Solaris

Hello All,
We have korn shell script running in production fine up untill now. The script will basically do grep "informatica" to check and see if any processes are running. If running get the process id's and kill them. we are using "kill -9" command to do this, it looks like today it could not able to kill the processes.

would it be possible there could be an issue with OS or shell?? or do i need to make my script wait for few minutes after issuing the "kill -9" command before checking if any processes are still running??

Appreciate your help.

Thank you.

The "kill -9 <PID>" command is the absolute last resort.

Unless the process has no database connections and is not firmly stuck in I/O (which would probably need a reboot anyway) there is no reason to ever issue this command.

There is no way that I would ever issue this command to a client process which is still attached to a database. There absolutely no way that I would ever issue this command to a database engine process unless I was in a disaster situation and could not shut down the system prior to restoring the database from backup.

Please read the documentation for your database engine and consider appropriate ways to stop extraneous or stuck processes.

ImHo. Do not issue "kill -9" to a process connected to a database or to a database engine process.

Hope this is clear.

The processes which my script is killing are no way related to the database, those are application scripts which were running at that time polling for a indicator file and sleeps for a while. Is there any draw back using "kill -9" command?? keeping drawbacks aside why is it not working? do i have to make the script sleep for a while after issuing the "kill -9" to give enough time to make it successfull.

Yes, as already stated, "kill -9" is a last resort command. It shouldn't be used routinely. "kill -9" bypasses all measures that a program and the libraries it uses put in place for the process to die gracefully.
You do not provide clear evidence "kill -9" did not work. How did you conclude that ? Can you post "ps" output ?
Don't use "kill -9" and explain why you are killing these processes in the first place.

Why not just do a

Code:

pkill informatica

Do you have a snippet of your code?

Please post:

Sample "ps -ef" type output showing lines containing the word "informatica" or just the relevant processes if they don't contain the word "informatica" all the time. If you can, please post the process tree containing the bad process and also show how the process was started.

The complete "kill" script.

The script you are trying to kill. Scripts with a sleep loop can be difficult to kill. It is often easier to build a stop flag process into the script. I also use Shell "trap" to cause a script to exit cleanly after "ctrl/C" or "kill -15".

Are you running a unix virus scanner (e.g. Sophos)?

If you are running an Informatica database process can you be sure that the word "informatica" will never appear in a "ps" display from a real Informatica process?

Just out of interest I have seen processes take hours to disappear after "kill -9" or not disappear at all. Beware that killing part of a process tree can leave orphan processes owned by init (pid 1) which are hard to distinguish from genuine background processes.
Also I have seen "ps" miss processes on numerous occasions. If the kernel is busy it may not reply.