[
https://issues.apache.org/jira/browse/YARN-1775?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13939599#comment-13939599
]
Chris Nauroth commented on YARN-1775:
-------------------------------------
bq. Tested with branch 2.4.0 on 20 node cluster.
Hi, [~rajesh.balamohan]. Can you please share more details on this testing?
What would be particularly interesting is pushing a cluster close to the limits
of its memory capacity and then observing the differences in behavior between
{{ProcfsBasedProcessTree}} and {{SMAPBasedProcessTree}}. The argument is that
{{ProcfsBasedProcessTree}} is effectively over-conservative and may kill
containers too quickly. Did this testing show that {{SMAPBasedProcessTree}}
was able to keep more containers running? If so, were there any negative
performance impacts observed? (I wondered about the potential for introducing
more thrashing as a side effect of killing containers less quickly.)
Thanks for taking on this effort!
> Create SMAPBasedProcessTree to get PSS information
> --------------------------------------------------
>
> Key: YARN-1775
> URL: https://issues.apache.org/jira/browse/YARN-1775
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: nodemanager
> Reporter: Rajesh Balamohan
> Priority: Minor
> Fix For: 2.5.0
>
> Attachments: yarn-1775-2.4.0.patch
>
>
> Create SMAPBasedProcessTree (by extending ProcfsBasedProcessTree), which will
> make use of PSS for computing the memory usage.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
