Monitoring using Collectd
written on Sunday, December 19, 2010
So. I've gotten my hands on a nice little Sheevaplug Server running Debian Squeeze, and I handily had a small network monitoring job to do...
First off, I tried out trusty old Munin, but the puny ARM processor couldn't quite keep up with the constant graph-generation. As I had also run into a few other of Munins limitations in past projects, I decided it was time to try something else.
Enter collectd
I had previously discarded Cacti, Nagios and friends for being overly complex for my uses; I just use the collected data to assert what has happended after the fact. (And, hopefully, fix it before it happens again.)
Then various sources suggested to take a look at collectd...
Up front, collectd, has only one minor drawback IMHO; it does not come with a built-in front-end for viewing the collected data. (I guess thats the collect-part of collectd).
Installing & Setting up
Install the packages and get going:
sudo apt-get install collectd
Before you start up collectd, you should fix the configuration file to your liking; namely, it defaults to collect data every tenth second, which you might want to change a bit. Take a look in /etc/collectd/collectd.conf:
# Poll once a minute
Interval 60
Please note that you can't (easily) change this interval after you started the server, as it will be "hard-coded" into the collected RRD-files.
It's also in this file you set up what the daemon should actually collect. As to testing that our network actually is alive, I want to ping the outside world:
# Uncomment the appropriate 'LoadPlugin' lines:
LoadPlugin ping
# Later in the file, there's the individual plugin setups:
<Plugin ping>
Host "sbhr.dk" # My own web server.
Host "192.168.1.1" # The local gateway
# I leave the default as-is
</Plugin>
Not it's time to start the server:
sudo service collectd start
(Also, you'll have something besides blank graphs to see later...)
Viewing the output
After browsing the front-end list, I randomly decided on using collectd-web. I chose to go with the trunk:
mkdir Source cd Source git clone https://github.com/httpdss/collectd-web.git cd collectd-web
Before moving on, a few packages are needed, though (and check it's all there by running the check_deps.sh-script.):
sudo apt-get install librrds-perl libjson-perl liburi-perl libhtml-parser-perl ./check_deps.sh
If the latter returns all-OK, then you can use the runserver.py-script to take it for a spin:
./runserver.py
It will run a small webserver on the machines' port 8888, where you can watch it work:
While I had some ideas about using Nginx and spawn-fcgi to run the above CGI-scripts, I've decided to settle for running the runserver.py in screen:
screen -d -m ./runserver.py
Done!