I have a third party app that is core dumping. Some days I get 8 to 10 dumps.
No one is complaining (except me) about any problems using the application.
This has been going on for several months (so I can forget trying to identify
any application changes that might point to the source).
I have a crontab set up to check /var/core once an hour and run the following
mdb commands on any new dump files (the files are moved to a directory on an
NFS mount to keep /var from filling):
mdb $bin $i <<EOA
=nn"** Core file status **"
="------------------------"
::status
=nn"** Thread stack **"
="----------------------"
::stack
=nn"** regs **"
="----------------------"
::regs
=nn"** Shared objects **"
="----------------------"
::objects
EOA
::status all most always shows (sometimes the signal is SIGBUS, sometimes
SIGSEGV)
** Core file status **
------------------------
debugging core file of f90webm (32-bit) from kermit
executable file: /u01/app/oracle/product/10gASforms/9.0.4/bin/f90webm
initial argv:
/u01/app/oracle/product/10gASforms/9.0.4/bin/f90webm server webfile=HTTP-0,0,0,
threading model: multi-threaded
status: process terminated by SIGILL (Illegal Instruction)
::stack almost always starts (ends?) with:
** Thread stack **
----------------------
0x1638adc(3, 1814490, ffbfa32c, 0, 0, 0)
ixwsru+0x114(12703e8, ffbfa394, 0, 0, 0, 0)
icgcud+0xe8(12703e8, ffbfa3f8, 100, 0, ffbfa464, 0)
ifzterm+0x1dc(12703e8, 1, 1, e, 81010100, ff0000)
siehjmpterm+0x2cc(e, 0, ffbfa698, 0, 0, 0)
libthread.so.1`__sighndlr+0xc(e, 0, ffbfa698, 1fab48, 0, 0)
libthread.so.1`call_user_handler+0x234(e, 0, ffbfa698, 0, 0, 0)
libthread.so.1`sigacthandler+0x64(e, 0, ffbfa698, 0, 0, 0)
libc.so.1`_read+0xc(6, 1290388, 400, 400, 7ffffff2, fece625c)
ixhgr_GenRead+0x6c(12703e8, 127dd78, 1290388, 400, ffbfab84, 0)
ixhrdh_ReadHeader+0x38(12703e8, 127dd78, 1290368, 10e3f54, 0, 0)
ixhhsm_HTTPStateMachine+0x250(12703e8, 127dd78, 1, fffffffe, 4, 0)
Have talked to the third party support people and they ... (well, don't have
anything usefull to add).
Without any other information, is there something I could get from the dump
that would help with making the error more "observable"? (Well, at least to
the casual observer [me])
The server is a V490 running Solaris 9 4/04. The application is an Oracle AS
based ERP.
Thanks for any ideas/insight/wisdom,
Glen Gunselman
This message posted from opensolaris.org
