You are viewing a plain text version of this content. The canonical link for it is here.
Posted to mapreduce-issues@hadoop.apache.org by "Peter Bacsko (JIRA)" <ji...@apache.org> on 2017/08/04 13:50:00 UTC
[jira] [Comment Edited] (MAPREDUCE-6870) Add configuration for MR
job to finish when all reducers are complete (even with unfinished mappers)
[ https://issues.apache.org/jira/browse/MAPREDUCE-6870?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16114240#comment-16114240 ]
Peter Bacsko edited comment on MAPREDUCE-6870 at 8/4/17 1:49 PM:
-----------------------------------------------------------------
What's your suggestion to the variables?
My ideas: {{finishJobWhenReducersDone}}, {{MRJobConfig.FINISH_JOB_WHEN_REDUCERS_DONE}}
Preventing {{TA_KILL}} events: basically I just store a state information in each {{MapTaskImpl}}. But it's unnecessary since you can store this in a single variable after sending the kill events. So your approach is better.
New test: in a real environment, certain events are coming from {{TaskImpl}} and {{TaskAttemptImpl}}. However in the tests, these are mocked inside {{JobImpl}}, so you have to generate them manually. To properly test the behavior of this change, it might make sense to use the real impl classes instead of mocks.
bq. Also, do we expect the job the succeed even when killMappers is set to false?
Only if we send the completion events. If we don't, then of course it stays in RUNNING. I took the idea of finishing mappers/reducers from this test: https://github.com/apache/hadoop/blob/78b487bde175544ebe40e4dafab35569baa1d79e/hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestJobImpl.java#L597-L625
was (Author: pbacsko):
What's your suggestion to the variables?
My ideas: {{finishJobWhenReducersDone}}, {{MRJobConfig.FINISH_JOB_WHEN_REDUCERS_DONE}}
Preventing {{TA_KILL}} events: basically I just store a state information in each {{MapTaskImpl}}. But it's unnecessary since you can store this in a single variable after sending the kill events. So your approach is better.
New test: in {{TestJobImpl}}, certain events are coming from {{TaskImpl}} and {{TaskAttemptImpl}}. However these are mocked inside {{JobImpl}}, so you have to generate them manually. To properly test the behavior of this change, it might make sense to use the real impl classes instead of mocks.
bq. Also, do we expect the job the succeed even when killMappers is set to false?
Only if we send the completion events. If we don't, then of course it stays in RUNNING. I took the idea of finishing mappers/reducers from this test: https://github.com/apache/hadoop/blob/78b487bde175544ebe40e4dafab35569baa1d79e/hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestJobImpl.java#L597-L625
> Add configuration for MR job to finish when all reducers are complete (even with unfinished mappers)
> ----------------------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-6870
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-6870
> Project: Hadoop Map/Reduce
> Issue Type: Improvement
> Affects Versions: 2.6.1
> Reporter: Zhe Zhang
> Assignee: Peter Bacsko
> Attachments: MAPREDUCE-6870-001.patch, MAPREDUCE-6870-002.patch, MAPREDUCE-6870-003.patch
>
>
> Even with MAPREDUCE-5817, there could still be cases where mappers get scheduled before all reducers are complete, but those mappers run for long time, even after all reducers are complete. This could hurt the performance of large MR jobs.
> In some cases, mappers don't have any materialize-able outcome other than providing intermediate data to reducers. In that case, the job owner should have the config option to finish the job once all reducers are complete.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: mapreduce-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: mapreduce-issues-help@hadoop.apache.org