[
https://issues.apache.org/jira/browse/FLINK-25398?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17463067#comment-17463067
]
Junfan Zhang edited comment on FLINK-25398 at 12/21/21, 8:26 AM:
-----------------------------------------------------------------
[~xtsong] Thanks for your quick reply
>From my perspective, this feature makes sense. Usually we use thread dump in
>runtime ui to debug the Flink job and analyze which calling the job stucked.
>However due to stacktrace frame depth limitation, I have to login to
>nodemanager to check the complete thread info. It's a terriable experience.
Hence when debugging the problems, i think the cost of dumping all complete
stack is deserved.
Do we need to introduce extra user config to enable stack depth limit? Maybe it
should be determinzed by user?
was (Author: zuston):
[~xtsong] Thanks for your quick reply
>From my perspective, this feature makes sense. Usually we use thread dump in
>runtime ui to debug the Flink job and analyze which calling the job stucked.
>However due to stacktrace frame depth limitation, I have to login to
>nodemanager to check this flink thread info. It's a terriable experience.
Hence when debugging the problems, i think the cost of dumping all complete
stack is deserved.
Do we need to introduce extra user config to enable stack depth limit? Maybe it
should be determinzed by user?
> Show complete stacktrace when requesting thread dump
> ----------------------------------------------------
>
> Key: FLINK-25398
> URL: https://issues.apache.org/jira/browse/FLINK-25398
> Project: Flink
> Issue Type: Improvement
> Reporter: Junfan Zhang
> Priority: Major
> Labels: pull-request-available
> Attachments: stacktrace.png
>
>
> h2. Why
> Now the stacktrace is not complete when clicking the task executor's
> threaddump in runtime webui. Hence it's hard to the initial calling
> according to the stacktrace.
> Now the thread stacktrace is limited to 8, refer to openjdk:
> [https://github.com/openjdk/jdk/blob/master/src/java.management/share/classes/java/lang/management/ThreadInfo.java#L597]
>
> h2. Solution
> Using the custom {{stringify}} method to return stacktrace instead of using
> {{ThreadInfo.toString}} directly
>
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
