Zabbix Failover suggestions

[registration_is_lame]

How can I setup failover for zabbix polling service without virtual IP.

So I have 2 zabbix servers (along with frontend) connected to the same database.
Zabbix 1: 10.10.10.10 -> Connect to Database 1. Current polling.
Zabbix 2: 20.20.20.20 -> Connect to Database 1. (Zabbix service is stopped to prevent polling and duplicate data in DB - It can't be idle right ??)
Database 1: 30.30.30.30

Agents have been configured to listen from Zabbix 1 and Zabbix 2.

I want setup failover in such a way that if zabbix 1 is down, zabbix 2 should start polling (temporarily).
If zabbix 1 is back online, zabbix 2 should stop and zabbix 1 should take over and start polling.

Any suggestions?

If Zabbix 2 could be idle then instant polling (or send signal to start polling after few minutes) could be possible I guess? if Not, a tool is needed for service to be started if Zabbix 1 is down for 5 minutes.

similar question but different: https://www.zabbix.com/forum/showthread.php?t=16744

Frontend failover isn't necessary because both zabbix frontend has been configured to use the same Database to retrieve data.

[hpeti2]

How can I setup failover for zabbix polling service without virtual IP.

So I have 2 zabbix servers (along with frontend) connected to the same database.
Zabbix 1: 10.10.10.10 -> Connect to Database 1. Current polling.
Zabbix 2: 20.20.20.20 -> Connect to Database 1. (Zabbix service is stopped to prevent polling and duplicate data in DB - It can't be idle right ??)
Database 1: 30.30.30.30

Agents have been configured to listen from Zabbix 1 and Zabbix 2.

I want setup failover in such a way that if zabbix 1 is down, zabbix 2 should start polling (temporarily).
If zabbix 1 is back online, zabbix 2 should stop and zabbix 1 should take over and start polling.

Any suggestions?

If Zabbix 2 could be idle then instant polling (or send signal to start polling after few minutes) could be possible I guess? if Not, a tool is needed for service to be started if Zabbix 1 is down for 5 minutes.

similar question but different: https://www.zabbix.com/forum/showthread.php?t=16744

Frontend failover isn't necessary because both zabbix frontend has been configured to use the same Database to retrieve data.

We use Corosync+Pacemaker to make a similar funcionality. 1 virtual IP for 2 server. HAProxy with 1 virtual IP for Percona XtraDB Cluster (mysql database).

More information see this:

https://www.zabbix.org/wiki/Docs/howto/high_availability

[registration_is_lame]

We use Corosync+Pacemaker to make a similar funcionality. 1 virtual IP for 2 server. HAProxy with 1 virtual IP for Percona XtraDB Cluster (mysql database).

More information see this:
https://www.zabbix.org/wiki/Docs/how...h_availability

I've been looking into corosync and pacemaker. But i'm unsure how to do it without virtual IP. Is virtual IP really necessary? I'm gonna do a trail run for some other service.

Can you post your config if you can? I'd love to know how the zabbix 2 service is set to start in when the main server is down.

[hpeti2]

I've been looking into corosync and pacemaker. But i'm unsure how to do it without virtual IP. Is virtual IP really necessary? I'm gonna do a trail run for some other service.

Can you post your config if you can? I'd love to know how the zabbix 2 service is set to start in when the main server is down.

There are nothing special in zabbix_server.conf just set SourceIP=VIP and corosync config is same as describes in above link.

I think the VIP is really useful if you set this to server's address in zabbix_agent.conf. Only the "active" zabbiy server recive data and avoid database collision.

[registration_is_lame]

There are nothing special in zabbix_server.conf just set SourceIP=VIP and corosync config is same as describes in above link.

I think the VIP is really useful if you set this to server's address in zabbix_agent.conf. Only the "active" zabbiy server recive data and avoid database collision.

I meant the corosync/pacemaker configs, anyway I've seen the documentation.
But i'm not sure how I should configure corosync/pacemaker without virtual IP address. Couldn't find any articles for without virtual IP setup

[peatb]

A virtual IP is just another resource in corosync/pacemaker. So making a cluster without it is the same as with it, just leave the IP out of the configuration (or you could use any address, like 192.168.1.1, and it will just not do anything)

However, why don't you want to use a virtual IP (doesn't your network topology allow it)? It's a really convenient way to setup high availability. It also makes your zabbix agent configs easier, you only have to specify the VIP for server and serveractive. Also you can just ssh to the VIP (and the webinterface is on the VIP as well).

final note: you can't have zabbix running in 'idle' mode and you certainly don't want them active at the same time.

[registration_is_lame]

A virtual IP is just another resource in corosync/pacemaker. So making a cluster without it is the same as with it, just leave the IP out of the configuration (or you could use any address, like 192.168.1.1, and it will just not do anything)

I did not know that. Will try this week.

Thanks for clarifying. Virtual IP isn't gonna happen unless I create another server in the same DC which I might in future.