[
https://issues.apache.org/jira/browse/HDFS-13908?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16620664#comment-16620664
]
Ayush Saxena commented on HDFS-13908:
-------------------------------------
Thanx [~elgoiri] for the suggestion.
One doubt regarding the suggestion.As far as I think the purpose here seems to
be like after the format with different ID when the datanode is stopped and
started again we need to ensure it doesn't come up.It was done by making the
thread wait for 10 seconds and then check that after 10 seconds also it didn't
come up and conclude that since it didn't come up in this time frame it won't
come up ultimately.But after the change I guess we will conclude after 500 msec
only immediately after the start command is triggered .The scenario that we are
thinking to change is something like DN was up and we are waiting for it to get
up.But actually it is down only and we are making sure it won't come up
either.Correct me if If I understood it wrong.
> TestDataNodeMultipleRegistrations is flaky
> ------------------------------------------
>
> Key: HDFS-13908
> URL: https://issues.apache.org/jira/browse/HDFS-13908
> Project: Hadoop HDFS
> Issue Type: New Feature
> Reporter: Íñigo Goiri
> Assignee: Ayush Saxena
> Priority: Major
> Attachments: Above Timeout.rar, HDFS-13908-01.patch,
> HDFS-13908-02.patch, HDFS-13908-03.patch, HDFS-13908-04.patch, Within
> TImeout.rar
>
>
> We have seen this issue in multiple runs:
> https://builds.apache.org/job/PreCommit-HADOOP-Build/15146/testReport/org.apache.hadoop.hdfs.server.datanode/TestDataNodeMultipleRegistrations/testClusterIdMismatchAtStartupWithHA/
> https://builds.apache.org/job/PreCommit-HADOOP-Build/15116/testReport/org.apache.hadoop.hdfs.server.datanode/TestDataNodeMultipleRegistrations/testDNWithInvalidStorageWithHA/
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
