You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@yunikorn.apache.org by "Wilfred Spiegelenburg (Jira)" <ji...@apache.org> on 2023/10/20 09:44:00 UTC

[jira] [Created] (YUNIKORN-2057) FindQueueByAppID is slow

Wilfred Spiegelenburg created YUNIKORN-2057:
-----------------------------------------------

             Summary: FindQueueByAppID is slow
                 Key: YUNIKORN-2057
                 URL: https://issues.apache.org/jira/browse/YUNIKORN-2057
             Project: Apache YuniKorn
          Issue Type: Improvement
          Components: core - scheduler
            Reporter: Wilfred Spiegelenburg


The preemptor calls {{p.queue.FindQueueByAppID()}} to figure out which queue the application runs in. This is an expensive way to retrieve the queue.

Especially when the queue structure grows large and there is no fencing turned on with could be traversing large numbers of queue and doing even more lookups. We also do this for every single victim. That could be a list and we could be trying multiple nodes increasing the number of times we make that call.

We should leverage the application tracking at the partition level to do a simple call:
    PartitionContext.getApplicationQueue(appID string) *Queue
This will require some "plumbing" like we did with the node functions but it will perform much better.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@yunikorn.apache.org
For additional commands, e-mail: dev-help@yunikorn.apache.org