views:

430

answers:

1

We have problems with JVMs running internet applications under changing load. This problem comes and goes. One day we see three VMs dying and after that, there is nothing for a week or two. We have not found a pattern yet, found nothing to reproduce or cause it. Also a search in the Sun bug database did not help.

We tried a suggested workaround (-XX:-CMSPermGenPrecleaningEnabled -XX:-CMSConcurrentMTEnabled) from http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6354939 Did not help. It seem to just change the thread that causes it... or at least make us believe so.

Also an upgrade to 1.6.0_13 did not help and the bug request to Sun never returned a response.

So my question is, has anyone seen that or has an idea what to look for? Could this be related to OS libraries?

Edited: OS is Linux, OpenSuse running on AMD cpus (Linux 2.6.18.8-0.1-dw #3 SMP Thu Mar 15 01:21:48 GMT 2007 x86_64 x86_64 x86_64 GNU/Linux)

#
# An unexpected error has been detected by Java Runtime Environment:
#
# SIGSEGV (0xb) at pc=0x062c75f5, pid=6667, tid=1090374560
#
# Java VM: Java HotSpot(TM) Server VM (11.2-b01 mixed mode linux-x86)
# Problematic frame:
# V [libjvm.so+0x2c75f5]
#
# If you would like to submit a bug report, please visit:
# http://java.sun.com/webapps/bugreport/crash.jsp
#

--------------- T H R E A D ---------------

Current thread (0x081ddc00): ConcurrentGCThread [stack: 0x40f5c000,0x40fdd000] [id=6679]

siginfo:si_signo=SIGSEGV: si_errno=0, si_code=1 (SEGV_MAPERR), si_addr=0x0000000c

Registers:
EAX=0x00000000, EBX=0x00000008, ECX=0x0bf5e510, EDX=0x42d6dcb0
ESP=0x40fdc150, EBP=0x40fdc168, ESI=0x40fdc200, EDI=0xa19e9640
EIP=0x062c75f5, CR2=0x0000000c, EFLAGS=0x00210206

Top of Stack: (sp=0x40fdc150)
0x40fdc150: 40fdc200 71c70000 0815a748 0815a704
0x40fdc160: a19e9640 40fdc200 40fdc198 062c74cb
0x40fdc170: 40fdc200 a19e9640 0bf5e510 0bf5e510
0x40fdc180: 080ea6f0 40fdc200 00000010 a19e9640
0x40fdc190: ad38a000 40fdc200 40fdc1c8 0629efaa
0x40fdc1a0: 40fdc200 a19e9640 00000100 00000100
0x40fdc1b0: 0815ab00 40fdc200 40fdc2b8 40fdc200
0x40fdc1c0: 080ea5f0 0815a638 40fdc2b8 062c2905

Instructions: (pc=0x062c75f5)
0x062c75e5: 53 83 ec 0c 8b 7d 0c 8b 75 08 8b 47 04 8d 58 08
0x062c75f5: 8b 53 04 89 d1 c1 f9 02 85 d2 7e 6f b8 04 00 00

Stack: [0x40f5c000,0x40fdd000], sp=0x40fdc150, free space=512k
Native frames: (J=compiled Java code, j=interpreted, Vv=VM code, C=native code)
V [libjvm.so+0x2c75f5]
V [libjvm.so+0x2c74cb]
V [libjvm.so+0x29efaa]
V [libjvm.so+0x2c2905]
V [libjvm.so+0x2bb461]
V [libjvm.so+0x2c9ef5]
V [libjvm.so+0x506929]
C [libpthread.so.0+0x52ab]
A: 

There have been several bug fixes related to GC crashes since 6u13 .. here are a few:

  • 6897143 - Stress test crashes during HeapInspection using ParallelGC.
  • 6845368 - large objects cause a crash or unexpected exception
  • 6840775 - Multiple JVM crashes seen with 1.6.0_10 through early access of 1.6.0_14 - possibly related to GC
  • 6806226 - Signed integer overflow in growable array code causes JVM crash

I suggest using at least u18.

unhillbilly