[
https://issues.apache.org/jira/browse/MAPREDUCE-4295?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13397865#comment-13397865
]
Robert Joseph Evans commented on MAPREDUCE-4295:
------------------------------------------------
The change looks good to me. I just have two minor comments.
# It is BuilderUtils.newContainerToken that could throw the
IllegalArgumentException because of a DNS outage. It might be good to move the
creation of the ContainerTokenIdentifier out of the try block so we don't
accidentally treat something else as a DNS issue. This is mostly because
IllegalArgumentException really does not indicate that there was a problem
resolving a host name, so I would prefer to reduce the chance of handling the
error incorrectly.
# Do we want to log the exception as a string or do we want to pass in the
exception as an argument to the log so we get the full stack trace? In the
case of a DNS outage we may not want to fill up the logs with useless messages,
but if there is a different error happening I would prefer to have more
information to be able to debug it.
> RM crashes due to DNS issue
> ---------------------------
>
> Key: MAPREDUCE-4295
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-4295
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: mrv2, resourcemanager
> Reporter: Thomas Graves
> Assignee: Thomas Graves
> Priority: Critical
> Fix For: 0.23.3
>
> Attachments: MAPREDUCE-4295.patch, MAPREDUCE-4295.patch,
> MAPREDUCE-4295.patch
>
>
> we had a DNS outage and the RM crashed with the following backtrace:
> 2012-05-29 19:17:34,492 FATAL
> org.apache.hadoop.yarn.server.resourcemanager.ResourceManager: Error in
> handling event type NODE_UPDATE to the scheduler
> java.lang.IllegalArgumentException: java.net.UnknownHostException:
> host.com at
> org.apache.hadoop.security.SecurityUtil.buildTokenService(SecurityUtil.java:430)
> at
> org.apache.hadoop.yarn.util.BuilderUtils.newContainerToken(BuilderUtils.java:261)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.LeafQueue.createContainer(LeafQueue.java:1184)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.LeafQueue.getContainer(LeafQueue.java:1167)
> at
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.LeafQueue.assignContainer(LeafQueue.java:1213)
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
