[
https://issues.apache.org/jira/browse/MAPREDUCE-4026?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Harsh J updated MAPREDUCE-4026:
-------------------------------
Attachment: MAPREDUCE-4026.patch
Bobby - Good point, I should've done that!
Addressed now, and also fixed some references to the MINIMUM_ALLOCATION config
where applicable. MAXIMUM_ALLOCATION doesn't seem to be tested/used for at the
moment, and I've filed MAPREDUCE-4171 for that.
Thanks for taking a look, please do review again!
To be sure: This only changes minimum allocation values, not job default memory
in MR. DistributedShell too in its test uses 128 MB as the min-alloc, for
example. So I don't see this change regressing any working features.
> Lower minimum-allocation-mb to sensible defaults
> ------------------------------------------------
>
> Key: MAPREDUCE-4026
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-4026
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: mrv2, scheduler
> Affects Versions: 2.0.0
> Reporter: Harsh J
> Assignee: Harsh J
> Attachments: MAPREDUCE-4026.patch, MAPREDUCE-4026.patch,
> MAPREDUCE-4026.patch, MAPREDUCE-4026.patch, MAPREDUCE-4026.patch
>
>
> The CapacityScheduler's minimum-allocation-mb is set to 1024.
> The FIFO's minimum-allocation-mb meanwhile, is 128.
> I propose changing the formers' minimum to that amount as well. 1024 is way
> too much as a default, wastes "slots" on NMs - and I also do not see why CS
> has to deviate that settings from the FIFO default.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
