[
https://issues.apache.org/jira/browse/IGNITE-2725?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15232619#comment-15232619
]
Ivan Veselovsky commented on IGNITE-2725:
-----------------------------------------
In
org.apache.ignite.internal.processors.hadoop.jobtracker.HadoopJobTracker#onTaskFinished,
line 541 we have
{code}
cache.invokeAsync(info.jobId(), new
UpdatePhaseProcessor(incrCntrs, PHASE_COMPLETE)).
listen(failsLog);
{code}
, so it looks like the HadoopJobMetadata#phase may asynchronously become
"PHASE_COMPLETE", so there is no surprise the problematic assertion fails.
> Assertion in
> org.apache.ignite.internal.processors.hadoop.jobtracker.HadoopJobTracker.CancelJobProcessor#update
> fails if a job failed with an exception
> -------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: IGNITE-2725
> URL: https://issues.apache.org/jira/browse/IGNITE-2725
> Project: Ignite
> Issue Type: Bug
> Components: hadoop
> Affects Versions: 1.6
> Reporter: Ivan Veselovsky
> Assignee: Ivan Veselovsky
>
> Assertion in
> org.apache.ignite.internal.processors.hadoop.jobtracker.HadoopJobTracker.CancelJobProcessor#update
> : 1584 failed if a Hadoop job failed with an exception. The problem is that
> assertion expects the phase to be PHASE_CANCELLING , while actual phase is
> PHASE_COMPLETE.
> {code}
> assert meta.phase() == PHASE_CANCELLING || err != null: "Invalid phase for
> cancel: " + meta;
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
