[
https://issues.apache.org/jira/browse/ARTEMIS-2670?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17063693#comment-17063693
]
Domenico Bruscino commented on ARTEMIS-2670:
--------------------------------------------
I can't reproduce this issue against Artemis 2.11.0 because this issue was
already fixed by Clebert with ARTEMIS-2592.
Before the fix, the QueueImpl.removeConsumer locks the queue until the deleting
of auto-create queue.
The deleting of a queue calls ManagementServiceImpl.unregisterQueue that is
synchronized and this can cause the deadlock.
After the fix, QueueManagerImpl has an executor so the deleting of auto-create
queue isn't executed in the same thread of QueueImpl.removeConsumer and this
prevents the deadlock.
> Deadlock when running in cluster mode
> -------------------------------------
>
> Key: ARTEMIS-2670
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2670
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Affects Versions: 2.9.0
> Reporter: Luis Miguel De Bello
> Priority: Major
> Attachments: Deadlock1, Deadlock2
>
>
> We are running Artemis 2.9.0 in production (AWS) with 4 instances creating a
> cluster (Jgroup) it usually works ok and was working fine but last week one
> of the
> instances got unhealthy, when login into the instance we noticed there was a
> deadlock and lot of connection in CLOSE_WAIT state. I am assuming the
> connection in CLOSE_WAIT are generated by this deadlock.
> I was checking the releases note for 2.10.0 - 2.10.1 - 2.11.0 and I read
> about two deadlock fixes but I don't think it is the same case.
> Our clients connect to the broker using Secure Web Socket (Mutual TLS).
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
