Help needed!! Novice with Garbage Collection problems.

Question

Hi Guys,
Really hoping somebody can help me here. I am a relative novice when it comes to all things Java but i am slowly trying to learn. I have come across an issue which i have identified but i am just not sure what to do about it.
Ok, in a nut shell the issue seems to be revolving around the frequency of garbage collection. From the default-err.log file i am seeing (on average) an Allocation Failure occur every 2 secs. Here is a sample from the log with verbose:gc active:
<AF[4986]: Allocation Failure. need 208480 bytes, 78 ms since last AF>
<AF[4986]: managing allocation failure, action=2 (559165976/1342176248)>
<GC: Mon Oct 11 11:51:12 2004
<GC(4986): freed 4101528 bytes in 1559 ms, 41% free (563267504/1342176248)>
<GC(4986): mark: 1301 ms, sweep: 258 ms, compact: 0 ms>
<GC(4986): refs: soft 0 (age >= 32), weak 0, final 0, phantom 0>
<AF[4986]: completed in 1563 ms>
<AF[4987]: Allocation Failure. need 208536 bytes, 78 ms since last AF>
<AF[4987]: managing allocation failure, action=2 (559138336/1342176248)>
<GC: Mon Oct 11 11:51:14 2004
<GC(4987): freed 4105128 bytes in 1563 ms, 41% free (563243464/1342176248)>
<GC(4987): mark: 1293 ms, sweep: 270 ms, compact: 0 ms>
<GC(4987): refs: soft 0 (age >= 32), weak 0, final 0, phantom 0>
<AF[4987]: completed in 1563 ms>
As you can see, allocation failures are occuring all the time, and with 2secs between events, and each GC taking around 1.5secs, i am having massive problems with the response of the server. It seems that the javaw.exe process is just pegged at 100% CPU the whole time and then it will eventually grind to a halt, and the users will get terrible response times.
OK, the questions are:
- Even to me (a novice) the above extract from the log doesnt look good. Am i right?
- what would be causing this? (i know - how long is a peice of string but i am hoping somebody can point me in the right direction so i can look some more)
- what can i do about it? Is there any parameters i can put into the java args to help me out?
Currently i am running -Xms of 128meg and an -Xmx of 1024m with no other settings. There are at the moment about 200users logged onto this server concurrently, and it seems to die a couple of hours into them all being logged on. I then have to kick everybody out and reboot to get it in a working state again.
I am in some serious need of help from some gurus!! any help would be invaluable, thanks heaps guys.
Tim

Answer 1

Hi Again guys, thanks for all your replies.
I have been working my butt of on this issue and i just cant seem to get anywhere... probably due to my complete lack of knowledge on this whole GC thing! :)
One thing that i have noticed that seems to be very consistant is the fact that as soon as i get an "action=2" in my default-err.log from an allocation failure, thats when the system goes nuts and tends to not recover.
At all other times it is an action=1, and the system seems to be running OK, but as soon as i get action=2, the time between GC events drop from seconds to miliseconds, and the bytes required just skyrockets. It keeps on this upward spiral till i just have to reboot the box.
From all my reading, i have found that an action=2 means "2 - The Garbage Collector has tried to allocate out of the wilderness, and failed."
This is the only item in the logs that i see is directly related to the server performance.
Can anybody please explain to me (in laymans terms) what the action=2 means, what may cause it and what i should be looking at changing to fix it?
The extract from the log files (in my first post in this thread) are still valid, as are the min/max memory settings.
Any help at all would be invaluable.
Thanks very much.
Tim