We've had frequent problems with perm-gen space OOME's in glassfish when deploying without restarting the server, so our deploy process is:
- undeploy app
- stop domain
- start domain
- deploy new .war
That said, about 90% of the time, glassfish dies (with a diff OOME) immediately after the deploy, despite the fact it was just restarted. However, when we start glassfish again (with the app already deployed from before the crash), everything is fine.
Attached is an hs_err file from the most recent incident. Can anybody help interpret this file to help us determine what caused the OOME, and why it's only a problem after deploying but not when the server starts up with the app already deployed?
JVM-args can be seen in the hs_err file, but to save you the time, here's the relevant ones as far as I know:
-server
-Xmx3000m
-Xmn1000m
-Xss128k
-XX:+AggressiveOpts
-XX:+AggressiveHeap
-XX:+DisableExplicitGC
-XX:+UseParallelGC
-XX:+UseParallelOldGC
-XX:ParallelGCThreads=8
-XX:MaxPermSize=192m
-XX:NewRatio=2
Here's the except on the heap:
PSYoungGen total 864512K, used 156081K [0xb6400000, 0xf4c00000, 0xf4c00000)
eden space 708416K, 0% used [0xb6400000,0xb6400000,0xe17d0000)
from space 156096K, 99% used [0xe17d0000,0xeb03c560,0xeb040000)
to space 159488K, 0% used [0xeb040000,0xeb040000,0xf4c00000)
ParOldGen total 200704K, used 184598K [0x39400000, 0x45800000, 0xb6400000)
object space 200704K, 91% used [0x39400000,0x44845ac0,0x45800000)
PSPermGen total 135168K, used 78034K [0x2d400000, 0x35800000, 0x39400000)
object space 135168K, 57% used [0x2d400000,0x320349f8,0x35800000)
The "from space" seems to be the culprit, though given that, I'm not sure what to think next.
[Message sent by forum member 'rwillie6']
http://forums.java.net/jive/thread.jspa?messageID=480154