[
https://issues.apache.org/jira/browse/HBASE-18771?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16159571#comment-16159571
]
Ashu Pachauri commented on HBASE-18771:
---------------------------------------
Nice find! The patch looks generally good. There is one nitpick in the tests
though:
{code:java}
+ // Split at this point should not result in the RS being aborted
+
assertEquals(util.getMiniHBaseCluster().getLiveRegionServerThreads().size(), 3);
{code}
The RS abort may not happen every time a split fails depending on where in the
process the split transaction failed (or it rolled back). Also, the RS may come
back up after aborting at the exact same time when this check happens and abort
again. A more robust check is on the number of online regions for the table,
which should be exactly one more than the previously online regions.
> Incorrect StoreFileRefresh leading to split and compaction failures
> -------------------------------------------------------------------
>
> Key: HBASE-18771
> URL: https://issues.apache.org/jira/browse/HBASE-18771
> Project: HBase
> Issue Type: Bug
> Affects Versions: 1.3.1
> Reporter: Abhishek Singh Chouhan
> Assignee: Abhishek Singh Chouhan
> Priority: Blocker
> Fix For: 1.4.0, 1.3.2, 1.5.0
>
> Attachments: HBASE-18771.branch-1.3.001.patch,
> HBASE-18771.branch-1.3.002.patch, HBASE-18771.branch-1.3.003.patch,
> HBASE-18771.master.001.patch
>
>
> We ran into issues of compaction and split failures with 1.3 similar to
> HBASE-18186 and HBASE-17406. Here's what i believe is happening -
> Lets say we have 4 store files that are compacted to form a new one. At this
> point we now have 5 store files, however only 1(the newly formed) is open now
> for the store and rest are waiting to get archived by HFileArchiver
> Now before the files are archived we get a FNFE in a scanner. This results in
> HRegion.RegionScannerImpl.handleFileNotFound(FileNotFoundException fnfe)
> being called which results in region.refreshStoreFiles(true) ->
> HStore.refreshStoreFiles()
> HStore.refreshStoreFiles now checks the hdfs dir and adds the previously
> compacted files back to the store, however these files are also present in
> StoreFileManager's compactedFiles list. Now at this point HFileArchiver runs,
> checks compactedFiles list and moves these files into the archive directory.
> Now when compaction runs it gets:
> 2017-09-04 12:30:13,899 ERROR [ctions-1504505399609]
> regionserver.CompactSplitThread - Compaction selection failed regionName =
> xxxx, storeName = 0, priority = 26, time = 1504528213899
> java.io.FileNotFoundException: File does not exist: hdfs://xxxx
> at
> org.apache.hadoop.hdfs.DistributedFileSystem$23.doCall(DistributedFileSystem.java:1337)
> at
> org.apache.hadoop.hdfs.DistributedFileSystem$23.doCall(DistributedFileSystem.java:1329)
> at
> org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:81)
> at
> org.apache.hadoop.hdfs.DistributedFileSystem.getFileStatus(DistributedFileSystem.java:1329)
> at
> org.apache.hadoop.fs.FilterFileSystem.getFileStatus(FilterFileSystem.java:422)
> at
> org.apache.hadoop.hbase.regionserver.StoreFileInfo.getReferencedFileStatus(StoreFileInfo.java:342)
> at
> org.apache.hadoop.hbase.regionserver.StoreFileInfo.getFileStatus(StoreFileInfo.java:355)
> at
> org.apache.hadoop.hbase.regionserver.StoreFileInfo.getModificationTime(StoreFileInfo.java:360)
> at
> org.apache.hadoop.hbase.regionserver.StoreFile.getModificationTimeStamp(StoreFile.java:325)
> at
> org.apache.hadoop.hbase.regionserver.StoreUtils.getLowestTimestamp(StoreUtils.java:63)
> at
> org.apache.hadoop.hbase.regionserver.compactions.RatioBasedCompactionPolicy.shouldPerformMajorCompaction(RatioBasedCompactionPolicy.java:65)
> at
> org.apache.hadoop.hbase.regionserver.compactions.SortedCompactionPolicy.selectCompaction(SortedCompactionPolicy.java:82)
> at
> org.apache.hadoop.hbase.regionserver.DefaultStoreEngine$DefaultCompactionContext.select(DefaultStoreEngine.java:107)
> at
> org.apache.hadoop.hbase.regionserver.HStore.requestCompaction(HStore.java:1679)
> Similarly if a split happens after archival we fail after PONR while opening
> daughter regions due to FNFE. This results in parent offline and daughters
> also in a limbo since they're unable to open. Since we get the error after
> PONR we also end up aborting the RS.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
