Remy Maucherat wrote:
Filip Hanik - Dev Lists wrote:
much appreciated, I'm still working on the "deregistering" of the JMX
resources, somehow they are slipping through the cracks
Originally, the purpose of the JMX stuff was to provide thread
monitoring. Now that the processors are not tied to threads anymore,
it becomes a bit useless IMO. Normally, Java 5+ provides JMX
monitoring of threads already, but of course the Tomcat JMX does
provide a little more data (that may or may not be worth keeping).
I've done the fix for the NIO connector. I believe I have been able to
take care of all the problems.
In the NIO protocol, I simply got rid of the thread local, and replaced
it with a queue.
This queue is able to deregister objects, should it be needed.
Only time they are deregistered, is if they number exceed the
processorCache=<int value for max number> in the NIO <Connector> element.
The memory leak exists in the APR and JIO connector, when and only when
they use the shared <Executor> and the Executor has a
minSpareThreads<maxThreads.
Since these connectors do not use an executor by default, I'd like to
get some feedback if these two need to get fixed for the upcoming release.
If we are to proceed with the release without fixing these, I would
simply note it in the release notes, that to use an executor, use the
workaround.
I'm open to suggestions...and once again...thanks for your patience
(especially since it's me that want this new release)
Filip
Rémy
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
