I'm trying to collect a java core on AIX (java version 1.6). There is only one dump agent configured, which collects a system dump, heap dump and java core on every SIGQUIT, so I used HealthCenter to request only a javacore. That worked fine, it didn't kick off a heap dump or system dump, but it caused the WebLogic (I know) server to hang for 172s (seconds, not ms) while it dumped out, and the resulting file was 88MB. I've no idea what could cause a javacore dump to take so long and why it would be so big.
Slide 3 in this presentation states that it should take a few hundred ms and be around 2MB, which is very similar to what I see with HotSpot. I know it's talking about Java 5, but a single major version increase couldn't make such a massive difference!
What could be the cause of it taking so long? Is there anything I could do to speed it up?
FYI: the heap size is 16GB, but it was less than half full when the dump was taken. There's also no native memory information in the dump and using IBM TMDA refuses to show render the thread detail pane properly. There are around 310 threads in each dump I've taken.
