[
https://issues.apache.org/jira/browse/MAPREDUCE-5471?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13757831#comment-13757831
]
Jason Lowe commented on MAPREDUCE-5471:
---------------------------------------
bq. If the MR AM scans staging directory, followed by intermediate directory
and the done directory before starting the new AM's jobs, we should be good.
The tricky part is to locate the file in the done directory. Let's see if we
can do that.
I missed the offline discussions, so apologies if this is all stuff you have
hashed over before. IMHO this is an issue that can be resolved by not having
the RM return from the app unregister call until the persistent store has
confirmed it has removed the app state. Otherwise we can return control back
to the AM which will proceed to remove its staging directory and the app will
fail to restart. In other words, we cannot fix this completely in the MR AM
with a subsequent attempt because it is highly likely the subsequent attempt
cannot be launched successfully.
As for trying to protect the MR AM from a re-run after success, isn't this
already a solved problem? See MAPREDUCE-4819. The MR AM will check for prior
success in the staging directory and avoid re-running the job if that's there.
I don't think we need to go scraping through the done_intermediate and done
directory since the history file is copied, not moved, to the done_intermediate
directory. And if the staging directory is missing then the subsequent AM
attempt isn't going to start anyway.
> Succeed job tries to restart after RMrestart
> --------------------------------------------
>
> Key: MAPREDUCE-5471
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5471
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Reporter: yeshavora
> Assignee: Jian He
> Priority: Blocker
> Attachments: MR5471-1AM.log, MR5471-2AM.log
>
>
> Run a job , restart RM when job just finished. It should not restart the job
> once it Succeed.
> After RM restart, The AM of restarted job fails with below error.
> AM log after Rmrestart:
> 013-08-19 17:29:21,144 INFO [main]
> org.apache.hadoop.mapreduce.jobhistory.JobHistoryEventHandler: Stopping
> JobHistoryEventHandler. Size of the outstanding queue size is 0
> 2013-08-19 17:29:21,145 INFO [main]
> org.apache.hadoop.mapreduce.jobhistory.JobHistoryEventHandler: Stopped
> JobHistoryEventHandler. super.stop()
> 2013-08-19 17:29:21,146 INFO [main]
> org.apache.hadoop.mapreduce.v2.app.MRAppMaster: Deleting staging directory
> hdfs://host1:port1/user/ABC/.staging/job_1376933101704_0001
> 2013-08-19 17:29:21,156 FATAL [main]
> org.apache.hadoop.mapreduce.v2.app.MRAppMaster: Error starting MRAppMaster
> org.apache.hadoop.yarn.exceptions.YarnRuntimeException:
> java.io.FileNotFoundException: File does not exist:
> hdfs://host1:port1/ABC/.staging/job_1376933101704_0001/job.splitmetainfo
> at
> org.apache.hadoop.mapreduce.v2.app.job.impl.JobImpl$InitTransition.createSplits(JobImpl.java:1469)
> at
> org.apache.hadoop.mapreduce.v2.app.job.impl.JobImpl$InitTransition.transition(JobImpl.java:1324)
> at
> org.apache.hadoop.mapreduce.v2.app.job.impl.JobImpl$InitTransition.transition(JobImpl.java:1291)
> at
> org.apache.hadoop.yarn.state.StateMachineFactory$MultipleInternalArc.doTransition(StateMachineFactory.java:385)
> at
> org.apache.hadoop.yarn.state.StateMachineFactory.doTransition(StateMachineFactory.java:302)
> at
> org.apache.hadoop.yarn.state.StateMachineFactory.access$300(StateMachineFactory.java:46)
> at
> org.apache.hadoop.yarn.state.StateMachineFactory$InternalStateMachine.doTransition(StateMachineFactory.java:448)
> at
> org.apache.hadoop.mapreduce.v2.app.job.impl.JobImpl.handle(JobImpl.java:922)
> at
> org.apache.hadoop.mapreduce.v2.app.job.impl.JobImpl.handle(JobImpl.java:131)
> at
> org.apache.hadoop.mapreduce.v2.app.MRAppMaster$JobEventDispatcher.handle(MRAppMaster.java:1184)
> at
> org.apache.hadoop.mapreduce.v2.app.MRAppMaster.serviceStart(MRAppMaster.java:995)
> at
> org.apache.hadoop.service.AbstractService.start(AbstractService.java:193)
> at
> org.apache.hadoop.mapreduce.v2.app.MRAppMaster$1.run(MRAppMaster.java:1394)
> at java.security.AccessController.doPrivileged(Native Method)
> at javax.security.auth.Subject.doAs(Subject.java:396)
> at
> org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1477)
> at
> org.apache.hadoop.mapreduce.v2.app.MRAppMaster.initAndStartAppMaster(MRAppMaster.java:1390)
> at
> org.apache.hadoop.mapreduce.v2.app.MRAppMaster.main(MRAppMaster.java:1323)
> Caused by: java.io.FileNotFoundException: File does not exist:
> hdfs://host1:port1/ABC/.staging/job_1376933101704_0001/job.splitmetainfo
> at
> org.apache.hadoop.hdfs.DistributedFileSystem$17.doCall(DistributedFileSystem.java:1121)
> at
> org.apache.hadoop.hdfs.DistributedFileSystem$17.doCall(DistributedFileSystem.java:1113)
> at
> org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:78)
> at
> org.apache.hadoop.hdfs.DistributedFileSystem.getFileStatus(DistributedFileSystem.java:1113)
> at
> org.apache.hadoop.mapreduce.split.SplitMetaInfoReader.readSplitMetaInfo(SplitMetaInfoReader.java:51)
> at
> org.apache.hadoop.mapreduce.v2.app.job.impl.JobImpl$InitTransition.createSplits(JobImpl.java:1464)
> ... 17 more
> 2013-08-19 17:29:21,158 INFO [Thread-2]
> org.apache.hadoop.mapreduce.v2.app.MRAppMaster: MRAppMaster received a
> signal. Signaling RMCommunicator and JobHistoryEventHandler.
> 2013-08-19 17:29:21,159 WARN [Thread-2]
> org.apache.hadoop.util.ShutdownHookManager: ShutdownHook
> 'MRAppMasterShutdownHook' failed, java.lang.NullPointerException
> java.lang.NullPointerException
> at
> org.apache.hadoop.mapreduce.v2.app.MRAppMaster$ContainerAllocatorRouter.setSignalled(MRAppMaster.java:805)
> at
> org.apache.hadoop.mapreduce.v2.app.MRAppMaster$MRAppMasterShutdownHook.run(MRAppMaster.java:1344)
> at
> org.apache.hadoop.util.ShutdownHookManager$1.run(ShutdownHookManager.java:54)
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
