[jira] Commented: (MAPREDUCE-1204) Fair Scheduler preemption may preempt tasks running in slots unusable by the preempting job

["Matei Zaharia (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MAPREDUCE-1204?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12776136#action_12776136 ] 

Matei Zaharia commented on MAPREDUCE-1204:
------------------------------------------

+1 on adding an API similar to obtainNewMapTask that does not launch a task. This would simplify locality-aware scheduling too, because it would be easier to check whether a job has a local task to launch on the same node.

In the meantime, one thing you might try is making JobInProgress.shouldRunOnTaskTracker public and calling it. It's not perfect though, because other runnability checks could be added to JobInProgress in the future. In fact, there is already one other check in the current codebase that will prevent tasks from being launched when there is no free disk space on a node. The right solution is to structure JobInProgress in a more modular fashion so that these kinds of checks can be performed by schedulers.

> Fair Scheduler preemption may preempt tasks running in slots unusable by the preempting job
> -------------------------------------------------------------------------------------------
>
>                 Key: MAPREDUCE-1204
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-1204
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>          Components: contrib/fair-share
>    Affects Versions: 0.21.0
>            Reporter: Todd Lipcon
>
> The current preemption code works by first calculating how many tasks need to be preempted to satisfy the min share constraints, and then killing an equal number of tasks from other jobs, sorted to favor killing of young tasks. This works fine for the general case, but there are some edge cases where this can cause problems.
> For example, if the preempting job has blacklisted ("marked flaky") a particular task tracker, and that tracker is running the youngest task, preemption can still kill that task. The preempting job will then refuse that slot, since the tracker has been blacklisted. The same task that just got killed then gets rescheduled in that slot. This repeats ad infinitum until a new slot opens in the cluster.
> I don't have a good test case for this, yet, but logically it is possible.
> One potential fix would be to add an API to JobInProgress that functions identically to obtainNewMapTask but does not schedule the task. The preemption code could then use this while iterating through the sorted preemption list to check that the preempting jobs can actually make use of the candidate slots before killing them.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.