Monitor your server with Zabbix (pretty graphs included)

Saturday 01/17/2009  –  Category: Uncategorized

(...or, how MRTG is too much trouble to setup)

My server was having spiking load averages this past week and I had no idea was going on; my top output didn't have any processes that stood out and I didn't have any cron jobs running either.  A friend suggested I use a monitoring tool like MRTG to keep track and graph my performance.  After downloading and compiling I couldn't for the life of me figure out how to get MRTG fully working (following the guide here).  I was able to get the free memory and TCP connection graphs working but had no luck with the CPU usage graph--the one I really wanted to see.

After a few hours I finally gave up on MRTG and started looking for other solutions: Cacti, Zenoss, Nagios...I eventually ended up choosing Zabbix because it looked more lightweight and less intimidating to setup...and most importantly it generated graphs of relevant data.

Some gotchas during the setup process:

• The install script installs the PHP admin interface into /etc/local/zabbix (or something like that), creates an apache conf file that aliases whatever you set your web dir to be, and tries to restart Apache. Since I'm using nginx this obviously didn't work so I ended up copying /etc/local/zabbix to a public web directory.
• It isn't completely apparent that you have to go to Administration > Installation to setup Zabbix for the first time
• Setup items to monitor under Configuration > Items
• Setup graphs under Configuration > Graphs