[icinga-users] Performance tuning/scaling Icinga

Devdas Bhagat devdas.bhagat at booking.com
Wed Feb 15 16:04:17 CET 2012

	We have multiple Nagios instances monitoring over multiple
datacenters (one per datacenter). We are now trying to simplify the
monitoring and alerting setup.

	We are trying to setup an Icinga instance to accept
notifications from our active pollers and simplify alerting and
reporting. At the moment, the required feature is the Icinga web
interface (to allow each group to have it's own custom views of data),
which requires us to use the ido backend.

	Current testing has shown that the ido backend isn't able to
keep up with data volumes. However, the non ido setup can keep up with
current feed volumes (about 70-80 passive check results/second).
We are monitoring about 21000 services over 1700 hosts (and growing).

Does anyone have suggestions on scaling this up/out?

I am looking at using multiple instances of Icinga as collectors (one
feed per datacenter), and one reporting instance, all talking to the
same backend database (if that is possible).

top(1) indicates:
  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND                                                                                         
 5309 icinga    25   0  293m 137m 2692 R 100.0  0.3  61:03.65 icinga  

That's on two feeds worth of input.

This host has spare CPU and RAM, but is relatively limited on disk-io.

Devdas Bhagat

More information about the icinga-users mailing list