Monitoring Zabbix

[mpureka]

A concern for us in our Zabbix implementation is - what happens if our Zabbix server experiences hardware failure or something similar. In this sort of situation, we essentially become blind to other problems in our infrastructure. Our previous monitoring solution was agentless, so it was not a problem to have a secondary install of the product keeping tabs on the "live" install, but since Zabbix self-monitors, and therefore the Zabbix server already has a Zabbix agent installed, reporting to the Zabbix server, I am uncertain of how best to implement this.

What have other people done to address this sort of concern?

[Rudd]

Quis custodiet ipsos custodes?

You can configure the zabbix agent to communicate with multiple zabbix servers by adding both addresses in the Server directive:

Server=192.168.10.10,192.168.20.10

So multiple zabbix servers can monitor each other, like your previous agentless solution.

[mpureka]

Excellent; I didn't realize that the agent would report to ALL servers in the list. I thought the multiple servers in the list option was there for redundancy/failover.

[ghoz]

Be carefull, according to the documentation, the agent will get its checks and send active items only to the first server...

[mpureka]

Oh. That's what I initially thought. That's not so useful then.

[Rudd]

It will allow your secondary zabbix server to monitor, and alert on, all agent items providing a completely functional health check of the "live" server. Was that not your primary objective?

[ghoz]

Oh. That's what I initially thought. That's not so useful then.

no, it just means you have to setup your cross server checks using passive checks...

other options :
* setup 2 agents on 2 ports
* use scripts and 2 zabbix-sender per item ...