[
https://issues.apache.org/jira/browse/CAMEL-25066?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
shashank reassigned CAMEL-25066:
--------------------------------
Assignee: shashank
> camel-core - Producer cache eviction can stop a dynamic endpoint that another
> producer cache still uses
> -------------------------------------------------------------------------------------------------------
>
> Key: CAMEL-25066
> URL: https://issues.apache.org/jira/browse/CAMEL-25066
> Project: Camel
> Issue Type: Improvement
> Components: camel-core
> Reporter: Claus Ibsen
> Assignee: shashank
> Priority: Minor
>
> Endpoints are shared per CamelContext, but producer caches are not. toD
> resolves its URI via context.getEndpoint(uri), and the endpoint registry
> caches dynamic endpoints, so two toD (or recipientList, routingSlip,
> ProducerTemplate) that resolve the same URI get the same Endpoint instance.
> Each of them has its own DefaultProducerCache (SendDynamicProcessor creates
> one per processor).
> When one cache evicts its producer, ServicePool (SinglePool.cleanupEvicts)
> stops the endpoint unless isEndpointInUse() finds it in the static endpoint
> registry or as a route consumer. It does not know about other producer
> caches, so the endpoint is stopped while another cache still holds a started
> producer for it.
> Example: two routes both do toD("seda:${header.region}"). Route A has a small
> cacheSize and sees many regions, so it evicts seda:eu and stops that endpoint
> while route B's cache still sends to it.
> This is not a regression: before CAMEL-25004 the endpoint was always stopped
> on eviction. Possible fixes: reference-count endpoint usage across producer
> caches, or leave stopping dynamic endpoints to the endpoint registry's own
> eviction.
> Follow-up from the review of https://github.com/apache/camel/pull/26862.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
