Please have you a script to find and kill zombie processes?

[Pitone]

Hi everybody, have you please a script to suggest me how to find and kill zombie processes?

Regards, thanks in advance

 

[jpor2003]

Have you tried a ps -ef |grep defunct
and then killed the zombie process with the kill command?

 

[Mag0007]

ps -ef | grep defunct | awk {'print $2'} | xargs kill -9

 

[hirschaj]

I have never seen anyone be able to kill a zombie process, and I don't think there is any benefit to it. A zombie process just sits around in a wait state doing nothing. The only method I have ever seen to clear zombie processes is to reboot the server, but again, since zombies don't use up any substantial resources I don't really see the point to making it much of a priority. Maybe someone else has experienced something different, but that is what I have always seen.


Jim Hirschauer

http://www.aixexpert.com

 

[Pitone]

Thanks sirs, my question is.. I have such a situation..
when users close their session by clicking the "X" on the terminal, they don't kill all the processes forked by the shell, so I have "zombie" processes using cpu.. I need to recover their PID and kill them.
regards

 

[hirschaj]

Hmmm, a true zombie process should have a "Z" listing in when you run a ps. Maybe what you are seeing are orphaned processes that are still using resources since by definition a zombie uses no resources besides a listing in the process table. Can you provide ps output for the processes you think are the zombies?


Jim Hirschauer

http://www.aixexpert.com

 

[Pitone]

Here is an example, the process listed is not linked to any session anymore

pintonf 442578 1 67 16:19:01 - 0:15 programnames

How can I detect these processes? and kill them?

 

[p5wizard]

Those are just orphaned processes, if you want to get rid of them, just "kill" 'em or if they do not respond: "kill -9" 'em...



HTH,

p5wizard

 

[Mag0007]

ps -ef | grep defunct | awk {' if ($2 != 1) kill -9 $1; else print "everything okay!"'}

Is this what you want?

 

[p5wizard]

You can get rid of real zombies (processes which have died, but their parent won't acknowledge that fact: the SIGCLD signal has not been received by the parent)

On AIX5 aixmibd is prone to leave zombies around.

You can find out the parent of the zombie processed from a

# ps -ef|grep defunct

output. If it is aixmibd, get rid of the zombies by stopping and restarting aixmibd:

# stopsrc -s aixmibd

(then wait while srcmstr kills aixmibd - takes longer because now aixmibd first has to acknowledge all it's zombies before terminating itself)

then restart aixmibd

# startsrc -s aixmibd

If the zombies are caused by a different process, sometimes a reboot is the only solution, because you can't go and indiscriminately kill any old process - well you can, but not without adverse effects...



HTH,

p5wizard

 

[Pitone]

Thanks guys, now I understood

 

[Pitone]

But one thing, If I want to get rid of parentless processes non linked to shell, by script? by crontab?

 

[RodKnowlton]

I've dealt with a similar situation, and found that the orphaned processes could be identified by examining lsof output, because they lacked stdout (file descriptor 0).

If you have lsof installed, you can check a known orphan's files with "lsof -p <pid>". If it has no entry with 0u in the FD column, this could be the way for you to identify them. The -c option was used in our case, because the orphans we were getting were all of one particular program.

I'd post the script, but process reaping is best developed on a case by case basis, with plenty of testing.


Rod Knowlton
IBM Certified Advanced Technical Expert pSeries and AIX 5L
CompTIA Linux+
CompTIA Security+

 

[mjldba]

Pitone:

I have a similar issue with users X'ing out of old Informix 4ge apps, so I have a cron job set-up to run the following afterhours:

kill -9 `ps -ef | grep rprp.4ge | awk '{print $2}'`

Works like a charm but (isn't there always a but .....) doesn't help during normal business hours