Just thought I’d share a little python script I wrote the other day and have found helpful on several occasions already. It’s basically the same as `tail -f file | grep pattern`, but, if it finds a match, it allows you to execute an arbitrary command.
For instance, the yesterday a customer had problems with Apache infrequently segfaulting. The problem was temporarily fixed by simply restarting Apache, so I used this script to automatically restart apache each time children started segfaulting while we looked for a more long term solution.
Here is the script:
$ cat pytail.py
#!/usr/bin/python
# This script mimics the behaviour of tail -f , but has the capablity to execute
# a command whenver a line with a matching pattern in it is found.
# As you probably see, there's no support for regex or anything of that
# flashy nature.
#
# Erik Ljungstrom 15/10/07
import time
import sys
import os
import socket
file = open(sys.argv[2],"r");
node=socket.gethostbyaddr(socket.gethostname())[0]
setdef = os.path.getsize(sys.argv[2])
file.seek(setdef-1)
while 1:
where = file.tell()
line = file.readline()
if not line:
time.sleep(1)
file.seek(where)
else:
pos = line.find(sys.argv[1])
if pos >= 0:
os.system("logger -p local6.crit -t postfix amavisd has bailed out on %s!" % (node))
sys.exit(1)
(As always with python code, make sure the intendation is not mangled by the copy/paste). If it does, you can download the script here.
You start it such as ./pytail.py pattern file. As my comment says, it’s very basic, but useful! Of course it would be nice if it forked into the background and could do regexp matching etc. But I simply haven’t had use for that yet, and `screen` is a good enoguh tool. The above ‘version’ was used under pretty similar circumstances. A customer’s amavisd kept bailing out due to BDB 3 limtiations (running out of locker entries). So I ran that script while we upgraded it on all their mail servers. So I ran ./pytail.py child_init_hook /var/log/messages The command being executed is the bit between the quotation marks in os.system(). In that case, I had previously set up syslog to log to a MySQL instance on one of my own servers which is then polled from our monitoring panel (2 x 50″ LCD screens in my corner of the office). So whenever this happened, we knew which server had the issue and could mitigate the problem.
