hadoop-mapreduce-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Haibo Chen (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (MAPREDUCE-6506) Make the reducer-preemption configs consistent in how they handle defaults
Date Thu, 02 Feb 2017 22:38:51 GMT

    [ https://issues.apache.org/jira/browse/MAPREDUCE-6506?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15850634#comment-15850634

Haibo Chen commented on MAPREDUCE-6506:

Talked with Karthik offline, we think there is no change needed given the current state. mapreduce.job.reducer.preempt.delay.sec
adds a delay to preempt reducers, so if we disable it, we'd preempt reducers instantly.  mapreduce.job.reducer.unconditional-preempt.delay.sec
preempts reducers unconditionally, so disabling it means we do not preempt reducers. They
are therefore consistent in their behaviors.

> Make the reducer-preemption configs consistent in how they handle defaults
> --------------------------------------------------------------------------
>                 Key: MAPREDUCE-6506
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-6506
>             Project: Hadoop Map/Reduce
>          Issue Type: Task
>          Components: applicationmaster
>    Affects Versions: 2.8.0
>            Reporter: Karthik Kambatla
>            Assignee: Gergő Pásztor
>            Priority: Critical
> {{mapreduce.job.reducer.preempt.delay.sec}} and {{mapreduce.job.reducer.unconditional-preempt.delay.sec}}
are two configs related to reducer preemption. These configs are not consistent in how they
handle non-positive values. Also, the way to disable them is different. 
> It would be nice to make them consistent somehow, and change the behavior of former if
need be. 

This message was sent by Atlassian JIRA

To unsubscribe, e-mail: mapreduce-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: mapreduce-issues-help@hadoop.apache.org

View raw message