[ 
https://issues.apache.org/jira/browse/CAMEL-27339?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Work on CAMEL-27339 started by shashank.
----------------------------------------
> camel-jms, camel-sjms - a request that passed the running check before the 
> producer stopped creates a reply manager after the stop, which is never 
> stopped
> ----------------------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: CAMEL-27339
>                 URL: https://issues.apache.org/jira/browse/CAMEL-27339
>             Project: Camel
>          Issue Type: Bug
>          Components: camel-jms, camel-sjms
>            Reporter: shashank
>            Assignee: shashank
>            Priority: Major
>
> h3. What happens
> {{JmsProducer.process}} and {{SjmsProducer.process}} check {{isRunAllowed()}} 
> and then call {{initReplyManager()}} for an InOut exchange. The stop does not 
> wait for a request between the two: {{doStop}} stops the reply manager and 
> sets {{started=false}}. A request that passed the check before the stop then 
> finds {{started=false}}, takes the producer's lock (the same lock 
> {{BaseService.stop}} holds, so it waits for the stop to end) and creates and 
> starts a new reply manager for the stopped producer: a JMS connection and 
> listener (a temporary queue by default), a timeout checker thread and an 
> onTimeout thread pool. Nothing stops it. If the producer is started again, 
> the next request finds {{started=true}} and uses that reply manager, and the 
> next stop stops it, so the leak lasts until then; a producer that is not 
> started again (a removed route, a stopped {{ProducerTemplate}}) leaks it for 
> good.
> h3. How to reproduce
> Hold the request where it reads the {{CamelJmsDestinationName}} header (after 
> the running check, before {{initReplyManager}}), stop the producer, and let 
> the request continue: a reply manager is created after the stop.
> Found with a TLA+ model of the reply manager and the producer stop.
> h3. Fix
> {{initReplyManager}} checks {{isRunAllowed()}} again under the producer's 
> lock and fails the request with {{RejectedExecutionException}}, as 
> {{process}} does for a stopped producer, instead of creating a reply manager. 
> Independent of CAMEL-27321 (every request completes when the reply manager 
> stops), which does not touch {{initReplyManager}}.
> _Filed with Claude Code on behalf of allthingssecurity._



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

Reply via email to