[
https://issues.apache.org/jira/browse/BUILDS-49?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14284100#comment-14284100
]
Andrew Purtell commented on BUILDS-49:
--------------------------------------
The HBase 0.98 builds are looking better since I changed the job configuration
to only run one unit test per forked JVM. I suspect memory on the build hosts
is overcommitted and if the surefire runner JVM commits one too many pages of
virtual memory it's bye bye.
> Surefire runner JVMs are being killed for HBase 0.98 Jenkins jobs
> -----------------------------------------------------------------
>
> Key: BUILDS-49
> URL: https://issues.apache.org/jira/browse/BUILDS-49
> Project: Infra Build Platform
> Issue Type: Bug
> Components: Jenkins
> Reporter: Andrew Purtell
> Assignee: Andrew Bayer
>
> Occasionally the JVMs executing forked runners from Surefire are being
> killed, failing HBase-0.98 Jenkins jobs.
> For example, see https://builds.apache.org/job/HBase-0.98/794:
> {noformat}
> Running org.apache.hadoop.hbase.security.access.TestCellACLs
> Running org.apache.hadoop.hbase.security.access.TestAccessController
> Killed
> Killed
> {noformat}
> or https://builds.apache.org/job/HBase-0.98/797/
> {noformat}
> Running org.apache.hadoop.hbase.snapshot.TestFlushSnapshotFromClient
> Running org.apache.hadoop.hbase.snapshot.TestRestoreFlushSnapshotFromClient
> Killed
> Killed
> {noformat}
> Is there something we can do to avoid this?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
