[
https://issues.apache.org/jira/browse/HDFS-4887?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13703809#comment-13703809
]
Jing Zhao commented on HDFS-4887:
---------------------------------
bq. The replication monitor thread accesses this variable only once when
terminating, so there will be no issue.
You're right. The thread setting checkNSRunning to false also interrupts the
replication monitor thread, thus the happen-before relationship between them
stands.
+1 for the patch.
> TestNNThroughputBenchmark exits abruptly
> ----------------------------------------
>
> Key: HDFS-4887
> URL: https://issues.apache.org/jira/browse/HDFS-4887
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: benchmarks, test
> Affects Versions: 3.0.0, 2.1.0-beta
> Reporter: Kihwal Lee
> Assignee: Kihwal Lee
> Priority: Blocker
> Attachments: HDFS-4887.patch, HDFS-4887.patch
>
>
> After HDFS-4840, TestNNThroughputBenchmark exits in the middle. This is
> because ReplicationMonitor is being stopped while NN is still running.
> This is only valid during testing. In normal cases, ReplicationMonitor thread
> runs all the time once started. In standby or safemode, it just skips
> calculating DN work. I think NNThroughputBenchmark needs to use ExitUtil to
> prevent termination, rather than modifying ReplicationMonitor.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
