[
https://issues.apache.org/jira/browse/MAPREDUCE-3032?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13125201#comment-13125201
]
Devaraj K commented on MAPREDUCE-3032:
--------------------------------------
bq.Can you debug why the diagnostics from the reportedStatus is empty? We need
to make sure the remote-diagnostics are not getting lost. Otherwise, we will
only see the diagnostics produce inside the AM.
I debugged the code, reportedStatus is getting updated from the task status
with statusUpdate API from the TaskUmbilicalProtocol. This task status is
getting creating with empty diagnostic info(code mentioned below) and no where
else it is getting updated after that.
{code:title=Task.java|borderStyle=solid}
this.taskStatus = TaskStatus.createTaskStatus(isMapTask(), this.taskId,
0.0f, numSlotsRequired,
TaskStatus.State.UNASSIGNED,
"", "", "",
isMapTask() ?
TaskStatus.Phase.MAP :
TaskStatus.Phase.SHUFFLE,
counters);
}
{code}
When the task fails, task is reporting using reportDiagnosticInfo from the
TaskUmbilicalProtocol and same is available in the taskAttempt.
bq.Also, what is the test that you used and what is the error message that you
saw?
As you have mentioned in the previous comment, deleting input file after the
job starts and simulated the same. I have attached history UI screen shot for
failed task after the changes.
bq.In any case, there is one more occurrence of usage of the diagnostics-info
from the 'reportedStatus' in TaskAttemptImpl.getReport(), we should fix that.
I missed it. I will look into this.
bq.Adding a test will greatly help this. You can look at
TestMRApp|TestFail|TestKill for inspiration.
I am very much interested to add new testcases. In this case, changes are in
private method, the method present in private class and also changes are very
minimal. Thanks for the information. I will add new test cases.
> JobHistory doesn't have error information from failed tasks
> -----------------------------------------------------------
>
> Key: MAPREDUCE-3032
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-3032
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: applicationmaster, mrv2
> Affects Versions: 0.23.0
> Reporter: Vinod Kumar Vavilapalli
> Assignee: Devaraj K
> Priority: Blocker
> Fix For: 0.23.0
>
> Attachments: JobFail.PNG, MAPREDUCE-3032.patch
>
>
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
