[
https://issues.apache.org/jira/browse/MAPREDUCE-2310?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13096859#comment-13096859
]
Harsh J commented on MAPREDUCE-2310:
------------------------------------
Hm, experimented a bit more here and I believe you should update the JIRA
summaries and description to illustrate the problem better.
The said problem of the TT crashing out _only_ occurs when the JobTracker has
not completely initialized and added in the specific TT node before its taken
down. If the TT faced proper initialization before the JT went down, you won't
see this.
Patch is OK to me, since waitForProxy anyway waits internally (while the look
implemented doesn't seem to sleep). However, this would be fixing a very rare
case where a JT is taken down immediately for some reason, and not in case of
general restarts. I suppose it is okay to have the uninited TTs running anyway,
like zombies awaiting their master… :)
> If we stop Job Tracker, Task Tracker is also getting stopped.
> -------------------------------------------------------------
>
> Key: MAPREDUCE-2310
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-2310
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: tasktracker
> Affects Versions: 0.20.2
> Reporter: Devaraj K
> Assignee: Devaraj K
> Priority: Minor
> Attachments: MAPREDUCE-2310.patch
>
>
> If we execute stop-jobtracker.sh for stopping Job Tracker, Task Tracker is
> also stopping.
> This is not applicable for the latest (trunk) code because stop-jobtracker.sh
> file is not coming.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
