[
https://issues.apache.org/jira/browse/YARN-6215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15884290#comment-15884290
]
Karthik Kambatla commented on YARN-6215:
----------------------------------------
Since the scheduler state is transient and the update thread runs only so
often, it is hard to ensure that containers are not preempted by mistake. That
said, you are right about not being worse than before. I guess it is okay to
start with the readLock and optimize the locking later if need be.
Comments on the patch:
# {{lock()}} should be called outside the try-block
# Can we add a comment capturing why we need to lock() here. Maybe, something
along the lines of: "Hold the scheduler readlock so this is not concurrent with
the update thread."
# {{FairScheduler.getSchedulerReadLock}} can be package-private.
> TestFairSchedulerPreemption.testPreemptionBetweenNonSiblingQueues fails in
> trunk
> --------------------------------------------------------------------------------
>
> Key: YARN-6215
> URL: https://issues.apache.org/jira/browse/YARN-6215
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: fairscheduler, test
> Reporter: Sunil G
> Assignee: Tao Jie
> Attachments: YARN-6215.001.patch
>
>
> *Error Message*
> Incorrect number of containers on the greedy app expected:<4> but was:<8>
> Failed test case
> [link|https://builds.apache.org/job/PreCommit-YARN-Build/15038/testReport/org.apache.hadoop.yarn.server.resourcemanager.scheduler.fair/TestFairSchedulerPreemption/testPreemptionBetweenNonSiblingQueues_FairSharePreemptionWithDRF_/]
> *Stacktrace*
> {noformat}
> java.lang.AssertionError: Incorrect number of containers on the greedy app
> expected:<4> but was:<8>
> at org.junit.Assert.fail(Assert.java:88)
> at org.junit.Assert.failNotEquals(Assert.java:743)
> at org.junit.Assert.assertEquals(Assert.java:118)
> at org.junit.Assert.assertEquals(Assert.java:555)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.fair.TestFairSchedulerPreemption.verifyPreemption(TestFairSchedulerPreemption.java:282)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.fair.TestFairSchedulerPreemption.testPreemptionBetweenNonSiblingQueues(TestFairSchedulerPreemption.java:323)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
