[
https://issues.apache.org/jira/browse/YARN-10855?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17381791#comment-17381791
]
Qi Zhu commented on YARN-10855:
-------------------------------
Thanks [~Jim_Brennan] for patch.
If we can change the LogAggregationTFileController resource close in finally to
try with resource.
It's a minor advice, the patch LGTM generally.
> yarn logs cli fails to retrieve logs if any TFile is corrupt or empty
> ---------------------------------------------------------------------
>
> Key: YARN-10855
> URL: https://issues.apache.org/jira/browse/YARN-10855
> Project: Hadoop YARN
> Issue Type: Bug
> Components: yarn
> Affects Versions: 3.2.2, 2.10.1, 3.4.0, 3.3.1
> Reporter: Jim Brennan
> Assignee: Jim Brennan
> Priority: Major
> Attachments: YARN-10855.001.patch
>
>
> When attempting to retrieve yarn logs via the CLI command, it failed with the
> following stack trace (on branch-2.10):
> {noformat}
> yarn logs -applicationId application_1591017890475_1049740 > logs
> 20/06/05 19:15:50 INFO client.RMProxy: Connecting to ResourceManager
> 20/06/05 19:15:51 INFO client.AHSProxy: Connecting to Application History
> server
> Exception in thread "main" java.io.EOFException: Cannot seek to negative
> offset
> at org.apache.hadoop.hdfs.DFSInputStream.seek(DFSInputStream.java:1701)
> at
> org.apache.hadoop.fs.FSDataInputStream.seek(FSDataInputStream.java:65)
> at org.apache.hadoop.io.file.tfile.BCFile$Reader.<init>(BCFile.java:624)
> at org.apache.hadoop.io.file.tfile.TFile$Reader.<init>(TFile.java:804)
> at
> org.apache.hadoop.yarn.logaggregation.AggregatedLogFormat$LogReader.<init>(AggregatedLogFormat.java:503)
> at
> org.apache.hadoop.yarn.logaggregation.LogCLIHelpers.dumpAllContainersLogs(LogCLIHelpers.java:227)
> at org.apache.hadoop.yarn.client.cli.LogsCLI.run(LogsCLI.java:333)
> at org.apache.hadoop.yarn.client.cli.LogsCLI.main(LogsCLI.java:367)
> {noformat}
> The problem was that there was a zero-length TFile for one of the containers
> in the application aggregated log directory in hdfs. When we removed the
> zero length file, {{yarn logs}} was able to retrieve the logs.
> A corrupt or zero length TFile for one container should not prevent loading
> logs for the rest of the application.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
