
guojh commented on HDFS-14768:
--
@[~surendrasingh] I update the UT in Description to reproducing the issue, So
just try it.
!1568275810244.jpg!
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: 1568275810244.jpg, HDFS-14768.000.patch,
> guojh_UT_after_deomission.txt, guojh_UT_before_deomission.txt,
> zhaoyiming_UT_after_deomission.txt, zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> private int replicationStreamsHardLimit =
> DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
> numDNs = dataBlocks + parityBlocks + 10;
> @Test(timeout = 24)
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
> .getINode4Write(ecFile.toString()).asFile();
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> BlockInfo firstBlock = fileNode.getBlocks()[0];
> DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
> // the first heartbeat will consume 3 replica tasks
> for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
> BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
> Block(i),
> new DatanodeStorageInfo[]{dStorageInfos[0]});
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
> //assertNull(checkFile(dfs, ecFile,
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16928344#comment-16928344
]
guojh commented on HDFS-14768:
--
[~zhaoyim] The key point about this issue is make one datanode busy. Then you
can reproducing this issue. Try the following code:
{code:java}
// code placeholder
private int replicationStreamsHardLimit =
DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
numDNs = dataBlocks + parityBlocks + 10;
@Test(timeout = 24)
public void testFileDecommission() throws Exception {
LOG.info("Starting test testFileDecommission");
final Path ecFile = new Path(ecDir, "testFileDecommission");
int writeBytes = cellSize * dataBlocks;
writeStripedFile(dfs, ecFile, writeBytes);
Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
.getINode4Write(ecFile.toString()).asFile();
LocatedBlocks locatedBlocks =
StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
.get(0);
DatanodeInfo[] dnLocs = lb.getLocations();
LocatedStripedBlock lastBlock =
(LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
DatanodeInfo[] storageInfos = lastBlock.getLocations();
//
DatanodeDescriptor datanodeDescriptor = cluster.getNameNode().getNamesystem()
.getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
BlockInfo firstBlock = fileNode.getBlocks()[0];
DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
// the first heartbeat will consume 3 replica tasks
for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
Block(i),
new DatanodeStorageInfo[]{dStorageInfos[0]});
}
assertEquals(dataBlocks + parityBlocks, dnLocs.length);
int[] decommNodeIndex = {3, 4};
final List decommisionNodes = new ArrayList();
// add the node which will be decommissioning
decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
//assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
// Ensure decommissioned datanode is not automatically shutdown
DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
assertEquals("All datanodes must be alive", numDNs,
client.datanodeReport(DatanodeReportType.LIVE).length);
FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
Assert.assertTrue("Checksum mismatches!",
fileChecksum1.equals(fileChecksum2));
StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
null, blockGroupSize);
}
{code}
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch, guojh_UT_after_deomission.txt,
> guojh_UT_before_deomission.txt, zhaoyiming_UT_after_deomission.txt,
> zhaoyiming_UT_beofre_deomission.txt
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
>
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Description:
Policy is RS-6-3-1024K, version is hadoop 3.0.2;
We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
index[3,4], increase the index 6 datanode's
pendingReplicationWithoutTargets that make it large than
replicationStreamsHardLimit(we set 14). Then, After the method
chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
[0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
In method scheduleReconstruction of BlockManager, the additionalReplRequired is
9 - 7 = 2. After Namenode choose two target Datanode, will assign a erasureCode
task to target datanode.
When datanode get the task will build targetIndices from liveBlockIndices and
target length. the code is blow.
{code:java}
// code placeholder
targetIndices = new short[targets.length];
private void initTargetIndices() {
BitSet bitset = reconstructor.getLiveBitSet();
int m = 0; hasValidTargets = false;
for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
if (!bitset.get) {
if (reconstructor.getBlockLen > 0) {
if (m < targets.length) {
targetIndices[m++] = (short)i;
hasValidTargets = true;
}
}
}
}
{code}
targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
The StripedReader is aways create reader from first 6 index block, and is
[0,1,2,3,4,5]
Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
bug. the block index6's data is corruption(all data is zero).
I write a unit test can stabilize repreduce.
