[mohitanchlia]

I see a process running from 2 days. When I try to kill -9 it returns with no error but that process still keeps showing in top command.

When I do lsof I see somehting like:

afaconv.l 5813 root txt REG 104,2 11396935 20021557 /home/afa_build/work/AFA/main/dist/deploy/jboss/server/afa/deploy/afa/afa-vert.rar/bin/afaconv.linux32 (deleted)
afaconv.l 5813 root 0r FIFO 0,7 2523565 pipe
afaconv.l 5813 root 1w FIFO 0,7 2523566 pipe
afaconv.l 5813 root 2w FIFO 0,7 2523567 pipe
afaconv.l 5813 root 3r REG 104,2 151307 21151952 /home/afa_build/work/AFA/main/dist/deploy/jboss/server/afa/deploy/afa/afa-vert.rar/form/wio03.1pe (deleted)

I think it's in uninterruptible state but sounds like a very bad thing if it happens in our production servers. Any suggestions of what I should do?

[khafa]

what is the process you want to kill and how does it show in top?
maybe its a zombie

[kurianthayil]

Hi,

Who is its parent process?

--Kurian.

[Thrillhouse]

Originally Posted by kurianthayil

Who is its parent process?

I think Kurian is on the right track here. It's entirely possible that the parent process re-spawns the one you're trying to kill if it sees it go down. If that's the case, you'll have to kill the parent process.

[mohitanchlia]

As I mentioned that process has been running from 2 days. top output shows that. But, when I do "ps" it doesn't come up in the list.

[Thrillhouse]

A standard ps (with no parameters) will only show processes started or running for your currently logged in user. To show all processes (including kernel routines) use ps -Af. To show most system processes (not including kernel processes) use ps -ef.

[kurianthayil]

Hi,

To show most system processes (not including kernel processes) use ps -ef.

As Thrillhouse has said use ps with -ef option. PID could be obtained from the top command and then use ps -ef to find out its details.

#ps -ef |grep <processid from top>

--Kurian.

[mohitanchlia]

I have already tried the following:

top and then kill -9 <pid> # from top
ps -ef|grep "<pid>"
ps -ef|grep "<process name"

Since this program is not Kernel routine it will not appear in ps -Af.

So going back to my real question. Why is kill -9 not able to kill the process that appears in top. The same process doesn't appear in ps -ef or ps -eaf or ps -Af

[kurianthayil]

Hi,

Were you able to trace its parent process??? The child process would be re-spawned by the parent so if the parent is not killed it will still spawn child even if you kill the child. Suppose if 6750 is the PID that's making problem (Child) and If you need to find its parent, do the command.

#ps -ef |grep 6750
kurianmt  6750  6749  0 06:34 ?        00:00:00 /bin/sh /opt/glib2.16/firefox3/firefox
kurianmt  6857  6790  0 06:36 pts/0    00:00:00 grep 6750

The parent of the child process 6750, is 6749. So killing the PID 6749 will kill its child too. You will have to trace like that. Make sure you won't kill PID 1. Hope you are aware of that.

--Kurian.

[mohitanchlia]

That's not the issue. There is no parent process trying to spawn child process. As I mentioned earlier, this process has entered in a state that doesn't appear in ps but is seen in top. Which means that it's not getting respawned because the PID in top output is same. Also I already mentioned that this process is running over 2 days, as reported by top. So it can't be a parent re-spawning child. Does somebody else has any more information about why process would enter such a state?