Script for monitoring hang process

[bong_3]

Hi all,

I just started to read/learn scripting for linux and find it very interesting. Now i'm trying to monitor a process that always hangup on my linux and perhaps kill it by running a script under cron.

see my attachment...

from there, when i run a top command it shows me that bash process under user1 takes up a lot of process which causes my linux hang.

I wonder whether anyone can suggest how to write a script for this or perhaps share your scripts if happens that you already have one.

Appreciate anyones help.

Bong

[Derek Whigham_2]

ps -efd|grep |awk '{print $2}'|xargs -x kill -9

This should kill the first instance of any entered

Derek

[Mark Grant]

You have two problems here. The first is that it is extremely unusual for "bash" to be taking up all your CPU unless you are running some script stuck in a loop that doesn sleep or wait or something. Secondly, how are you going to determin that a process has hung. My personal favourite for determining a process that definately isn't very well is to check the "C" field as determined by "ps" and look for 255 as a consistant value.

Rather than kill off hung processes, I would suggest it is more important to determine why the thing is hanging or hogging CPU in the first place. I would be interested in your example to know exactly what "user1" is doing.

[Stuart Browne]

I'm with Mark's school of thinking here.

Don't write stop-gag's to get *around* issues.. Fix the issue.

[bong_3]

Hi,

Agree with both Mark & Stuart, of course I need to find out what causing this bash hang. But as of the moment, at least make a temporary solution for this.

Bong

[Stuart Browne · Accepted solution]

As Mark states, you need to watch the CPU usage.

Using the 6th column (C) in the output of 'ps -elf' is by var the easiest way to go.

When the system is in it's bad state, check tos ee what the 'C' column is doing in a ps output.

To continually monitor it, a simple routine similar to this should suffice:

while sleep 5
do
ps -elf | awk '{
if ($6 > 250) {
system("kill -HUP " $4);
}
}'
done

The two arbitary numbers are the 5 in the 'sleep' (how long between checks), and the 250 in in the awk's if statement (whatever is a bad value from your ps checks earlier).

.. just an idea atleast ..

[Ian Kidd_1]

Like the majority, I'm in favor of finding out what's causing bash to run at such a high CPU%.

The next time you see the problem, try using `ps -ef` and follow the parent/child process numbers (or see if `pstree` can give you enough info) to see what's causing this problem.

You could write a script to kill a bash shell if it reaches a certain percentage value, but you run the danger of killing something you might not have wanted to be killed.