Noel J. Bergman wrote:
I won't have time to test this before tomorrow night at the earliest, so if someone has time to run a bench test and look for memory leaks, that would be great. With the latest code, I'm seeing my OS reported memory increasing fairly steadily with no plateau so far.
The OS return what the JVM uses. And the JVM in the long term tend to use all of the memory in the Xmx if this helps perfomance. I *always* see my java applications (james, tomcat, jetty, weblogic) growing fairly steadily until they reach the Xmx and start running *full* gc... maybe I don't understand what you're trying to point out. Btw if a leak is there then sooner or later it will throw an OOM (try reducing the Xmx if you prefer sooner to later).
Are you saying that James 2.2.0 was not behaving this way?
I am not saying that there is a memory leak; just that I would like to see memory use level off, and it hasn't so far, so I'm hoping that someone has time to run a test to specifically monitor for a leak. --- Noel
I ran a 12 hour "release" (the default is 24h, but I had no time ;-) ) and I didn't notice any memory problem.
I also profiled a 1 hour session (release ran for 1 hour) with a memory/allocation profiler and I have verified that after one hour and a manual full gc in memory I only had objects created in the first 10 minutes or in the last 5 minutes. Usually I find leaks by looking at objects created in the middle of the test and never released.
My tests are far from being exaustive, but I have no time for further tests, now.
I ran both tests using almost default config.xml (only changed gateway/servername things for postage usage).
Stefano --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]