[
https://issues.apache.org/jira/browse/HDFS-16999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17720385#comment-17720385
]
ASF GitHub Bot commented on HDFS-16999:
---------------------------------------
ayushtkn commented on code in PR #5622:
URL: https://github.com/apache/hadoop/pull/5622#discussion_r1187022192
##########
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/blockmanagement/TestBlockManager.java:
##########
@@ -2066,4 +2069,55 @@ public void
testValidateReconstructionWorkAndRacksNotEnough() {
// validateReconstructionWork return false, need to perform resetTargets().
assertNull(work.getTargets());
}
+
+ /**
+ * Test whether the first block report after DataNode restart is completely
+ * processed.
+ */
+ @Test
+ public void testBlockReportAfterDataNodeRestart() throws Exception {
+ Configuration conf = new HdfsConfiguration();
+ try (MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf)
+ .numDataNodes(3).storagesPerDatanode(1).build()) {
+ cluster.waitActive();
+ BlockManager blockManager = cluster.getNamesystem().getBlockManager();
+ DistributedFileSystem fs = cluster.getFileSystem();
+ final Path filePath = new Path("/tmp.txt");
+ final long fileLen = 1L;
+ DFSTestUtil.createFile(fs, filePath, fileLen, (short) 3, 1L);
+ DFSTestUtil.waitForReplication(fs, filePath, (short) 3, 60000);
+ ArrayList<DataNode> datanodes = cluster.getDataNodes();
+ assertEquals(datanodes.size(), 3);
+
+ // Stop RedundancyMonitor.
+ blockManager.setInitializedReplQueues(false);
+
+ // Delete the replica on the first datanode.
+ File dnDir =
+ datanodes.get(0).getFSDataset().getVolumeList().get(0)
+ .getCurrentDir();
+ String[] children = FileUtil.list(dnDir);
+ for (String s : children) {
+ if (!s.equals("VERSION")) {
+ FileUtil.fullyDeleteContents(new File(dnDir, s));
+ }
+ }
+
+ // The number of replicas is still 3 because the datanode has not sent
+ // a new block report.
+ FileStatus stat = fs.getFileStatus(filePath);
+ BlockLocation[] locs = fs.getFileBlockLocations(stat, 0, stat.getLen());
+ assertEquals(3, locs[0].getHosts().length);
+
+ // Restart the first datanode.
+ cluster.restartDataNode(0, true);
+
+ // Wait for the block report to be processed.
+ Thread.sleep(5000);
Review Comment:
can you use specific wait methods or at worst some GenericTestUtils.waitFor
instead of a specific sleep. MiniDfsCluster has some
``waitDataNodeFullyStarted`` or ``waitFirstBRCompleted``, check if you can use
that.
> Fix wrong use of processFirstBlockReport()
> ------------------------------------------
>
> Key: HDFS-16999
> URL: https://issues.apache.org/jira/browse/HDFS-16999
> Project: Hadoop HDFS
> Issue Type: Bug
> Reporter: Shuyan Zhang
> Assignee: Shuyan Zhang
> Priority: Major
> Labels: pull-request-available
>
> `processFirstBlockReport()` is used to process first block report from
> datanode. It does not calculating `toRemove` list because it believes that
> there is no metadata about the datanode in the namenode. However, If a
> datanode is re registered after restarting, its `blockReportCount` will be
> updated to 0. That is to say, the first block report after a datanode
> restarts will be processed by `processFirstBlockReport()`. This is
> unreasonable because the metadata of the datanode already exists in namenode
> at this time, and if redundant replica metadata is not removed in time, the
> blocks with insufficient replicas cannot be reconstruct in time, which
> increases the risk of missing block. In summary, `processFirstBlockReport()`
> should only be used when the namenode restarts, not when the datanode
> restarts.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