{code:java}
// code placeholder
private int replicationStreamsHardLimit =
DFSConfigKeys.DFS_NAMENODE_REPLICATION_STREAMS_HARD_LIMIT_DEFAULT;
numDNs = dataBlocks + parityBlocks + 10;
@Test(timeout = 24)
public void testFileDecommission() throws Exception {
LOG.info("Starting test testFileDecommission");
final Path ecFile = new Path(ecDir, "testFileDecommission");
int writeBytes = cellSize * dataBlocks;
writeStripedFile(dfs, ecFile, writeBytes);
Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
final INodeFile fileNode = cluster.getNamesystem().getFSDirectory()
.getINode4Write(ecFile.toString()).asFile();
LocatedBlocks locatedBlocks =
StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
.get(0);
DatanodeInfo[] dnLocs = lb.getLocations();
LocatedStripedBlock lastBlock =
(LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
DatanodeInfo[] storageInfos = lastBlock.getLocations();
//
DatanodeDescriptor datanodeDescriptor = cluster.getNameNode().getNamesystem()
.getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
BlockInfo firstBlock = fileNode.getBlocks()[0];
DatanodeStorageInfo[] dStorageInfos = bm.getStorages(firstBlock);
// the first heartbeat will consume 3 replica tasks
for (int i = 0; i <= replicationStreamsHardLimit + 3; i++) {
BlockManagerTestUtil.addBlockToBeReplicated(datanodeDescriptor, new
Block(i),
new DatanodeStorageInfo[]{dStorageInfos[0]});
}
assertEquals(dataBlocks + parityBlocks, dnLocs.length);
int[] decommNodeIndex = {3, 4};
final List decommisionNodes = new ArrayList();
// add the node which will be decommissioning
decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
bm.getDatanodeManager().removeDatanode(datanodeDescriptor);
//assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
// Ensure decommissioned datanode is not automatically shutdown
DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
assertEquals("All datanodes must be alive", numDNs,
client.datanodeReport(DatanodeReportType.LIVE).length);
FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
Assert.assertTrue("Checksum mismatches!",
fileChecksum1.equals(fileChecksum2));
StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
null, blockGroupSize);
}
{code}
was:
Policy is RS-6-3-1024K, version is hadoop 3.0.2;
We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
index[3,4], increase the index 6 datanode's
pendingReplicationWithoutTargets that make it large than
replicationStreamsHardLimit(we set 14). Then, After the method
chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
[0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
In method scheduleReconstruction of BlockManager, the additionalReplRequired is
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16928146#comment-16928146
]
guojh commented on HDFS-14768:
--
[~zhaoyim] My UT is not good, you need to check the block manually, I will
update a new UT to check the block corruption automatic.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16928145#comment-16928145
]
guojh commented on HDFS-14768:
--
[~surendrasingh] Sorry about my UT. I will update a new UT to check the block
corruption.
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.
[jira] [Comment Edited] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16925298#comment-16925298
]
guojh edited comment on HDFS-14768 at 9/9/19 2:16 AM:
--
[~surendrasingh] Thanks for you replay. you need run the code below, and you
need check the block index 6 that recontruct on local path like
'./hadoop/hadoop-hdfs-project/hadoop-hdfs/target/test/data/dfs/*/current/BP-xxx/current/finalized/*/*/'
{code:java}
// code placeholder
public void testFileDecommission() throws Exception {
LOG.info("Starting test testFileDecommission");
final Path ecFile = new Path(ecDir, "testFileDecommission");
int writeBytes = cellSize * dataBlocks;
writeStripedFile(dfs, ecFile, writeBytes); Assert.assertEquals(0,
bm.numOfUnderReplicatedBlocks());
FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
LocatedBlocks locatedBlocks = StripedFileTestUtil.getLocatedBlocks(ecFile,
dfs);
LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
.get(0);DatanodeInfo[] dnLocs = lb.getLocations();
LocatedStripedBlock lastBlock =
(LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
DatanodeInfo[] storageInfos = lastBlock.getLocations();
DatanodeDescriptor datanodeDescriptor = cluster.getNameNode().getNamesystem()
.getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
for (int i = 0; i < 100; i++) {
datanodeDescriptor.incrementPendingReplicationWithoutTargets();
}
assertEquals(dataBlocks + parityBlocks, dnLocs.length);
int[] decommNodeIndex = {3, 4};
final List decommisionNodes = new ArrayList decommisionNodes = new
