[
https://issues.apache.org/jira/browse/YARN-4216?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14939791#comment-14939791
]
Jason Lowe commented on YARN-4216:
----------------------------------
The problem is the NM is thinking it is being torn down _not_ for a restart and
is trying to clean up. From the NM log:
{noformat}
2015-10-01 14:58:40,688 ERROR
org.apache.hadoop.yarn.server.nodemanager.NodeManager: RECEIVED SIGNAL 15:
SIGTERM
2015-10-01 14:58:40,720 INFO
org.apache.hadoop.yarn.server.nodemanager.NodeStatusUpdaterImpl: Successfully
Unregistered the Node localhost:38153 with ResourceManager.
2015-10-01 14:58:40,731 INFO org.mortbay.log: Stopped
[email protected]:8042
2015-10-01 14:58:40,836 INFO
org.apache.hadoop.yarn.server.nodemanager.containermanager.ContainerManagerImpl:
Applications still running : [application_1443685464627_0007]
2015-10-01 14:58:40,836 INFO
org.apache.hadoop.yarn.server.nodemanager.containermanager.ContainerManagerImpl:
Waiting for Applications to be Finished
2015-10-01 14:58:40,837 INFO
org.apache.hadoop.yarn.server.nodemanager.containermanager.application.ApplicationImpl:
Application application_1443685464627_0007 transitioned from RUNNING to
FINISHING_CONTAINERS_WAIT
2015-10-01 14:58:40,837 INFO
org.apache.hadoop.yarn.server.nodemanager.containermanager.container.ContainerImpl:
Container container_1443685464627_0007_01_000014 transitioned from RUNNING to
KILLING
2015-10-01 14:58:40,837 INFO
org.apache.hadoop.yarn.server.nodemanager.containermanager.container.ContainerImpl:
Container container_1443685464627_0007_01_000001 transitioned from RUNNING to
KILLING
{noformat}
For a proper recovery the NM should not be trying to kill containers. Part of
the issue here is having the NM distinguish a shutdown that will be restarted
from a shutdown that won't be restarted. In the former it should _not_ kill
containers since the restart will recover them. For the latter it _should_
kill containers since there won't be an NM around later to control them. See
YARN-1362 for more details.
Does this problem occur if you stop the NM with kill -9 before restarting?
> Container logs not shown for newly assigned containers after NM recovery
> --------------------------------------------------------------------------
>
> Key: YARN-4216
> URL: https://issues.apache.org/jira/browse/YARN-4216
> Project: Hadoop YARN
> Issue Type: Bug
> Components: log-aggregation, nodemanager
> Reporter: Bibin A Chundatt
> Assignee: Bibin A Chundatt
> Priority: Critical
> Attachments: NMLog, ScreenshotFolder.png, yarn-site.xml
>
>
> Steps to reproduce
> # Start 2 nodemanagers with NM recovery enabled
> # Submit pi job with 20 maps
> # Once 5 maps gets completed in NM 1 stop NM (yarn daemon stop nodemanager)
> (Logs of all completed container gets aggregated to HDFS)
> # Now start the NM1 again and wait for job completion
> *The newly assigned container logs on NM1 are not shown*
> *hdfs log dir state*
> # When logs are aggregated to HDFS during stop its with NAME (localhost_38153)
> # On log aggregation after starting NM the newly assigned container logs gets
> uploaded with name (localhost_38153.tmp)
> History server the logs are now shown for new task attempts
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
