[
https://issues.apache.org/jira/browse/YARN-3542?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15098214#comment-15098214
]
Varun Vasudev commented on YARN-3542:
-------------------------------------
{quote}
This effectively means the old code is not used anymore, and that the new
code is stable.
And it seems like we are stating that
[..], there should be no issue hooking into the new handler using the old
configuration mechanism.?
Given this and the fact that we are internally overriding to use the new
handlers, is there a reason for keeping the old code at all? Also if we are
using the new handler code internally anyways, we can proceed with the
deprecation (or better deletion) of LCEResourcesHandler interface,
DefaultLCEResourcesHandler etc?
{quote}
I think there are two issues here -
1) should we keep the CgroupsLCEResourcesHandler code and
2) should we do away with the resource-handler code altogether
The answer to (1) is that it is not required but I think it's useful to keep
around for some time for two reasons -
1. As documentation/reference - in case some discrepancy is seen in the new
implementation
2. In the off-chance that someone has inherited from it. Unfortunately the
resource-handler classes are public via config but are not annotated as such.
The answer to (2) is no - the resource handler interface is public and we
shouldn't break it. Deprecating it is fine once we provide an alternative(which
we currently don't have) or decide we will not support it any longer. Either
way that decision should be a seperate JIRA, not this one.
> Re-factor support for CPU as a resource using the new ResourceHandler
> mechanism
> -------------------------------------------------------------------------------
>
> Key: YARN-3542
> URL: https://issues.apache.org/jira/browse/YARN-3542
> Project: Hadoop YARN
> Issue Type: Bug
> Components: nodemanager
> Reporter: Sidharta Seethana
> Assignee: Varun Vasudev
> Priority: Critical
> Attachments: YARN-3542.001.patch, YARN-3542.002.patch,
> YARN-3542.003.patch, YARN-3542.004.patch, YARN-3542.005.patch,
> YARN-3542.006.patch, YARN-3542.007.patch
>
>
> In YARN-3443 , a new ResourceHandler mechanism was added which enabled easier
> addition of new resource types in the nodemanager (this was used for network
> as a resource - See YARN-2140 ). We should refactor the existing CPU
> implementation ( LinuxContainerExecutor/CgroupsLCEResourcesHandler ) using
> the new ResourceHandler mechanism.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