ArrayList In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeU
[jira] [Comment Edited] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16925298#comment-16925298
]
guojh edited comment on HDFS-14768 at 9/9/19 2:14 AM:
--
[~surendrasingh] Thanks for you replay. you need run the code below, and you
need check the block index 6 that recontruct on local path like
'./hadoop/hadoop-hdfs-project/hadoop-hdfs/target/test/data/dfs/*/current/BP-xxx/current/finalized/*/*/'
{code:java}
// code placeholder
public void testFileDecommission() throws Exception { LOG.info("Starting test
testFileDecommission"); final Path ecFile = new Path(ecDir,
"testFileDecommission"); int writeBytes = cellSize * dataBlocks;
writeStripedFile(dfs, ecFile, writeBytes); Assert.assertEquals(0,
bm.numOfUnderReplicatedBlocks()); FileChecksum fileChecksum1 =
dfs.getFileChecksum(ecFile, writeBytes); LocatedBlocks locatedBlocks =
StripedFileTestUtil.getLocatedBlocks(ecFile, dfs); LocatedBlock lb =
dfs.getClient().getLocatedBlocks(ecFile.toString(), 0) .get(0); DatanodeInfo[]
dnLocs = lb.getLocations(); LocatedStripedBlock lastBlock =
(LocatedStripedBlock)locatedBlocks.getLastLocatedBlock(); DatanodeInfo[]
storageInfos = lastBlock.getLocations(); DatanodeDescriptor datanodeDescriptor
= cluster.getNameNode().getNamesystem()
.getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
for (int i = 0; i < 100; i++) {
datanodeDescriptor.incrementPendingReplicationWithoutTargets(); }
assertEquals(dataBlocks + parityBlocks, dnLocs.length); int[] decommNodeIndex =
{3, 4}; final List decommisionNodes = new
ArrayList In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.
[jira] [Comment Edited] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16925298#comment-16925298
]
guojh edited comment on HDFS-14768 at 9/9/19 2:12 AM:
--
[~surendrasingh] Thanks for you replay. you need run the code below, and you
need check the block index 6 that recontruct on local path like
'./hadoop/hadoop-hdfs-project/hadoop-hdfs/target/test/data/dfs/*/current/BP-xxx/current/finalized/*/*/'
{code:java}
// code placeholder
public void testFileDecommission() throws Exception {
LOG.info("Starting test testFileDecommission");
final Path ecFile = new Path(ecDir, "testFileDecommission");
int writeBytes = cellSize * dataBlocks;
writeStripedFile(dfs, ecFile, writeBytes);
Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
LocatedBlocks locatedBlocks = StripedFileTestUtil.getLocatedBlocks(ecFile,
dfs);
LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
.get(0); DatanodeInfo[] dnLocs = lb.getLocations();
LocatedStripedBlock lastBlock =
(LocatedStripedBlock)locatedBlocks.getLastLocatedBlock(); DatanodeInfo[]
storageInfos = lastBlock.getLocations();
DatanodeDescriptor datanodeDescriptor = cluster.getNameNode().getNamesystem()
.getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
for (int i = 0; i < 100; i++) {
datanodeDescriptor.incrementPendingReplicationWithoutTargets(); }
assertEquals(dataBlocks + parityBlocks, dnLocs.length); int[] decommNodeIndex =
{3, 4}; final List decommisionNodes = new
ArrayList(); // add the node which will be decommissioning
decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
decommisionNodes.add(dnLocs[decommNodeIndex[1]]); decommissionNode(0,
decommisionNodes, AdminStates.DECOMMISSIONED);
assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
//assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs)); // Ensure
decommissioned datanode is not automatically shutdown DFSClient client =
getDfsClient(cluster.getNameNode(0), conf); assertEquals("All datanodes must be
alive", numDNs, client.datanodeReport(DatanodeReportType.LIVE).length);
FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
Assert.assertTrue("Checksum mismatches!", fileChecksum1.equals(fileChecksum2));
StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes, null,
blockGroupSize); }
{code}
was (Author: gjhkael):
[~surendrasingh] Thanks for you replay. you need run the code below, and you
need check the block index 6 that recontruct on local path like
'./hadoop/hadoop-hdfs-project/hadoop-hdfs/target/test/data/dfs/*/current/BP-xxx/current/finalized/*/*/'
{code:java}
// code placeholder
public void testFileDecommission() throws Exception { LOG.info("Starting test
testFileDecommission"); final Path ecFile = new Path(ecDir,
"testFileDecommission"); int writeBytes = cellSize * dataBlocks;
writeStripedFile(dfs, ecFile, writeBytes); Assert.assertEquals(0,
bm.numOfUnderReplicatedBlocks()); FileChecksum fileChecksum1 =
dfs.getFileChecksum(ecFile, writeBytes); LocatedBlocks locatedBlocks =
StripedFileTestUtil.getLocatedBlocks(ecFile, dfs); LocatedBlock lb =
