zabbix using all cpu

I only ever see a high number of zabbix processes if several agents aren't reachable (e.g. outage of a part of the network or one of our branches) and active checks have to run into timeouts. This of course also leads to higher CPU usage, but not to the extent that it (severely) impacts the system...
 
Don't put the database and zabbix server on the same machine. Split these up. They both have different loads and putting them on separate machines will make it easier to tune that machine for its specific purpose.

Also check your zabbix service health.
 
Alain De Vos said:
My first thoughts now are disabling some snmp monitoring
Click to expand...
Make sure to have 'use bulk requests' enabled for all SNMP hosts (if they support it and don't crash on bulkwalks... :rolleyes: ). Using single requests for a bunch of items drastically increases load on the server and agents! Also some devices (printers...) might be *extremely* slow and even stall for several seconds when replying to large bulk requests. Limit the requested items on those hosts to the ones you actually care about to speed up data gathering by snmp.


For larger environments definitely run zabbix and the DB on different hosts as SirDice recommended. For anything with <50-100 hosts (depending on the number of items/triggers per host) you should be fine with everything on a single machine (depending on its hardware). We're monitoring ~30 hosts and ~15 printers with zabbix and postgres in jails on a single Atom C3558 host just fine, but YMMV.
 
You must log in or register to reply here.
Back
Top