Zabbix sever muliti-homing redunancy for hosts reporting in

[greavette]

Hello,

At work we use MS System Center Operations Manager but I'd like to teach myself Zabbix for home use. SCOM has the ability to use multiple management servers whereby agents can report into either management server for redundancy. If one is down the other management server takes over as primary.

Does Zabbix have this ability like I explained above?

I have a small cluster of virtual host servers I've built with a common backend storage also in a cluster. I don't have HA enabled on the virtural machines yet but I would like to run multiple Zabbix Servers on each node. I would like to have the ability for my Zabbix to be redundant in that all hosts report into one Zabbix server and if that one is down they will switch to report into another Zabbix server.

What do I have to do in my setup to achieve this? Do I need to have my database in a cluster behind multiple Zabbix servers? Is there a name in Zabbix for this type of setup like mulit homing?

Any advise or existing documentation you can point me to would be greatly appreciated.

[greavette]

Follow-up to my post...

I have seen this wiki page here:

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

But this line here says that I need to have load balancing setup for both Zabbix Servers (unless I'm not reading this correctly):

share a virtual IP which is grouped together with Zabbix server (that is, Zabbix server follows the IP).

I guess what I was looking for was if the Zabbix Config file on each Host can point to mulitiple Zabbix Servers and decide which is the primary to talk too. Or do I need to setup a virtual IP that all hosts connect too for reporting in.

Thank you.

[kloczek]

Hello,

At work we use MS System Center Operations Manager but I'd like to teach myself Zabbix for home use. SCOM has the ability to use multiple management servers whereby agents can report into either management server for redundancy. If one is down the other management server takes over as primary.

Does Zabbix have this ability like I explained above?

I have a small cluster of virtual host servers I've built with a common backend storage also in a cluster.

Central part of the zabbix software stack in reality is not what it may you think.
It is not zabbix server process per se but it is zabbix server bakend database. System and database administrators engineered in past decades many different techniques securing such resource like database. Most basic technique is using master and slave DB.
In other words thinking about multiple layers applications like zabbix as king of monolith may work only in quite simple scenarios and/or it would be very hard to manage and develop such stack.

Try to think that in some extreme cases zabbix servers are writing every minute few GB of new monitoring data when in most typical cases even quite big monitored envs such batches of data are stored every few hours or only within whole day.
Doubling every part of the whole stack to have this kind of magical effect in real case scenarios is not practical when on top of this needs to be added scalability.
From this point of view architecting something from ground to have such ability which you thing that it would be good in reality would be very complicated and additionally would be way more complicated to maintain.

Sometimes on engineering things you must balance between what you like and what you can do or what is possible to do. And of course most important here is KISS principle
From this point of view and few others as well zabbix architecture is IMO very close to kind of SweetSpot(tm)
Instead trying to use zabbix like tools which you already know try to understand why zabbix is exactly like it is now