[
https://issues.apache.org/jira/browse/HDFS-12578?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16202827#comment-16202827
]
Xiao Chen commented on HDFS-12578:
----------------------------------
Thank you for the investigation [~ajayydv]. And sorry for my delayed response.
I think the reason there is a 1+ second delay in branch-2.7, is in
{{BlockManagerTestUtil.checkHeartbeat}}:
{code:title=HDFS-11224's change, which is in 2.8+}
public static void checkHeartbeat(BlockManager bm) {
- bm.getDatanodeManager().getHeartbeatManager().heartbeatCheck();
+ HeartbeatManager hbm = bm.getDatanodeManager().getHeartbeatManager();
+ hbm.restartHeartbeatStopWatch();
+ hbm.heartbeatCheck();
}
{code}
The jira HDFS-11224 was mainly to fix a bug in feature HDFS-9239, which is only
in branch-2.8+.
So here is what I propose:
- for branch-2.7, in addition to what you have found, we should also restart
the stopwatch.
- for branch-2.8+, let's do what you did in the patch, to give a wider range
than 1 millisecond. (Please upload a patch based on trunk, so pre-commit can be
triggered)
Does this make sense?
> TestDeadDatanode#testNonDFSUsedONDeadNodeReReg failing in branch-2.7
> --------------------------------------------------------------------
>
> Key: HDFS-12578
> URL: https://issues.apache.org/jira/browse/HDFS-12578
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: test
> Reporter: Xiao Chen
> Assignee: Ajay Kumar
> Priority: Blocker
> Attachments: HDFS-12578-branch-2.7.001.patch
>
>
> It appears {{TestDeadDatanode#testNonDFSUsedONDeadNodeReReg}} is consistently
> failing in branch-2.7. We should investigate and fix it.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
