[
https://issues.apache.org/activemq/browse/CAMEL-2716?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=59321#action_59321
]
Hadrian Zbarcea commented on CAMEL-2716:
----------------------------------------
r943644 - first cut.
Also added a way to reset the counters to have predictable ObjectNames during
tests. DefaultNodeIdFactory.resetAllCounters() could lead to name conflicts if
used without care, reason why I made it protected, so it cannot be invoked
directly.
> Incomplete cleanup of jmx mbeans after context stop
> ---------------------------------------------------
>
> Key: CAMEL-2716
> URL: https://issues.apache.org/activemq/browse/CAMEL-2716
> Project: Apache Camel
> Issue Type: Bug
> Affects Versions: 2.2.0
> Reporter: Hadrian Zbarcea
> Assignee: Hadrian Zbarcea
>
> There is a fair amount of cleanup on context stop, but it is incomplete. Most
> of it has to do with JMX and if one creates and starts a new CamelContext she
> may end up with managed objects pointing nowhere or have duplicates or other
> minor side-effects. One of the nasty side effects is that we are pretty much
> forced to use "<forkMode>pertest</forkMode>" in our surefire plugin
> configuration, ensuring a clean slate for every test.
> For long running apps in which you may have multiple CamelContext running at
> various times during the lifetime of the application we must properly cleanup
> when stopping the context. I have a series of patches fixing that which I
> hope to finish before 2.3. Once we're comfortable and tested with multiple
> jdk implementations and versions, I'd change the forkMode to "once". In my
> local copy this reduces the test time in core from 40+ mins to under 15 mins.
> Since interactions between tests is still possible and might lead us on a
> wild-goose we could add another profile that uses the safer yet time
> consuming "pertest" forkMode.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