dfs.getClient().getLocatedBlocks(ecFile.toString(), 0) .get(0); DatanodeInfo[]
dnLocs = lb.getLocations(); LocatedStripedBlock lastBlock =
(LocatedStripedBlock)locatedBlocks.getLastLocatedBlock(); DatanodeInfo[]
storageInfos = lastBlock.getLocations(); // DatanodeDescriptor
datanodeDescriptor = cluster.getNameNode().getNamesystem()
.getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
for (int i = 0; i < 100; i++) {
datanodeDescriptor.incrementPendingReplicationWithoutTargets(); }
assertEquals(dataBlocks + parityBlocks, dnLocs.length); int[] decommNodeIndex =
{3, 4}; final List decommisionNodes = new
ArrayList(); // add the node which will be decommissioning
decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
decommisionNodes.add(dnLocs[decommNodeIndex[1]]); decommissionNode(0,
decommisionNodes, AdminStates.DECOMMISSIONED);
assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
//assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs)); // Ensure
decommissioned datanode is not automatically shutdown DFSClient client =
getDfsClient(cluster.getNameNode(0), conf); assertEquals("All datanodes must be
alive", numDNs, client.datanodeReport(DatanodeReportType.LIVE).length);
FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
Assert.assertTrue("Checksum mismatches!", fileChecksum1.equals(fileChecksum2));
StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes, null,
blockGroupSize); }
{code}
> In some cases, erasure blocks are corruption when they are reconstruct.
> -
[jira] [Comment Edited] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16925298#comment-16925298
]
guojh edited comment on HDFS-14768 at 9/9/19 2:12 AM:
--
[~surendrasingh] Thanks for you replay. you need run the code below, and you
need check the block index 6 that recontruct on local path like
'./hadoop/hadoop-hdfs-project/hadoop-hdfs/target/test/data/dfs/*/current/BP-xxx/current/finalized/*/*/'
{code:java}
// code placeholder
{code}
was (Author: gjhkael):
[~surendrasingh] Thanks for you replay. you need run the code below, and you
need check the block index 6 that recontruct on local path like
'./hadoop/hadoop-hdfs-project/hadoop-hdfs/target/test/data/dfs/*/current/BP-xxx/current/finalized/*/*/'
{code:java}
// code placeholder
public void testFileDecommission() throws Exception {
LOG.info("Starting test testFileDecommission");
final Path ecFile = new Path(ecDir, "testFileDecommission");
int writeBytes = cellSize * dataBlocks;
writeStripedFile(dfs, ecFile, writeBytes);
Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
LocatedBlocks locatedBlocks = StripedFileTestUtil.getLocatedBlocks(ecFile,
dfs);
LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
.get(0); DatanodeInfo[] dnLocs = lb.getLocations();
LocatedStripedBlock lastBlock =
(LocatedStripedBlock)locatedBlocks.getLastLocatedBlock(); DatanodeInfo[]
storageInfos = lastBlock.getLocations();
DatanodeDescriptor datanodeDescriptor = cluster.getNameNode().getNamesystem()
.getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
for (int i = 0; i < 100; i++) {
datanodeDescriptor.incrementPendingReplicationWithoutTargets(); }
assertEquals(dataBlocks + parityBlocks, dnLocs.length); int[] decommNodeIndex =
{3, 4}; final List decommisionNodes = new
ArrayList(); // add the node which will be decommissioning
decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
decommisionNodes.add(dnLocs[decommNodeIndex[1]]); decommissionNode(0,
decommisionNodes, AdminStates.DECOMMISSIONED);
assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
//assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs)); // Ensure
decommissioned datanode is not automatically shutdown DFSClient client =
getDfsClient(cluster.getNameNode(0), conf); assertEquals("All datanodes must be
alive", numDNs, client.datanodeReport(DatanodeReportType.LIVE).length);
FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
Assert.assertTrue("Checksum mismatches!", fileChecksum1.equals(fileChecksum2));
StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes, null,
blockGroupSize); }
{code}
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16925298#comment-16925298
]
guojh commented on HDFS-14768:
--
[~surendrasingh] Thanks for you replay. you need run the code below, and you
need check the block index 6 that recontruct on local path like
'./hadoop/hadoop-hdfs-project/hadoop-hdfs/target/test/data/dfs/*/current/BP-xxx/current/finalized/*/*/'
{code:java}
// code placeholder
public void testFileDecommission() throws Exception { LOG.info("Starting test
testFileDecommission"); final Path ecFile = new Path(ecDir,
"testFileDecommission"); int writeBytes = cellSize * dataBlocks;
writeStripedFile(dfs, ecFile, writeBytes); Assert.assertEquals(0,
