hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Steve Loughran (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-4937) EntityGroupFSTimelineStore to close cached leveldb instances in serviceStop()
Date Sat, 09 Apr 2016 12:00:28 GMT

    [ https://issues.apache.org/jira/browse/YARN-4937?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15233516#comment-15233516

Steve Loughran commented on YARN-4937:

I'm going to see if I can do a workaround for this, of having separate directories per test
case, hence leveldb instances. However, it means that the test cases are leaking leveldb instances,
so I may hit resource limitations instead.

> EntityGroupFSTimelineStore to close cached leveldb instances in serviceStop()
> -----------------------------------------------------------------------------
>                 Key: YARN-4937
>                 URL: https://issues.apache.org/jira/browse/YARN-4937
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: timelineserver
>    Affects Versions: 2.8.0
>         Environment: JUnit test runs with
>            Reporter: Steve Loughran
> When I enable ATS1.5 and non-summary data, all my JUnit test runs with it fail  pass
if called individually. But if executed together, the 2nd+ service fails, key exception text
> {code}
> org.fusesource.leveldbjni.internal.NativeDB$DBException: IO error: lock /Users/stevel/Projects/spark/yarn-timeline/yarn-timeline-history/target/tmp/integration/leveldb/timelineEntityGroupId_1111_0_application_1111_0000-timeline-cache.ldb/LOCK:
already held by process
> {code}
> I believe the problem is that {{EntityGroupFSTimelineStore.cachedLogs}} isn't cleaning
up its logs in {{serviceStop()}}, and as my tests use the same directory and appIds in its
test runs, the next attempt to create a cache item fails.

This message was sent by Atlassian JIRA

View raw message