[ 
https://issues.apache.org/jira/browse/YARN-9525?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16991592#comment-16991592
 ] 

Adam Antal commented on YARN-9525:
----------------------------------

Validated that this patch on top of YARN-9607 actually fixes the bug.

Log aggregation using the IndexedFileController targeting an S3a URI as 
remote-app-folder works in simple and in the rolling case.

As the latest jenkins result was green, could you commit this [~snemeth] if 
there are no more concerns?

> IFile format is not working against s3a remote folder
> -----------------------------------------------------
>
>                 Key: YARN-9525
>                 URL: https://issues.apache.org/jira/browse/YARN-9525
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: log-aggregation
>    Affects Versions: 3.1.2
>            Reporter: Adam Antal
>            Assignee: Adam Antal
>            Priority: Major
>         Attachments: IFile-S3A-POC01.patch, YARN-9525-001.patch, 
> YARN-9525.002.patch, YARN-9525.003.patch, YARN-9525.004.patch, 
> YARN-9525.005.patch, YARN-9525.006.patch
>
>
> Using the IndexedFileFormat {{yarn.nodemanager.remote-app-log-dir}} 
> configured to an s3a URI throws the following exception during log 
> aggregation:
> {noformat}
> Cannot create writer for app application_1556199768861_0001. Skip log upload 
> this time. 
> java.io.IOException: java.io.FileNotFoundException: No such file or 
> directory: 
> s3a://adamantal-log-test/logs/systest/ifile/application_1556199768861_0001/adamantal-3.gce.cloudera.com_8041
>       at 
> org.apache.hadoop.yarn.logaggregation.filecontroller.ifile.LogAggregationIndexedFileController.initializeWriter(LogAggregationIndexedFileController.java:247)
>       at 
> org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggregation.AppLogAggregatorImpl.uploadLogsForContainers(AppLogAggregatorImpl.java:306)
>       at 
> org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggregation.AppLogAggregatorImpl.doAppLogAggregation(AppLogAggregatorImpl.java:464)
>       at 
> org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggregation.AppLogAggregatorImpl.run(AppLogAggregatorImpl.java:420)
>       at 
> org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggregation.LogAggregationService$1.run(LogAggregationService.java:276)
>       at 
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
>       at 
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
>       at java.lang.Thread.run(Thread.java:748)
> Caused by: java.io.FileNotFoundException: No such file or directory: 
> s3a://adamantal-log-test/logs/systest/ifile/application_1556199768861_0001/adamantal-3.gce.cloudera.com_8041
>       at 
> org.apache.hadoop.fs.s3a.S3AFileSystem.s3GetFileStatus(S3AFileSystem.java:2488)
>       at 
> org.apache.hadoop.fs.s3a.S3AFileSystem.innerGetFileStatus(S3AFileSystem.java:2382)
>       at 
> org.apache.hadoop.fs.s3a.S3AFileSystem.getFileStatus(S3AFileSystem.java:2321)
>       at 
> org.apache.hadoop.fs.DelegateToFileSystem.getFileStatus(DelegateToFileSystem.java:128)
>       at org.apache.hadoop.fs.FileContext$15.next(FileContext.java:1244)
>       at org.apache.hadoop.fs.FileContext$15.next(FileContext.java:1240)
>       at org.apache.hadoop.fs.FSLinkResolver.resolve(FSLinkResolver.java:90)
>       at org.apache.hadoop.fs.FileContext.getFileStatus(FileContext.java:1246)
>       at 
> org.apache.hadoop.yarn.logaggregation.filecontroller.ifile.LogAggregationIndexedFileController$1.run(LogAggregationIndexedFileController.java:228)
>       at java.security.AccessController.doPrivileged(Native Method)
>       at javax.security.auth.Subject.doAs(Subject.java:422)
>       at 
> org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1730)
>       at 
> org.apache.hadoop.yarn.logaggregation.filecontroller.ifile.LogAggregationIndexedFileController.initializeWriter(LogAggregationIndexedFileController.java:195)
>       ... 7 more
> {noformat}
> This stack trace point to 
> {{LogAggregationIndexedFileController$initializeWriter}} where we do the 
> following steps (in a non-rolling log aggregation setup):
> - create FSDataOutputStream
> - writing out a UUID
> - flushing
> - immediately after that we call a GetFileStatus to get the length of the log 
> file (the bytes we just wrote out), and that's where the failures happens: 
> the file is not there yet due to eventual consistency.
> Maybe we can get rid of that, so we can use IFile format against a s3a target.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org

Reply via email to