[ 
https://issues.apache.org/jira/browse/YARN-6736?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16328653#comment-16328653
 ] 

Rohith Sharma K S commented on YARN-6736:
-----------------------------------------

Thanks [~vrushalic] for post review.. Since there were no reply from anyone I 
went ahead with committing this patch. We can created follow up JIRA if any 
issues has been observed from this patch. 

bq. In ApplicationMaster # init, the patch removes the setting of 
timelineServiceV1Enabled at line 638 but uses it at line 709. Since it is not 
set, this will not invoke publishApplicationAttemptEvent I think? Similarly for 
timelineServiceV2Enabled at lines 701. 
Before publishing, *startTimelineClient(conf);* is being called which 
initializes these variables at line no. 700 which is not present in this patch. 
You can see startTimelineClient(conf) method change in patch.

bq. Perhaps a good idea to catch this and ignore and proceed. 
IIUC, if wrong configuration value has been set then we need to inform admin 
immediately. So I thought let it through exception and fail the service start 
up.

bq. why the conf settings for timeline server address & port had to be moved 
out of synchronized in MiniYARNCluster were in this patch?
Configuration is global settings which same config is being passed to 
ApplicationHistoryServerWrapper service. Doing this, all the services gets same 
values. I believe this should not cause any impact. Btw, I guess it changed 
because of test failure after this patch change which I see from history of 
patches.


> Consider writing to both ats v1 & v2 from RM for smoother upgrades
> ------------------------------------------------------------------
>
>                 Key: YARN-6736
>                 URL: https://issues.apache.org/jira/browse/YARN-6736
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: timelineserver
>            Reporter: Vrushali C
>            Assignee: Aaron Gresch
>            Priority: Major
>             Fix For: 3.1.0, 2.10.0, 3.0.1, yarn-7055
>
>         Attachments: YARN-6736-YARN-5355.001.patch, 
> YARN-6736-YARN-5355.002.patch, YARN-6736-YARN-5355.003.patch, 
> YARN-6736-YARN-5355.004.patch, YARN-6736-YARN-5355.005.patch, 
> YARN-6736.001.patch, YARN-6736.002.patch
>
>
> When the cluster is being upgraded from atsv1 to v2, it may be good to have a 
> brief time period during which RM writes to both atsv1 and v2. This will help 
> frameworks like Tez migrate more smoothly. 



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org

Reply via email to