[ 
https://issues.apache.org/jira/browse/YARN-6610?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Wangda Tan updated YARN-6610:
-----------------------------
    Attachment: YARN-6610.YARN-3926.perf-wangda-001.patch

Investigated more, it looks like original perf patch has some typos inside 
{{calculateShares2}}, I tried to run all CapacityScheduler tests with the old 
perf patch, it causes many failures. 

With attached patch {{YARN-6610.YARN-3926.perf-wangda-001.patch}}, it has 2 
failures, but looks unrelated. (Ran unit test w/o the patch, it fails as well).

[~templedf], it looks like existing tests in TestResourceCalculator are not 
enough to capture the issue, I tried to add some (testCompare2) but it looks 
not enough as well. If you have any cycles, could you help to add more unit 
tests?

I also tried to run testUserLimitThroughput again, the latest patch shows very 
similar result.

Since this is blocker of YARN-3926 merge, and it doesn't show performance 
regression, I plan to commit this patch to branch by end of today if no 
opposite opinions.

> DominantResourceCalculator.getResourceAsValue() dominant param is no longer 
> appropriate
> ---------------------------------------------------------------------------------------
>
>                 Key: YARN-6610
>                 URL: https://issues.apache.org/jira/browse/YARN-6610
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: resourcemanager
>    Affects Versions: YARN-3926
>            Reporter: Daniel Templeton
>            Assignee: Daniel Templeton
>            Priority: Critical
>         Attachments: YARN-6610.001.patch, YARN-6610.YARN-3926.002.patch, 
> YARN-6610.YARN-3926.003.patch, YARN-6610.YARN-3926.004.patch, 
> YARN-6610.YARN-3926.005.patch, YARN-6610.YARN-3926.006.patch, 
> YARN-6610.YARN-3926.006-rebased.patch, YARN-6610.YARN-3926.perf.patch, 
> YARN-6610.YARN-3926.perf-rebased.patch, 
> YARN-6610.YARN-3926.perf-wangda-001.patch
>
>
> The {{dominant}} param assumes there are only two resources, i.e. true means 
> to compare the dominant, and false means to compare the subordinate.  Now 
> that there are _n_ resources, this parameter no longer makes sense.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to