Monitor log file

Shell Programming and Scripting
1

Hi,

I need to amend an existing ksh script so that it runs a process (stop weblogic) and in parallel needs to monitor a log file (startup.log) in the background for a certain string (e.g. unable to stop weblogic). If the string appears in the log i need to kill the stop weblogic process.

Currently the script just runs the stop weblogic process and hangs if a certain string appears in the log file.

2

How about running stop in background and monitoring the log+child-process in the foreground like this:
(assumption is you have /proc filesystem, if not might need to call ps instead of checking the /proc/$CHILD directory).

stop weblogin &
CHILD=$!
while ! grep -q "unable to stop weblogin" startup.log && [ -d /proc/$CHILD ]
do
   sleep 5
done
grep -q "unable to stop weblogin" startup.log && [ -d /proc/$CHILD ] && kill $CHILD

You could also add a saftey check, eg if after X loops it kills the child anyway (for taking too long).

3

I will give this a try. Many thanks Chubler_XL!

---------- Post updated at 11:07 AM ---------- Previous update was at 08:27 AM ----------

Ok, this is a little more complicated then i thought.

Unfortunately i can't just grep the startup.log for the string "unable to stop weblogin", i need to tail the startup.log everytime my script is run as i need to ignore any previous instances of this string in the log.

Therefore i need to able to run 2 processes simultaneously

Process 1) Tail the startup.log for "unable to stop weblogin"
Process 2) Stop weblogic

and If process 1 finds the string before process 2 has completed I need it to kill process 2

...and if process 2 completes with process having found the string i need my script to kill process 1.

Any help would be much appreciated.

4

Could you age the log such that the current log only applies to the current instance. This would be normal practice.

5

That would definately make sense but unfortunately i am unable to change that process. The startup.log is archived per start/stop and therefore i have no choice but to 'tail -f' it.

6

You could use awk to only look for new occurances of string within log:

(first line creates logfile if none exists - this simplifies rest of code as we can ignore file not existing case).

[ -f logfile ] || > logfile
LN=$(wc -l < logfile)
stop weblogin &
CHILD=$!
 
while [ -d /proc/$CHILD ] && awk "NR>$LN && /unable to stop weblogin/{exit 1}" logfile
do
   sleep 1
done
[ -d /proc/$CHILD ] && ! awk "NR>$LN && /unable to stop weblogin/{exit 1}" logfile && kill $CHILD
1 Like