[
https://issues.apache.org/jira/browse/YARN-4966?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15256423#comment-15256423
]
Junping Du commented on YARN-4966:
----------------------------------
Agree with [~vvasudev] that the run() method is too long now and we should do
some refactor work (may be in another JIRA?). Within the same method call, we
are construct/destruct yarnclient for several times, like:
getApplicationState(), getContainerReport(), etc. - that should be improved.
Also, there is question here:
{noformat}
+ } else if (!isApplicationFinished(appState)) {
+ System.err.println("Unable to get logs for this container:"
+ + containerIdStr + "for the application:" + appId);
+ System.out.println("Please specify the nodeAddress, and use "
+ + "yarn logs -applicationId <appId> -containerId <containerId> "
+ + "--nodeAddress <nodeAddress> to get the container logs");
+ return -1;
}
{noformat}
Why we must need node address if app is not finished? Can we ask RM to get the
nodeAddress instead?
> More improvement to get Container logs without specify nodeId
> -------------------------------------------------------------
>
> Key: YARN-4966
> URL: https://issues.apache.org/jira/browse/YARN-4966
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Xuan Gong
> Assignee: Xuan Gong
> Attachments: YARN-4966.1.patch, YARN-4966.2.patch, YARN-4966.3.patch
>
>
> Currently, for the finished application, we can get the container logs
> without specify node id, but we need to enable
> yarn.timeline-service.generic-application-history.enabled.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
