[
https://issues.apache.org/jira/browse/YARN-6570?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16179240#comment-16179240
]
Jason Lowe commented on YARN-6570:
----------------------------------
-1 for the branch-2.8 patch. This broke a lot of things since nobody, not even
the RM, expects the new SCHEDULED state being reported by the NM. That causes
the RM to think any container that gets reported as SCHEDULED is actually
COMPLETED because of this existing logic in RMNodeImpl:
{code}
// Process running containers
if (remoteContainer.getState() == ContainerState.RUNNING) {
[...]
} else {
// A finished container
{code}
I've seen a lot of MapReduce jobs fail as a result because containers are
reported as completing as soon as they start localizing.
> No logs were found for running application, running container
> -------------------------------------------------------------
>
> Key: YARN-6570
> URL: https://issues.apache.org/jira/browse/YARN-6570
> Project: Hadoop YARN
> Issue Type: Bug
> Components: yarn
> Reporter: Sumana Sathish
> Assignee: Junping Du
> Priority: Critical
> Fix For: 2.9.0, 3.0.0-beta1, 2.8.3, 3.1.0
>
> Attachments: YARN-6570-branch-2.8.001.patch,
> YARN-6570-branch-2.8.002.patch, YARN-6570.poc.patch, YARN-6570-v2.patch,
> YARN-6570-v3.patch
>
>
> 1.Obtain running containers from the following CLI for running application:
> yarn container -list appattempt
> 2. Couldnot fetch logs
> {code}
> Can not find any log file matching the pattern: ALL for the container
> {code}
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