bm.numOfUnderReplicatedBlocks()); FileChecksum fileChecksum1 =
dfs.getFileChecksum(ecFile, writeBytes); LocatedBlocks locatedBlocks =
StripedFileTestUtil.getLocatedBlocks(ecFile, dfs); LocatedBlock lb =
dfs.getClient().getLocatedBlocks(ecFile.toString(), 0) .get(0); DatanodeInfo[]
dnLocs = lb.getLocations(); LocatedStripedBlock lastBlock =
(LocatedStripedBlock)locatedBlocks.getLastLocatedBlock(); DatanodeInfo[]
storageInfos = lastBlock.getLocations(); // DatanodeDescriptor
datanodeDescriptor = cluster.getNameNode().getNamesystem()
.getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
for (int i = 0; i < 100; i++) {
datanodeDescriptor.incrementPendingReplicationWithoutTargets(); }
assertEquals(dataBlocks + parityBlocks, dnLocs.length); int[] decommNodeIndex =
{3, 4}; final List decommisionNodes = new
ArrayList(); // add the node which will be decommissioning
decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
decommisionNodes.add(dnLocs[decommNodeIndex[1]]); decommissionNode(0,
decommisionNodes, AdminStates.DECOMMISSIONED);
assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
//assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs)); // Ensure
decommissioned datanode is not automatically shutdown DFSClient client =
getDfsClient(cluster.getNameNode(0), conf); assertEquals("All datanodes must be
alive", numDNs, client.datanodeReport(DatanodeReportType.LIVE).length);
FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
Assert.assertTrue("Checksum mismatches!", fileChecksum1.equals(fileChecksum2));
StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes, null,
blockGroupSize); }
{code}
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Assignee: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeS
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.000.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.001.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.000.patch)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.001.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.001.patch
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch, HDFS-14768.001.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are reconstruct.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Summary: In some cases, erasure blocks are corruption when they are
reconstruct. (was: In some cases, erasure blocks are corruption when they are
rebuilt.)
> In some cases, erasure blocks are corruption when they are reconstruct.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are rebuilt.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16915395#comment-16915395
]
guojh commented on HDFS-14768:
--
[~drankye] Can we talk about this issue?
> In some cases, erasure blocks are corruption when they are rebuilt.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are rebuilt.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: (was: HDFS-14768.000.patch)
> In some cases, erasure blocks are corruption when they are rebuilt.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are rebuilt.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.000.patch
Fix Version/s: 3.3.0
Target Version/s: 3.3.0
Labels: patch (was: )
Status: Patch Available (was: Open)
fixed the block swell and build bug
> In some cases, erasure blocks are corruption when they are rebuilt.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
> Labels: patch
> Fix For: 3.3.0
>
> Attachments: HDFS-14768.000.patch, HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.o
[jira] [Updated] (HDFS-14768) In some cases, erasure blocks are corruption when they are rebuilt.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
guojh updated HDFS-14768:
-
Attachment: HDFS-14768.000.patch
> In some cases, erasure blocks are corruption when they are rebuilt.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
> Attachments: HDFS-14768.000.patch
>
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are rebuilt.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16913961#comment-16913961
]
guojh commented on HDFS-14768:
--
[~ayushtkn] yes, we have solved the bug. But I still need your advice for the
solution.
> In some cases, erasure blocks are corruption when they are rebuilt.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14768) In some cases, erasure blocks are corruption when they are rebuilt.
[
https://issues.apache.org/jira/browse/HDFS-14768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16913954#comment-16913954
]
guojh commented on HDFS-14768:
--
[~hemanthboyina] Thanks for you attention. I post the detail Description above.
The bug Damaged our production data, I think it is terrible.
> In some cases, erasure blocks are corruption when they are rebuilt.
