Shared publicly  - 
1
1
Ian Ward's profile photoBart Trojanowski's profile photoMichael Richardson's profile photoPatrick Naubert's profile photo
6 comments
 
I don't think it does the same thing as munin? okay, perhaps I don't use munin for what's it's intended, or I'm confusing it with something else.
 
munin-clients run scripts and records status of a system they are on. Every 5 minutes a munin-server collects the data from each of the clients and munges up some plots.

The last step is time consuming, and if I don't intend to view the plots all that often it's wasted CPU cycles.

collectd on the other hand seems to have several front ends, all/most of which do the plotting at time of viewing. Some of them even do it on the client side.

Switching to collectd would mean my CPU usage on my KVM host wouldn't be a regular ^^^^^^^^^ plot with spikes every 5 minutes.

The other difference between collectd and munin, is that the former is a daemon so it can collect data at a finer granularity. Munin-client on the other hand runs perl scripts from cron.
 
Munin can be set to only create the graphs when you access them. 
 
sounds like less effort, I like it!
 
unless it's really show when it's configured to "draw on demand" and that's why no one deploys munin that way.
Add a comment...