[
https://issues.apache.org/jira/browse/APEXCORE-330?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15271358#comment-15271358
]
Pramod Immaneni commented on APEXCORE-330:
------------------------------------------
I noticed that the individual containers are sending a JSON serialized string
of the stack trace information to STRAM. The serialization is typically handled
by RPC and the containers just send objects. Could you convert the Map<Thread,
StackTraceElement[]> into an object that can be serialized, maybe into a
Map<ThreadInfo, StackTraceElement[]>.
Secondly the web service call results in a request flag to be set in the
container agent which gets reset immediately after initiating a request to the
container to send the stack trace. I believe this flag should be reset after
receiving the stack trace from the container as there is a chance the request
may not make it to the container if there are any intermittent network errors
and the web service call is anyway waiting for a timeout period of 10s for the
stack trace which is several heartbeats.
I apologize I was not able to give these comments during the pull request
review.
> Ability to obtain a thread dump from a container
> ------------------------------------------------
>
> Key: APEXCORE-330
> URL: https://issues.apache.org/jira/browse/APEXCORE-330
> Project: Apache Apex Core
> Issue Type: Improvement
> Reporter: Thomas Weise
> Assignee: Sandesh
> Fix For: 3.4.0
>
>
> To analyze issues such as "stuck operator", it is useful to obtain stats from
> the running JVM. Often needed are GC stats and thread dump. In production
> environments users often don't have direct access to the machines, making it
> available through the REST API will help.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
