Warning! Bias rant follows.

I’ve been struggling to find a perfect solution for server monitoring for a while now. At the moment it’s a combination of nagios and cacti plus ipmitool. And recently I have also looked at / tried opennms, zenoss, ganglia, argus and munin.

Nagios + cacti has been serving me well, but I would be happier to only have to maintain one system and to have some auto discover of both systems and their services would make life much easier. Most of the apps I looked though have failed my expectations by either being hogs (some even java based), complicated mashups, or too simple for my needs (not able to replace both nagios and cacti). And that left me with zabbix. But even zabbix has left me with a bad taste by requiring the zabbix-agent be installed on client machines. Are my requirements really to much to ask? Surely there is enough to be gleamed from snmp, ipmi and nmap to monitor both server and desktops.

So. How does you see the required agent? A negative? I can easily install it on all my linux server with puppet, but what about windows, the desktop machines and routers etc? And I have yet to see anything come close to the asset management I’d like either.

But on the plus side for zabbix I have found it fairly well thought out, albeit a little confusing at first.

I cant remember the name of that windows thing I tried a little while ago that sent me on this mission to find a decent linux all-in-one system manager, but it was reasonably forgettable because of the auto-discovery problems it kept presenting with the doubling up of systems on each discovery scan, and it’s inability to be uninstalled.

Anyway I’ll be off for a week to shift house but I’ll give zabbix a good go and report back.