Makes it easy to monitor and report just about anything to Graphite from the commandline.
Created initially to report the useful numbers from
service memcached status
to Graphite. More examples are added as needed.
Write a script that outputs some
name value pairs tha your care about.
pipe-to-graphite will take care of formatting them and sending them
to graphite at regular intervals. The expected output is something like this:
some.thing.count 53453 SomeOtherThing 3522
NOTE: Your script must finish with an exit code of 0 or the output won't be sent to graphite (though it will be logged along with a failure message).
Lets say you want to monitor the
glorkd program. Let's say glorkd comes with
/usr/glork/status that spits out something like this:
Glorkd - v2.1.352 Running: pid 3242 Blargs: 1230 Blorks: 452
You only care about the Blargs and Blorks (those are counts since program start). All you have to do is write a script that does some grepping and awking to get that into this:
Blargs 1230 Blorks 452
Your script (lets say
~/glork-stats.sh) would look something like this:
/usr/glork/status | tail -n +3 | sed "s/: / /"
To get that regularly reported to Graphite, run this command
./pipe-to-graphite.sh ~/glork-stats.sh >> /var/log/glork-stats.log
Or to report it to graphite at any time, if you want to use cron or something. The '-' argument indicates input should be read form stdin
~/glork-stats.sh | ./pipe-to-graphite.sh -
Output from each run is prepended with a timestamp and echoed to stdout Redirect that wherever you like.