[
https://issues.apache.org/activemq/browse/SM-1868?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=52228#action_52228
]
Freeman Fang commented on SM-1868:
----------------------------------
Hi Maciej,
Maybe this situation is different with SM-1824. But the fix for SM-1824 will
aslo take care this issue.
If you've read SM-1824, you will know the fix for the trunk(3.3.1 snapshot) now
is that the Flow.suspend() will never be invoked which means the
writeLock().lock() will never be invoked, so what you mentioned here
"deployment of another SA is started - it calls suspend on SedaFlow, acquires
SedaFlow object as monitor and waits for writeLock"
will never happen.
Did you ever try the 3.3.1 SNAPSHOT? I'm quite sure it will address your
problem.
Freeman
> Deadlock using SedaFlow & Camel Component during SMX start
> ----------------------------------------------------------
>
> Key: SM-1868
> URL: https://issues.apache.org/activemq/browse/SM-1868
> Project: ServiceMix
> Issue Type: Bug
> Components: servicemix-core
> Affects Versions: 3.3
> Environment: JDK 1.5, SedaFlow
> Reporter: Maciej Prochniak
>
> The scenario is as follows:
> - http request is sent to HttpEndpoint
> - it's sent to Component A which uses push delivery (e.g. Camel) .
> - the camel component processes request - which holds readLock on SedaFlow as
> push delivery is used.
> - deployment of another SA is started - it calls suspend on SedaFlow,
> acquires SedaFlow object as monitor and waits for writeLock
> - Component A sends request to Component B (ODE in our case). The SedaQueue
> object for Component B is not created yet, so createQueue method in SedaFlow
> is called, but it's synchronized on the flow object - and deadlock appears
> Note: this is different situation than in
> http://issues.apache.org/activemq/browse/SM-1824
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
