-server does not cause incremental gc. Use -incgc for that.

Also keep in mind that with gc you can either have efficient and erratic
or inefficient and predictable, it's always a trade off. So incgc will have
shorter pauses, but more of them, so the total time paused will be
greater than normal gc. concurrent gc will have even shorter pauses,
but will use far more cpu. Generally, don't use concurrent gc unless you're
on an SMP system with a program. Only use incgc if you can't
tolerate the pauses to collect garbage.


Also, garbage collection time is entirely determined by the number of
live objects. A larger heap will make gc occur less frequently, but it
won't be any faster, or slower when it does happen.

Also, keep in mind that if you use the -server option, then the vm will
try to use all of its heap to avoid pauses. So it will slowly fill up the heap,
and then do a gc when it starts getting low on space, so don't give it too much
heap. I've seen it slowly fill a gig of space and then gc it down to less than 100
MB. The client VM tries to use minimal space, but the server tries to use all
the space you give it for maximal performance.




-Tyler
[EMAIL PROTECTED]





On Feb 18, 2004, at 6:56 AM, Kevin Z Grey wrote:

Heya,



The JVM has the following command line options to tweak the heap size:




-Xms


initial java heap size


-Xmx


maximum java heap size


-Xmn


the size of the heap for the young generation



e.g.
-Xmn100M -Xms500M -Xmx500M -install



While this does not fix memory leaks, it will allow you to put a cap on
the memory usage.  I believe when the max is hit, it will initiate gc.
There is also a -server option which I think will cause gc to occur
incrementally.



~~K



-----Original Message-----
From: J Malcolm [mailto:[EMAIL PROTECTED]
Sent: Wednesday, February 18, 2004 12:44 AM
To: James Users List
Subject: Memory Leaks?



Recently, I have been noticing my memory utilization creeping up by
about 10MB per hour. I've got James, Apache, Tomcat, BIND, and MySQL
running on my server (WinXP). I've tried restarting each service one at
a time. James appears to be the culprit. I can drop 80MB of committed
memory by recycling James. Then the creep slowly starts again.


Is there anything I could have done in James configuration that would
cause memory to be allocated and never returned?  Does this sound like
anything that is known?

I'm running James 2.1 with a file system based mail store. No special
mailets other than the ones shipped with the base package. Sun JRE 1.4.
XPPro with all the latest patches.


Thanks.

Jerry

_____





---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]



Reply via email to