Process Stack Failure

[Raghavendra GL]

Hi,

I have a Small N-class server.After i upgraded the server to newer version i am facing this problem.

The exact error is:

PID 6353 received a SIGSEGV for the stack growth failure
possible cause insufficient memory or swap space
or stack size exceeded maxssiz.
ENTERING MAIN PROXY LOOP :.....
PROXY : WAITING

The machine is HP-UX 11.00.


gcsomc01 #sysdef |grep max
dbc_max_pct 10 - - -
maxdsiz 0 - 0-655360 Pages -
maxfiles 1000 - 30-2048 -
maxfiles_lim 2048 - 30-2048 -
maxssiz 0 - 0-655360 Pages -
maxswapchunks 1200 - 1-16384 -
maxtsiz 0 - 0-655360 Pages -
maxuprc 256 - 3- -
maxvgs 10 - - -
netmemmax 0 - - -

gcsomc01 #grep -i max /stand/system
dbc_max_pct 25
max_thread_proc 200
maxdsiz 0XC800000
maxfiles 1000
maxfiles_lim 2048
maxswapchunks 1200
maxuprc 256
maxusers 64
shmmax 0XA000000
[1] + Done sam &

Both maxssiz and maxssziz_64 bit were 8MB and changed them to 16MB.But still the problem is not solved.Pls let me know what else can be done.

[Eknath]

Hi Raghvendra,

If you have installed a new kernel try to reboot the system.
If you have psuedoswap confirm it is enabled.

Cheers!!!
eknath

[Ermin Borovac]

You'll need to find stack size of a process before it died. Then you'll know what to set maxssiz to.

One way to find out stack size of a process is with kmeminfo program (unsupported utility available on request from HP response centre).

# kmeminfo -pid | grep -ie stack -e type
type space vaddr ref virt phys swap
STACK 0x15b6400.0x00000000680f0000 1 16 8 16

Stack size is shown in 4k memory pages (in this example it's 16 x 4K).

Otherwise, if you have glance you can see stack size from 'memory regions' window in 'process detail' screen. From the main screen press s to select process, type its PID and then press F3.

If process left a core file and is multithreaded you may be able to find stack size before it died with following commands.

# echo '$m' | adb core | grep -i stack
Stack: b = 680F0000 e = 68100000 f = 625FC

Stack size is difference between e and b.

[D Block 2]

11.0 - maybe you need to check your System patch environment.

do you have any Quality Patches installed for 11.0 ?

http://www.informatik.uni-frankfurt.de/RBI/hp-service/patches/s700_800/11.X/PHCO_24958.txt

[vinod_25]

hi raghav

After much troubleshooting the following steps were used to prevent the error
from occuring:

# cd /usr/lib
# /sbin/ln libc.2 libc.2.saved
# /sbin/rm libc.2; /sbin/ln /usr/lib/pa11_32/libc.2 libc.2

NOTE: The above commands can affect a live production system. There is a small
window between when libc.2 is removed and linked that can cause programs
calling libc.2 to fail. The above procedure should only be performed during a
scheduled downtime or maintenence window.

regards

Vinod K