[
https://issues.apache.org/jira/browse/YARN-6249?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15892305#comment-15892305
]
Tao Jie commented on YARN-6249:
-------------------------------
I debugged this test and detected the root cause of the failure.
In the test, FsLeafQueues are initialized before {{scheduler.setClock(clock)}}
is called in setup(). As a result, {{lastTimeAtMinShare}} in FsLeafQueue is
initialized to the long value of current time(a big number), and it will
compare to the time of {{ControlledClock}} which starts from 0.
In {{FsLeafQueue#minShareStarvation}} invoked in update()
{code}
long now = scheduler.getClock().getTime();
if (!starved) {
// Record that the queue is not starved
setLastTimeAtMinShare(now);
}
if (now - lastTimeAtMinShare < getMinSharePreemptionTimeout()) {
// the queue is not starved for the preemption timeout
starvation = Resources.clone(Resources.none());
}
{code}
If {{starved}} is true here at the first time this method is called, this queue
would never satisfy the min preemption timeout.
However I don't think it is a bug in the real world, because this issue is
related to ControlledClock only used in test.
> TestFairSchedulerPreemption is inconsistently failing on trunk
> --------------------------------------------------------------
>
> Key: YARN-6249
> URL: https://issues.apache.org/jira/browse/YARN-6249
> Project: Hadoop YARN
> Issue Type: Bug
> Components: fairscheduler, resourcemanager
> Affects Versions: 2.9.0
> Reporter: Sean Po
> Assignee: Yufei Gu
>
> Tests in TestFairSchedulerPreemption.java will inconsistently fail on trunk.
> An example stack trace:
> {noformat}
> Tests run: 24, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 24.879 sec
> <<< FAILURE! - in
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.fair.TestFairSchedulerPreemption
> testPreemptionSelectNonAMContainer[MinSharePreemptionWithDRF](org.apache.hadoop.yarn.server.resourcemanager.scheduler.fair.TestFairSchedulerPreemption)
> Time elapsed: 10.475 sec <<< FAILURE!
> 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:288)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.fair.TestFairSchedulerPreemption.testPreemptionSelectNonAMContainer(TestFairSchedulerPreemption.java:363)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
