Kill -15 issued on the database process

[niranjan_9]

Hi,
We are running progress-database on HP-Unix-11. Last week end, the database was shut down b-coz of a kill -15 issued on the db- process.

Is it possible to find out who (or which script) issued the kill -15 command ??

The `last` command is showing the users (logged in at that time) who do not have root access.


Any insight will be very helpful and appreciated.

Thanks & Regards
niranjan

[Nicolas Dumeige]

Hello,

You could take a look at the .sh_history in all users home directory - given taht they use sh or ksh - or .bask_history ... but for that you'll need root privilege as all files are not readable for others.

Are you sure that the kill was issue by a user ? Do you have a core dump ? Is there anything anormal in the log of the database ?

Cheers

Nicolas

[KapilRaj]

The oracle owner id can also do a kill -15 hence check the .sh_history for oracle owner and root.

Also if somebody has done a su to root with his personal id , it will be logged in his history file!

Regds,

Kaps

[Shaikh Imran]

Hi,
How do you know that kill -15 command was issued to the database and b'coz of which it shutdown ?
what i feel is if a process uses a shared memory or semaphores and that process is terminated with errors then a kill -15 command is issued by the system itself which
in turn terminates the process.
I advise you to check the error logs of the database for more details.
Regards

[niranjan_9]

The below database log says that kill -15 is issued at 17:09 hrs. We are using PROGRESS-database.
Does unix issue kill signal on DB process ???

Pl. refer "BROKER 0: SIGTERM" in this log:

16:28:31 Usr 9: Login by athk on /dev/pts/ta. (452)
16:33:02 Usr 9: Logout by athk on /dev/pts/ta. (453)
17:00:00 Usr 9: Login by root on batch. (452)
17:00:21 Usr 9: Logout by root on batch. (453)
17:09:00 BROKER 0: SIGTERM received. (3694)
17:09:01 BROKER 0: SIGTERM received. (3694)
17:09:01 Usr 16: HANGUP signal received. (562)
17:09:01 Usr 16: Logout by aktan on /dev/pts/tlb. (453)
17:09:01 BROKER 0: Begin normal shutdown (2248)
17:09:01 BIW 7: Stopped. (2520)
17:09:01 APW 5: Stopped. (2520)
17:09:01 APW 6: Stopped. (2520)
17:09:07 WDOG 8: Stopped. (2520)
17:09:08 BROKER : Removed shared memory with segment_id: 57351
17:09:08 BROKER : Multi-user session end. (334)

Regards,
niranjan

[KapilRaj]

Do you have memory problems ?. I think unix kills some processes if your memory and swap in insufficient !

Kaps

[niranjan_9]

Hi Kaps,
I have script which kills all the users at 17:09 hrs. That means there were no users at that time. So memory should not be a problem . right ??

(is there any command to check this ?)

There is a batch-run script which uses this command to kill users @ 17:09hrs:
processid=`who -u | grep "pts" | awk '{print $7}'`
kill -15 $processid

(This script will never pick the root process.)

regards,
niranjan

[Nicolas Dumeige · Accepted solution]

When a process try to access a forbiden page the Unix kernel will take necessary steps to stop that process.

That's why I asked if you have seen a core dump. If the database process was killed by the system, the core dump will contain "segmentation fault", "segmentation violation" or "bus error" message.

Cheers

Nicolas

[niranjan_9]

Hi,
I checked for "segmentation fault", "segmentation violation" or "bus error" messages in these files:

But I could not see any such messages,

/usr/adm/core
/var/opt/cifsclient/core
/var/adm/core
/u02/disk4/MFG-REPORTS/mrp/daily/core
/core
/reports/mrp/daily/core

Pl. note this happened on 04-Apr-04, I checked these files on 14-Apr-04.

I am sending these files to HP-helpdesk for further analysis.

Thanks
Niranjan