[ 
https://issues.apache.org/jira/browse/CAMEL-27333?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Work on CAMEL-27333 started by shashank.
----------------------------------------
> camel-sjms - after the temporary reply queue's connection fails, an InOut 
> request waits for the new queue with no timeout, and for ever when the 
> producer stops
> ---------------------------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: CAMEL-27333
>                 URL: https://issues.apache.org/jira/browse/CAMEL-27333
>             Project: Camel
>          Issue Type: Bug
>          Components: camel-sjms
>            Reporter: shashank
>            Assignee: shashank
>            Priority: Major
>
> h3. What happens
> With a temporary reply queue, when the connection of the reply listener 
> fails, {{TemporaryReplyQueueExceptionListener}} marks the temporary queue for 
> a refresh. Every InOut request then waits in 
> {{TemporaryQueueReplyManager.getReplyTo()}} ({{destinationReady}}) until the 
> connection recovery creates the new temporary queue. The wait is 
> {{condition.await()}} without a timeout:
> * when the producer is stopped during the wait, the listener container stops 
> and cancels the recovery, and nothing signals the condition: the request 
> thread waits for ever, holding the JMS connection and session opened by 
> {{SjmsTemplate.send}}, and the exchange never completes. The request is not 
> registered yet, so neither its requestTimeout nor the stop of the reply 
> manager completes it;
> * until the recovery has created the new temporary queue (it starts 1 s after 
> the failure, then retries every recoveryInterval), every InOut request blocks 
> in that wait, even when the broker is reachable again, instead of failing 
> after requestTimeout.
> The await has no loop either, so a spurious wakeup lets a request go on with 
> the old queue.
> h3. How to reproduce
> After a first InOut request (which creates the reply manager), trigger the 
> refresh of the temporary queue without a recovery, send an InOut request, and 
> stop the producer: the request thread stays in {{destinationReady}} (WAITING) 
> and its callback is never called.
> Found with a TLA+ model of the sjms request/reply.
> h3. Fix
> Wait in a loop, at most the endpoint's requestTimeout, and stop waiting when 
> the reply manager stops ({{TemporaryQueueReplyManager.doStop}} signals the 
> waiters before the listener container stops). When the queue is not ready, 
> {{getReplyTo()}} throws, and the producer completes the exchange with that 
> exception without sending the request.
> _Filed with Claude Code on behalf of allthingssecurity._



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

Reply via email to