Prabhu Joseph commented on YARN-10309:

Thanks [~gandras] for the patch.

1. Why we need a separate StateDump if most of the metrics are already exposed 
via RMWebService /ws/v1/cluster/scheduler. Sorry, did not think about this when 
raising the Jira. It is unnecessary effort to maintain the metrics at both the 

Most of the useful metrics are already captured in scheduler response, it is 
better to add missing ones into the same instead of rewriting a new one. Thanks.

> Dump scheduler and queue state information into CapacityScheduler statedump 
> ----------------------------------------------------------------------------
>                 Key: YARN-10309
>                 URL: https://issues.apache.org/jira/browse/YARN-10309
>             Project: Hadoop YARN
>          Issue Type: Improvement
>          Components: capacityscheduler
>            Reporter: Prabhu Joseph
>            Assignee: Andras Gyori
>            Priority: Major
>         Attachments: YARN-10309.001.patch
> CapacityScheduler dump with Scheduler, FiCaSchedulerNode, FiCaSchedulerApp, 
> ParentQueue, LeafQueue, app and queue level ordering, multi node lookup 
> ordering details will make it ease to debug scheduler related issues instead 
> of correlating Debug Logs with CapacityScheduler code to get values for the 
> above.
> This is similar to FairScheduler statedump YARN-6042.

This message was sent by Atlassian Jira

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