[ 
https://issues.apache.org/jira/browse/CAMEL-26915?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18144679#comment-18144679
 ] 

shashank commented on CAMEL-26915:
----------------------------------

Correction to the description: a route stop and start (or suspend and resume) 
does not hit this, because the route creates a new consumer each time, so the 
old task can never write the fields of the new run. The case is a route policy 
that stops and starts the same consumer instance, such as the throttling route 
policies, the Hazelcast route policy, ScheduledRoutePolicy and the DNS route 
policy. The test of the fix uses that path.

_Claude Code on behalf of allthingssecurity_

> camel-nats - a quick stop and start of a consumer: the task of the stopped 
> run can subscribe and take over the fields of the new run
> ------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: CAMEL-26915
>                 URL: https://issues.apache.org/jira/browse/CAMEL-26915
>             Project: Camel
>          Issue Type: Bug
>          Components: camel-nats
>            Reporter: shashank
>            Assignee: shashank
>            Priority: Minor
>
> Follow-up of CAMEL-26462 and CAMEL-26884, the gap left open by both (listed 
> in their PRs, #29042 and #29788).
> {{NatsConsumer}} subscribes on a task of its executor, and the task checks 
> {{isRunAllowed()}} to see whether the consumer stopped meanwhile. That is the 
> state of the consumer, not of a run. On a quick stop and start (for example a 
> route restart) while the task of the first run is still in its subscribe:
> * the stop of the first run does not see the dispatcher or the subscription 
> the task is about to set, and the interrupt of {{shutdownNow}} does not 
> always end the subscribe;
> * when the subscribe returns, the consumer runs again, so the task of the old 
> run sees {{isRunAllowed()}} true, keeps its subscription and writes 
> {{dispatcher}}, {{dispatcherHandler}} and {{jetStreamSubscription}}, which 
> may overwrite those of the new run;
> * the next stop then ends only what the fields point at: the dispatcher or 
> subscription of the other run is never closed, and the route can receive each 
> message twice (a standard or push subscription of each run, or two pull fetch 
> loops).
> Also, after the task of a run gave up (its subscribe failed), the fields 
> still point at the closed dispatcher of that run until the next run sets them.
> Expected: each start of the consumer gets a run token (a counter or an 
> object), the task works on the token of its run, it sets the fields only 
> while its run is the current one, and it ends what it made itself when its 
> run is no longer current (with the interrupt cleared for the unsubscribe, as 
> in CAMEL-26884). A stop clears the fields of its own run.
> Not seen in production; found while reviewing the stop path for CAMEL-26462 
> and CAMEL-26884. The window is a subscribe in flight across a stop and a 
> start.
> _Filed with Claude Code on behalf of allthingssecurity._



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

Reply via email to