Derby is not being used with the default configuration; at least I have deleted the derby jar file and run with no problem. The JVM parameter is just specifying where derby should write files, if it happens to be called.
The JMX figures look like EEG scans of an AMQ broker in dream state :) Eddie On Thu, Jun 19, 2008 at 7:36 AM, Marshall Schor <[EMAIL PROTECTED]> wrote: > Marshall Schor wrote: > >> Marshall Schor wrote: >> >>> Marshall Schor wrote: >>> >>>> After running some tests on uima-as on a clean windows machine, I >>>> connected a jmx console to the broker and watched it "idle" (i.e., no uima >>>> activity was running) for several hours, overnight, and can see there is no >>>> memory leaks. There is some behavior that I don't understand, though - I'm >>>> wondering if anyone has delved into this and can explain what these screen >>>> shots show (this is the first time I've attempted to paste screen shots >>>> into >>>> mail to the dev list - so apologies if these don't come thru - we'll see). >>>> >>> I see the screen shots don't come thru. I'll see if I can post to the >>> wiki... >>> >> Wiki did not accept screen shot either... Trying another mail client... >> > Here's the link to the wiki after I uploaded individually the images for > the screenshots. Would be nice if there was an easier way to do this... > > http://cwiki.apache.org/UIMA/testing-activemq-broker.html > > -Marshall > > >>> -Marshall >>> >>>> >>>> First - the vm-args seem to indicate that maybe Derby is being used - >>>> not sure - but I had thought derby was 'disabled' by our default >>>> configuration? >>>> >>>> >>>> >>>> Then, here are some JMX Summary screens, over 3 time spans - a few >>>> minutes, an hour, and overnight. The memory used drops after maybe about >>>> 1/2 an hour, coincident with some classes being unloaded. The number of >>>> live threads goes up and down on a regular basis. >>>> >>>> >>>> >>>> >>>> >>>> >>>> >>>> -Marshall >>>> >>>> >>>> >>> >>> >>> >> >> >> >
