[
https://issues.apache.org/jira/browse/YARN-1643?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14646709#comment-14646709
]
MENG DING commented on YARN-1643:
---------------------------------
bq. IIUC, the main benefit is that we don't need to synchronize on the
enforceResourceLimits call, which can be heavy, right? If that is the case, we
probably also need to have proper synchronization for
ResourceCalculatorProcessTree, e.g.,
ProcfsBasedProcessTree/WindowsBasedProcessTree? These objects could be updated
by multiple threads as well. I was afraid that the code change may be too much?
I think I find a way without having to synchronize on
{{ResourceCalculatorProcessTree}}. All that is needed for synchronization in
this class is the trackingContainers map and the access to the
vmemLimit/pmemLimit/cpuVcores fields. The actual resource limit enforcement can
still be handled in the {{MonitoringThread.run}} thread only.
> Make ContainersMonitor can support change monitoring size of an allocated
> container in NM side
> ----------------------------------------------------------------------------------------------
>
> Key: YARN-1643
> URL: https://issues.apache.org/jira/browse/YARN-1643
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: nodemanager
> Reporter: Wangda Tan
> Assignee: MENG DING
> Attachments: YARN-1643-YARN-1197.4.patch,
> YARN-1643-YARN-1197.5.patch, YARN-1643-YARN-1197.6.patch, YARN-1643.1.patch,
> YARN-1643.2.patch, YARN-1643.3.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
