HOWTO monitor your servers via Twitter

Alert: your server has failed!

The other day I got inspired to write a script that would allow me to monitor my servers via Twitter. The idea of having a column in Tweetdeck set aside to inform me of my servers’ statuses’ sounded cool, plus, it’s quicker than checking email.  I know sending tweets from the command-line had been done before, but after seeing briealeida’s post titled Tweeting Cron Jobs I really got inspired. While hers was written in Perl, I didn’t want to go that route since I had a few, self imposed, restrictions I wanted to stick to. One, I only wanted to use standard shell commands, the ones you get by default in Linux, so you would have absolutely no dependencies to install for this to work. Two, I wanted to see how much info I could stuff into a 140 character tweet, and still have it make sense. While I’m still working on adding more info, the current state of the script gives me a quick snapshot of seven specifics metrics on a selected server, which I’m quite happy with. To try it yourself only takes a few minutes.First you’ll need a Twitter account, either the one you already use, or a dedicated one that only your servers post to (this is what I’ve done). For a tad bit more security I protected my tweets and made the account private – I don’t think I’ll get hacked by anyone that sees my system load is too high, but I’d rather not hear about it from security experts that I’m leaving myself open, so I’ll continue to keep things blocked by default. Also, I’ve setup the update method to use SSL to login and transmit, so you can rest a little easier knowing that things are staying encrypted while they move over the wire.  The only option I’ve coded in is that you can use wget (by default) or curl (not installed by default in any Linux distro I’m running, but is on OS X if someone wants it for that), so you have a little bit of flexibility if you need it.  Otherwise it’s pretty much ready to go out of the box, define your username and password at the top of the script, chmod it, and away we go.  Slap it in your crontab  for daily/hourly updates, and the rest, as they say, will be handled by simple, beautiful, BASH.

[codesyntax lang="bash" lines="no" container="pre_valid" capitalize="no"]

#!/bin/bash
#
###############################
# twitter username/password #
###############################
user="1user"
pass="sekrit"
#
###############################
# run tasks for the report #
###############################
HOST=`hostname -s`
UP=`uptime | cut -d" " -f4,5 | cut -d"," -f1`
LOAD=`uptime | cut -d":" -f5,6`
PING=`ping -q -c 3 google.com | tail -n1 | cut -d"/" -f5 | cut -d"." -f1`
MEM=`ps aux | awk '{ sum += $4 }; END { print sum }'`
CPU=`ps aux | awk '{ sum += $3 }; END { print sum }'`
if [ -x "/usr/bin/lsb_release" ]; then
DIST="`lsb_release -s -i`/`lsb_release -s -c` on `uname -m`"
else
DIST="`uname -o` on `uname -m`"
fi
#
###############################
# build the report for post #
###############################
tweet="(HOST) ${HOST} (UP) ${UP} (CPU) ${CPU}% (MEM) ${MEM}% (LOAD) ${LOAD} (PING) ${PING}ms (DIST) ${DIST}"
#
################################
# check that post is <140 char #
################################
if [ $(echo "${tweet}" | wc -c) -gt 140 ]; then
echo "FATAL: The tweet is longer than 140 characters!"
exit 1
fi
#
################################
# post the report to twitter #
################################
### via wget (default)
wget -q --user="${user}" --password="${pass}" --post-data=status="${tweet}" https://twitter.com/statuses/update.xml
### via curl
#curl -k -u ${user}:${pass} -d status="${tweet}" https://twitter.com/statuses/update.xml >/dev/null 2>&1
#
################################
# if no errors, successful #
################################
if [ $? -eq '0' ]; then
echo "Tweet sent using `echo ${tweet} | wc -c`/140 characters."
fi
rm update.xml
#
################################
exit 0

[/codesyntax]

Told you it was pretty simple, now I get updates to my ‘secret’ server account on Twitter that look like this:

(HOST) mookie (UP) 8 days (CPU) 6% (MEM) 44.1% (LOAD) 0.07, 0.06, 0.07 (PING) 29ms (DIST) Debian/squeeze on i686

So what other statistics would be helpful to gather that we can tack on to what we already have? Heck, we could even drop the (DIST) section, that just gives me a reminder if what distro each box is running; it could be shortened to just say Debian or Slackware. One thing I’m working on, I would like a monitor to report when any partitions get over 90% full, I haven’t been able to figure that one out yet without using an external app not installed by default, like monit.  So, is this helpful? Can you think of better way to do any/all of the above, while fulfilling my basic requirements? Does it blend?