[
https://issues.apache.org/jira/browse/HADOOP-13762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15609219#comment-15609219
]
Chris Nauroth commented on HADOOP-13762:
----------------------------------------
See HDFS-11063 for a similar proposal I've created on the HDFS NameNode side.
For example, upon entering a public S3A method, we could append useful
information to the thread name, such as the user, the start time of the
operation and the file path argument. This information would be visible in the
thread names when running {{jstack}}. That way, we could see that not only is
a thread spending a long time in a {{globStatus}} call, but also which user is
making the call, which path is referenced, and the start time.
This proposal would get trickier in combination with some of the plans around
asynchronous and parallel execution. We'd need to pass along that contextual
information to all of the threads that make up the high-level operation.
It's important that after the S3A operation completes, we restore the prior
value of the thread name. S3A is called from user applications that own the
lifecycle of the thread. If applications have set meaningful information into
the thread name already, then we don't want that to remain changed after the
thread exits the S3A code.
> S3A: Set thread names with more specific information about the call.
> --------------------------------------------------------------------
>
> Key: HADOOP-13762
> URL: https://issues.apache.org/jira/browse/HADOOP-13762
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
> Reporter: Chris Nauroth
>
> Running {{jstack}} on a hung process and reading the stack traces is a
> helpful way to determine exactly what code in the process is stuck. This
> would be even more helpful if we included more descriptive information about
> the specific file system method call.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
