[
https://issues.apache.org/jira/browse/HADOOP-3447?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Amar Kamat resolved HADOOP-3447.
--------------------------------
Resolution: Duplicate
Will be fixed by HADOOP-5083.
> JobTracker should only keep runnable jobs in the priority list
> --------------------------------------------------------------
>
> Key: HADOOP-3447
> URL: https://issues.apache.org/jira/browse/HADOOP-3447
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred
> Reporter: Amar Kamat
> Assignee: Amar Kamat
>
> Currently the JobTracker keeps the _non-runnable jobs_ in the priority list.
> Non runnable jobs are deleted only when they are *too* old (by default if
> they have finished 24hrs ago). A newly submitted job gets added into this
> list and the whole list is sorted. Here the JobTracker does some useless
> computation by sorting all the non-runnable jobs too. Also the scheduling
> code scans the priority list *searching* for a _runnable_ job. We could either
> 1) have a separate list for _non running_ jobs
> 2) give runnable jobs higher priority over non-runnable ones.
> ----
> _Note: Here, runnable jobs means jobs that are either running or queued._
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
