[
https://issues.apache.org/jira/browse/YARN-9525?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16852026#comment-16852026
]
Adam Antal commented on YARN-9525:
----------------------------------
Turned out I have problems changing the jars in my cluster.
Looks like the fix is good for simple log aggregation (so uploading all the
logs once, when the container is finished), but in the rolling case it throws
the known "append is not supported exception" on s3a. This is an already known
limitation, but since the patch unblocks the regular case on this file format,
I'm +1 pushing this in, if you all agree (non-binding).
For sake of completeness I also attach the stack trace for the rolling case:
{noformat}
Cannot create writer for app application_1559231240331_0001. Skip log upload
this time.
java.io.IOException: java.lang.UnsupportedOperationException: Append is not
supported by S3AFileSystem
at
org.apache.hadoop.yarn.logaggregation.filecontroller.ifile.LogAggregationIndexedFileController.initializeWriter(LogAggregationIndexedFileController.java:228)
at
org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggregation.AppLogAggregatorImpl.uploadLogsForContainers(AppLogAggregatorImpl.java:309)
at
org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggregation.AppLogAggregatorImpl.doAppLogAggregation(AppLogAggregatorImpl.java:476)
at
org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggregation.AppLogAggregatorImpl.run(AppLogAggregatorImpl.java:432)
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.lang.UnsupportedOperationException: Append is not supported by
S3AFileSystem
at
org.apache.hadoop.fs.s3a.S3AFileSystem.append(S3AFileSystem.java:1061)
at org.apache.hadoop.fs.FileSystem.primitiveCreate(FileSystem.java:1237)
at
org.apache.hadoop.fs.DelegateToFileSystem.createInternal(DelegateToFileSystem.java:102)
at
org.apache.hadoop.fs.AbstractFileSystem.create(AbstractFileSystem.java:613)
at org.apache.hadoop.fs.FileContext$3.next(FileContext.java:697)
at org.apache.hadoop.fs.FileContext$3.next(FileContext.java:693)
at org.apache.hadoop.fs.FSLinkResolver.resolve(FSLinkResolver.java:90)
at org.apache.hadoop.fs.FileContext.create(FileContext.java:699)
at
org.apache.hadoop.yarn.logaggregation.filecontroller.ifile.LogAggregationIndexedFileController.initializeWriterInRolling(LogAggregationIndexedFileController.java:297)
at
org.apache.hadoop.yarn.logaggregation.filecontroller.ifile.LogAggregationIndexedFileController.access$700(LogAggregationIndexedFileController.java:97)
at
org.apache.hadoop.yarn.logaggregation.filecontroller.ifile.LogAggregationIndexedFileController$1.run(LogAggregationIndexedFileController.java:199)
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:176)
... 7 more
{noformat}
> 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: Bug
> Components: log-aggregation
> Affects Versions: 3.1.2
> Reporter: Adam Antal
> Assignee: Peter Bacsko
> Priority: Major
> Attachments: IFile-S3A-POC01.patch, YARN-9525-001.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
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
