[
https://issues.apache.org/jira/browse/ARTEMIS-5139?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17895434#comment-17895434
]
Josh Byster commented on ARTEMIS-5139:
--------------------------------------
In the screenshot, it shows JFR events of waiting to acquire the lock on these
messages, but looking at the GC logs around that time, it's possible that it's
just at a safepoint, capturing a thread dump, and then the thread dump capture
takes 10ms so then we show this as being "blocked" for this time. So possibly a
red herring. I am trying to investigate why we are seeing 10ms+ latency spikes
on Artemis but not on Classic, so it may be another issue.
> Remove locking on TypedProperties getters
> -----------------------------------------
>
> Key: ARTEMIS-5139
> URL: https://issues.apache.org/jira/browse/ARTEMIS-5139
> Project: ActiveMQ Artemis
> Issue Type: Task
> Reporter: Josh Byster
> Priority: Minor
> Attachments: Screenshot 2024-10-30 at 10.38.54 PM.png
>
>
> We are noticing a performance degradation when switching to Artemis in places
> where we are sensitive on the millisecond level. We have confirmed that
> moving to ZGC and having minimal GC pause times cannot explain this spike.
> Looking at a flight recording, it does appear we are spending a lot of time
> blocked on TypedProperties getters (see screenshot). It seems like we can
> probably remove some of the locking here with a concurrent collection or be
> smarter about it with a ReadWriteLock, given that it appears we have threads
> competing for read calls.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
For further information, visit: https://activemq.apache.org/contact
