[
https://issues.apache.org/jira/browse/HADOOP-4018?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12629078#action_12629078
]
Amar Kamat commented on HADOOP-4018:
------------------------------------
More comments
1) Computing allocated tasks based on internal datastructures of
{{JobInProgress}} is incorrect as their count will always be missing.
{{garbageCollect()}} for the job frees up the memory taken up by the caches.
Note that {{garabgeCollect()}} will be invoked before {{retireJob()}}.
I think we should use {{numMapTasks}} and {{numReduceTasks}} instead of the
sizes of the caches. Also reset the values of {{numMapTasks/numReduceTasks}} if
the limit is crossed. If someone introduces partial expansion of tasks then we
can think about it later.
2) I am not sure if changing the log level of one class is sufficient. It would
be nice if we can set the log level of the complete framework in the testcase.
But again is it required?
3) Did not check the test case completely but should take care of the following
cases
- submit multiple jobs such that all of them should be accommodated. Both
while the previous ones are running and also when the previous jobs are cleaned
up.
- submit some small jobs and then a large job that exceeds the limit.
Submit a small job after a limit-crossing job and make sure that the job gets
accepted. The reason is to test if the cleanup is done and there is no side
effect of large job being submitted and getting rejected
- submit two jobs back to back such that {{job1.totalTasks() +
job2.totalTasks() > limit}}. Hence the first one should be accepted and the
next one should be rejected.
- anything else?
4) I dont see in the test case where the
{{mapred.jobtracker.completeuserjobs.maximum}} is set. By default it will be
100 and the test case might never test the cleanup process.
> limit memory usage in jobtracker
> --------------------------------
>
> Key: HADOOP-4018
> URL: https://issues.apache.org/jira/browse/HADOOP-4018
> Project: Hadoop Core
> Issue Type: Bug
> Components: mapred
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
> Attachments: maxSplits.patch, maxSplits2.patch, maxSplits3.patch,
> maxSplits4.patch, maxSplits5.patch, maxSplits6.patch, maxSplits7.patch
>
>
> We have seen instances when a user submitted a job with many thousands of
> mappers. The JobTracker was running with 3GB heap, but it was still not
> enough to prevent memory trashing from Garbage collection; effectively the
> Job Tracker was not able to serve jobs and had to be restarted.
> One simple proposal would be to limit the maximum number of tasks per job.
> This can be a configurable parameter. Is there other things that eat huge
> globs of memory in job Tracker?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