>
>
> Key: HDFS-14768
> URL: https://issues.apache.org/jira/browse/HDFS-14768
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, erasure-coding, hdfs, namenode
>Affects Versions: 3.0.2
>Reporter: guojh
>Priority: Major
>
> Policy is RS-6-3-1024K, version is hadoop 3.0.2;
> We suppose a file's block Index is [0,1,2,3,4,5,6,7,8], And decommission
> index[3,4], increase the index 6 datanode's
> pendingReplicationWithoutTargets that make it large than
> replicationStreamsHardLimit(we set 14). Then, After the method
> chooseSourceDatanodes of BlockMananger, the liveBlockIndices is
> [0,1,2,3,4,5,7,8], Block Counter is, Live:7, Decommission:2.
> In method scheduleReconstruction of BlockManager, the additionalReplRequired
> is 9 - 7 = 2. After Namenode choose two target Datanode, will assign a
> erasureCode task to target datanode.
> When datanode get the task will build targetIndices from liveBlockIndices
> and target length. the code is blow.
> {code:java}
> // code placeholder
> targetIndices = new short[targets.length];
> private void initTargetIndices() {
> BitSet bitset = reconstructor.getLiveBitSet();
> int m = 0; hasValidTargets = false;
> for (int i = 0; i < dataBlkNum + parityBlkNum; i++) {
> if (!bitset.get) {
> if (reconstructor.getBlockLen > 0) {
> if (m < targets.length) {
> targetIndices[m++] = (short)i;
> hasValidTargets = true;
> }
> }
> }
> }
> {code}
> targetIndices[0]=6, and targetIndices[1] is aways 0 from initial value.
> The StripedReader is aways create reader from first 6 index block, and is
> [0,1,2,3,4,5]
> Use the index [0,1,2,3,4,5] to build target index[6,0] will trigger the isal
> bug. the block index6's data is corruption(all data is zero).
> I write a unit test can stabilize repreduce.
> {code:java}
> // code placeholder
> public void testFileDecommission() throws Exception {
> LOG.info("Starting test testFileDecommission");
> final Path ecFile = new Path(ecDir, "testFileDecommission");
> int writeBytes = cellSize * dataBlocks;
> writeStripedFile(dfs, ecFile, writeBytes);
> Assert.assertEquals(0, bm.numOfUnderReplicatedBlocks());
> FileChecksum fileChecksum1 = dfs.getFileChecksum(ecFile, writeBytes);
> LocatedBlocks locatedBlocks =
> StripedFileTestUtil.getLocatedBlocks(ecFile, dfs);
> LocatedBlock lb = dfs.getClient().getLocatedBlocks(ecFile.toString(), 0)
> .get(0);
> DatanodeInfo[] dnLocs = lb.getLocations();
> LocatedStripedBlock lastBlock =
> (LocatedStripedBlock)locatedBlocks.getLastLocatedBlock();
> DatanodeInfo[] storageInfos = lastBlock.getLocations();
> //
> DatanodeDescriptor datanodeDescriptor =
> cluster.getNameNode().getNamesystem()
>
> .getBlockManager().getDatanodeManager().getDatanode(storageInfos[6].getDatanodeUuid());
> for (int i = 0; i < 100; i++) {
> datanodeDescriptor.incrementPendingReplicationWithoutTargets();
> }
> assertEquals(dataBlocks + parityBlocks, dnLocs.length);
> int[] decommNodeIndex = {3, 4};
> final List decommisionNodes = new ArrayList();
> // add the node which will be decommissioning
> decommisionNodes.add(dnLocs[decommNodeIndex[0]]);
> decommisionNodes.add(dnLocs[decommNodeIndex[1]]);
> decommissionNode(0, decommisionNodes, AdminStates.DECOMMISSIONED);
> assertEquals(decommisionNodes.size(), fsn.getNumDecomLiveDataNodes());
> //assertNull(checkFile(dfs, ecFile, 9, decommisionNodes, numDNs));
> // Ensure decommissioned datanode is not automatically shutdown
> DFSClient client = getDfsClient(cluster.getNameNode(0), conf);
> assertEquals("All datanodes must be alive", numDNs,
> client.datanodeReport(DatanodeReportType.LIVE).length);
> FileChecksum fileChecksum2 = dfs.getFileChecksum(ecFile, writeBytes);
> Assert.assertTrue("Checksum mismatches!",
> fileChecksum1.equals(fileChecksum2));
> StripedFileTestUtil.checkData(dfs, ecFile, writeBytes, decommisionNodes,
> null, blockGroupSize);
> }
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
