[
https://issues.apache.org/jira/browse/YARN-8737?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17204467#comment-17204467
]
Tao Yang commented on YARN-8737:
--------------------------------
Hi, [~Amithsha], [~wangda], [~bteke]. Sorry for missing this issue so long.
I haven't dug into this issue or checked if the exception never happen again (I
have just search the key words "Comparison method violates its general
contract" from RM logs of our YARN clusters which can only be stored for 7
days, nothing returned) since this exception can't crash or affect the
scheduling process in our internal versions.
After looking into YARN-10178, I think this problem may be raised by multiple
causes, the same point is that some resources like capacity-resource or
used-resource in child queues(leaf or parent queue) changed while parent queue
is sorting them.
I think this patch can solve the problem for the configurations-updating
scenario, adding read lock in ParentQueue#sortAndGetChildrenAllocationIterator
can avoid the child queues' configured capacity be updated while being sorted.
[~wangda], [~bteke] very appreciate if you can help to review and commit this
patch.
And we should also fix the problem for the scheduling scenario in YARN-10178.
> Race condition in ParentQueue when reinitializing and sorting child queues in
> the meanwhile
> -------------------------------------------------------------------------------------------
>
> Key: YARN-8737
> URL: https://issues.apache.org/jira/browse/YARN-8737
> Project: Hadoop YARN
> Issue Type: Bug
> Components: capacityscheduler
> Affects Versions: 3.3.0, 2.9.3, 3.2.2, 3.1.4
> Reporter: Tao Yang
> Assignee: Tao Yang
> Priority: Critical
> Attachments: YARN-8737.001.patch
>
>
> Administrator raised a update for queues through REST API, in RM parent queue
> is refreshing child queues through calling ParentQueue#reinitialize,
> meanwhile, async-schedule threads is sorting child queues when calling
> ParentQueue#sortAndGetChildrenAllocationIterator. Race condition may happen
> and throw exception as follow because TimSort does not handle the concurrent
> modification of objects it is sorting:
> {noformat}
> java.lang.IllegalArgumentException: Comparison method violates its general
> contract!
> at java.util.TimSort.mergeHi(TimSort.java:899)
> at java.util.TimSort.mergeAt(TimSort.java:516)
> at java.util.TimSort.mergeCollapse(TimSort.java:441)
> at java.util.TimSort.sort(TimSort.java:245)
> at java.util.Arrays.sort(Arrays.java:1512)
> at java.util.ArrayList.sort(ArrayList.java:1454)
> at java.util.Collections.sort(Collections.java:175)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.policy.PriorityUtilizationQueueOrderingPolicy.getAssignmentIterator(PriorityUtilizationQueueOrderingPolicy.java:291)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.ParentQueue.sortAndGetChildrenAllocationIterator(ParentQueue.java:804)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.ParentQueue.assignContainersToChildQueues(ParentQueue.java:817)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.ParentQueue.assignContainers(ParentQueue.java:636)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler.allocateOrReserveNewContainers(CapacityScheduler.java:2494)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler.allocateOrReserveNewContainers(CapacityScheduler.java:2431)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler.allocateContainersOnMultiNodes(CapacityScheduler.java:2588)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler.allocateContainersToNode(CapacityScheduler.java:2676)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler.scheduleBasedOnNodeLabels(CapacityScheduler.java:927)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler$AsyncScheduleThread.run(CapacityScheduler.java:962)
> {noformat}
> I think we can add read-lock for
> ParentQueue#sortAndGetChildrenAllocationIterator to solve this problem, the
> write-lock will be hold when updating child queues in
> ParentQueue#reinitialize.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
