[
https://issues.apache.org/jira/browse/YARN-7943?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16367829#comment-16367829
]
Tarun Parimi commented on YARN-7943:
------------------------------------
Attached a trivial patch which increases the
yarn.resourcemanager.nodemanagers.heartbeat-interval-ms to 10s
Alternatively, calling nodeUpdate() on the nodes, instead of
CapacityScheduler.schedule() should also work.
> testCSReservationWithRootUnblocked() and testCSQueueBlocked() in
> TestCapacityScheduler are flaky
> ------------------------------------------------------------------------------------------------
>
> Key: YARN-7943
> URL: https://issues.apache.org/jira/browse/YARN-7943
> Project: Hadoop YARN
> Issue Type: Task
> Components: capacity scheduler, test
> Reporter: Tarun Parimi
> Priority: Major
> Attachments: YARN-7943.001.patch
>
>
> testCSReservationWithRootUnblocked() and testCSQueueBlocked() test for the
> expected value of Used Resources in different queues after calling allocate()
> and CapacityScheduler.schedule().
> If CapacityScheduler.schedule() gets executed more than a second after
> NodeAddedSchedulerEvent was handled, then the node will be skipped from
> scheduling. This results in test failure as the node resources will not be
> allocated and so the expected Used Resource value differs.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
