[
https://issues.apache.org/jira/browse/HDFS-16994?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17724746#comment-17724746
]
ASF GitHub Bot commented on HDFS-16994:
---------------------------------------
hfutatzhanghb commented on PR #5600:
URL: https://github.com/apache/hadoop/pull/5600#issuecomment-1556462000
> First it is a behaviour change, hence incompatible.
>
> Second, I don't find any reason we shouldn't accumulate, it is just the
metrics telling since last failover/restart how many blocks timedout in pending
reconstruction stage. No relation of getNumTimedOut, there can be multiple
consumers as well....
@ayushtkn , Thanks sir, agree with you. I will close this PR.
> NumTimedOutPendingReconstructions metrics should not be accumulated
> -------------------------------------------------------------------
>
> Key: HDFS-16994
> URL: https://issues.apache.org/jira/browse/HDFS-16994
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: farmmamba
> Priority: Minor
> Labels: pull-request-available
>
> For now, the NumTimedOutPendingReconstructions metric is computed by below
> statement:
> {code:java}
> timedOutCount + timedOutItems.size(); {code}
> Therefore, its value would always be accumulated unless namenode failover
> happens.
> In fact, the NumTimedOutPendingReconstructions metric should not be
> accumulated, So, we should set it to zero after we execute getNumTimedOuts
> method.
> The UT TestPendingReconstruction#testPendingReconstruction was passed.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
