[
https://issues.apache.org/jira/browse/YARN-1109?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13806587#comment-13806587
]
Sandy Ryza commented on YARN-1109:
----------------------------------
Thanks for working on this [[email protected]].
As ContainerImpl already prints out an INFO message when a container's state
changes, I think this log message does not need to do so, and demoting it to
DEBUG is sufficient.
We should surround the debug message with if (LOG.isDebugEnabled()) so that we
don't pay the cost of string construction in normal operation.
Also, as Vinod says, the log message should use a more compact representation
of the container ID, i.e. what ContainerId#toString returns. Doing this
probably requires changing ContainerStatusPBImpl#toString to not rely on the
protobuf default debug string and to construct the string explicitly. We
should make sure to use a StringBuilder to avoid unnecessary string copies.
> Consider throttling or demoting NodeManager "Sending out status for
> container" logs
> -----------------------------------------------------------------------------------
>
> Key: YARN-1109
> URL: https://issues.apache.org/jira/browse/YARN-1109
> Project: Hadoop YARN
> Issue Type: Improvement
> Components: nodemanager
> Affects Versions: 2.1.0-beta
> Reporter: Sandy Ryza
> Labels: newbie
> Attachments: YARN-1109.patch
>
>
> Diagnosing NodeManager and container launch problems is made more difficult
> by the enormous number of logs like
> {code}
> Sending out status for container: container_id {, app_attempt_id {,
> application_id {, id: 18, cluster_timestamp: 1377559361179, }, attemptId: 1,
> }, id: 1337, }, state: C_RUNNING, diagnostics: "Container killed by the
> ApplicationMaster.\n", exit_status: -1000
> {code}
> On an NM with a few containers I am seeing tens of these per second.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
