[ 
https://issues.apache.org/jira/browse/MAPREDUCE-5502?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13823462#comment-13823462
 ] 

Zhijie Shen commented on MAPREDUCE-5502:
----------------------------------------

bq. So in this scenario the AM has finished the job but not unregistered yet. 
AM is telling clients that connect to it that the job status is 
SUCCEEDED/FAILED/KILLED (i.e.: not RUNNING but in some terminal state) but the 
AM has yet to unregister with the RM so the RM is directing clients to the AM 
when asked.

It should no longer be the case after MAPREDUCE-5505, because the final state 
will not be release until AM get unregistered with RM, when history tracking 
URL should be updated.

Anyway, the condition check here seems to be a bug. However, I'm not sure the 
patch here is fixing the scenario [~vrushalic] has mentioned

bq. From our investigation, it appears that the client's kill sends a KILL to 
the App Master as well as to the RM for this App.

If this is the case, obviously the first killApplication is not executed but 
the second. This means killJob has been executed, but after 10s we find the job 
is still not killed. It is possible that the code path doesn't check the other 
two final states, SUCCEEDED and FAILED, therefore, kill is invoked on RM. 
However, it should no longer be problem after MAPREDUCE-5505, because when the 
job returns SUCCEEDED/FAILED, it has already unregistered with RM, and updated 
the tracking URL.

It is also possible that process kill is delayed on AM due to some random 
issue, and the job is still in RUNNING after 10s. On the other side while RM 
moves so fast to complete the kill. Then, the tracking URL will not be updated. 
I'm afraid the patch is not going to take care of this possibility.

> History link in resource manager is broken for KILLED jobs
> ----------------------------------------------------------
>
>                 Key: MAPREDUCE-5502
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5502
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>          Components: resourcemanager
>    Affects Versions: 2.0.5-alpha
>            Reporter: Vrushali C
>            Assignee: Vrushali C
>              Labels: ui
>         Attachments: patch_5502.txt, patch_5502_2.txt, patch_5502_3_1.txt
>
>
> History link in resource manager is broken for KILLED jobs.
> Seems to happen with jobs with State 'KILLED' and FinalStatus 'KILLED'. If 
> the State is 'FINISHED' and FinalStatus is 'KILLED', then the "History" link 
> is fine.
> It isn't easy to reproduce the problem since the time at which the app is 
> killed determines the state it ends up in, which is hard to guess. these 
> particular jobs seem to get a Diagnostics message of "Application killed by 
> user." where as the other killed jobs get " Kill Job received from client 
> job_1378766187901_0002
> Job received Kill while in RUNNING state. "



--
This message was sent by Atlassian JIRA
(v6.1#6144)

Reply via email to