[jira] Commented: (MAPREDUCE-1342) Potential JT deadlock in faulty TT tracking

["Amar Kamat (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MAPREDUCE-1342?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12796092#action_12796092 ] 

Amar Kamat commented on MAPREDUCE-1342:
---------------------------------------

Simply making _potentiallyFaultyTrackers_ a concurrent HashMap and removing the *synchronized* keyword might introduce more issues. I think the reason for synchronizing on _potentiallyFaultyTrackers_ was to perform some operations in an atomic manner. Have you checked if the semantics remain same after removing the synchronized keyword? I think making _potentiallyFaultyTrackers_ as concurrent HashMap is better but might be dangerous.

One other way to avoid the deadlock would be by marking few non-private apis in JobTracker.FaultyTrackerInfo as synchronized. Mainly
{code}
JobTracker.FaultyTrackerInfo.incrementFaults // called via Heartbeat and testcases
JobTracker.FaultyTrackerInfo.markTrackerHealthy // called via Heartbeat
JobTracker.FaultyTrackerInfo.shouldAssignTasksToTracker // called via Heartbeat and testcases
JobTracker.FaultyTrackerInfo.isBlacklisted // called in multiple cases .. need to check
JobTracker.FaultyTrackerInfo.getFaultCount // called via Heartbeat and testcases
JobTracker.FaultyTrackerInfo.getReasonForBlackListing // never used!
JobTracker.FaultyTrackerInfo.setNodeHealthStatus // called via Heartbeat and testcases
{code}

So except JobTracker.FaultyTrackerInfo.isBlacklisted(), all the calls are centrally locked on JobTracker. Hence adding the synchronized keyword in the method signature wouldnt introduce any overhead. Need to check on JobTracker.FaultyTrackerInfo.isBlacklisted().

> Potential JT deadlock in faulty TT tracking
> -------------------------------------------
>
>                 Key: MAPREDUCE-1342
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-1342
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>          Components: jobtracker
>    Affects Versions: 0.22.0
>            Reporter: Todd Lipcon
>         Attachments: cycle0.png, mapreduce-1342-1.patch
>
>
> JT$FaultyTrackersInfo.incrementFaults first locks potentiallyFaultyTrackers, and then calls blackListTracker, which calls removeHostCapacity, which locks JT.taskTrackers
> On the other hand, JT.blacklistedTaskTrackers() locks taskTrackers, then calls faultyTrackers.isBlacklisted() which goes on to lock potentiallyFaultyTrackers.
> I haven't produced such a deadlock, but the lock ordering here is inverted and therefore could deadlock.
> Not sure if this goes back to 0.21 or just in trunk.

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