Zabbix Active/Active Server Cluster

[tmroberts]

This has been posted on here a few times but I've not see much in the way of concrete yes/no answers or how-to's....

We are deploying a new Zabbix instance to AWS and we are trying to figure out if its possible, and if so how, to create an H/A and load balanced Zabbix server. I know its possible and even fairly easy to create a load balanced and scalable web front end and the database backend, but there seem to be NO answers for the server component.

So... yes or no, is it possible to create a zabbix server that is h/a AND load balanced with active/active nodes

If yes, how?

[nelsonab]

At present it is not possible to Active/Active the Zabbix server. This is because the Zabbix server maintains a cached state of the data in the DB. You can however have multiple web frontends. If you want to HA the Zabbix server, you can use a VIP and fail that over between an active and passive node. You'll need to leverage the sourceip configuration options in the server conf file.

[GPegel]

I totally agree with Nelsonab. In my case I have a created a master/slave setup for the Zabbix database. And I have setup 2 Zabbix web-frontends behind a load balancer. It doesn't get more HA than this I guess.

[kloczek]

Active/active DB backend setup is possible with sitting LB between zabbix srv and DB backends.
Nevertheless there is no something for nothing and active/active setup is possible paying maximum possible performance. In other words this will be working up-to-some-performance beyond which will be possible to use only master/slave DB backend.

Conclusion: if someone is planing to reach ~+1K NVPS scale starting from active/active DB backent is pure waste of time.

[GPegel]

@kloczek, true but sometimes companies want to have a 'disaster and recovery' setup. So that's why I had to create a master/slave setup for the DB backend. And no I'm not doing 1K of NVPS. I just hit the 200

[kloczek]

+Two databases in master/slave setup + console as orchestration automatically promoting slave as new master can do this.
Using console it is possible to orchestrate rebuilding new slave when previously unavailable master needs to be rebuild to provide functions of new slave.
Instead console can be used any other multi node orchestration infrastructure.

[GPegel]

I totally agree about using console. I'm trying out the posibility to use VAMP in a test environment. https://vamp.io/

[sauron-le-noir]

VMWARE can do what you want

Vmware offer the full tolerance.

It keep in sync 2 VM when one fail the other take over without any interruption
Be carrefull this opération take a lot of bandwith imagine every bytes changed
in one machine is transmited in the seconde one but you have full tolerance
for all product and that include zabix if the PGSQL ou Mysql is on the same VM

We use this to sync our jboss7 server and our infra server dns,dhcp webserver