[jira] [Updated] (MAPREDUCE-4936) JobImpl uber checks for cpu are wrong

["Arun C Murthy (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/MAPREDUCE-4936?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Arun C Murthy updated MAPREDUCE-4936:
-------------------------------------

    Attachment: MAPREDUCE-4936.patch

Fixed the cpu check to not bork TestUberAM.

I've also fixed it to use not use hard-coded default, but it's systemic and I've opened MAPREDUCE-4943 to fix other instances since I'm looking to unblock 2.0.3-alpha for now.

No new tests since TestUberAM already covers it.
                
> JobImpl uber checks for cpu are wrong
> -------------------------------------
>
>                 Key: MAPREDUCE-4936
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-4936
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>          Components: mrv2
>    Affects Versions: 3.0.0, 2.0.3-alpha
>            Reporter: Daryn Sharp
>            Assignee: Arun C Murthy
>            Priority: Critical
>         Attachments: MAPREDUCE-4936.patch
>
>
> The cpu checks for uberizing have two issues:
> # the defaults are hardcoded instead of using the conf defaults
> # the comparison against the sys cpu size is using {{<}} instead of {{<=}}
> {code}
>     boolean smallCpu =
>         (
>             Math.max(
>                 conf.getInt(MRJobConfig.MAP_CPU_VCORES, 1), 
>                 conf.getInt(MRJobConfig.REDUCE_CPU_VCORES, 1)) < 
>              sysCPUSizeForUberSlot
>         );
> {code}
> Everything is defaulting to 1, so uber cpu checks are now disabled causing {{TestUberAM}} to fail.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira