Adam Antal created YARN-9525:
--------------------------------
Summary: TFile 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
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: [email protected]
For additional commands, e-mail: [email protected]
