Zabbix Redundancy

[SilenceTi]

Hi. I'm implementing zabbix to a large enterprise. I've a few years of experience of zabbix. but now, i need to have something like HA in Zabbix. So if server or rack is down, I need another zabbix from where i can see the Triggers/alarms , so my monitoring don't stay in the dark. What is the best procedure for this? thanks.

[kloczek]

Zabbix server and proxy are stateless and if one will ie you can spawn another one as long as they will be using the same addresses.
All what needs HA is db backend (active-standby) of the server and proxy and here you can use well known HA techniques on securing HA this type of the components.

[zdenek.spichal]

Ahoj

Some simple example:

- setup at least two servers for DB and setup MariaDB active-active Galera cluster for example
(recom. is at least 3 members, you can spread them over more locations ..)
- setup two servers with Pacemaker and Corosync for app+web server
- setup VIP for this servers as app and web IP
- setup HAproxy for connection to DB servers
- setup backup
- enjoy the HA setup

Best regards

Zdenek

[tjyang]

Hi Zdenek
I assume your example is for HA in LAN environment,right ?
Do you have suggestion of HA over WAN ?
active-active MariaDB A and B sit in two D.C. across US East and central which has ping time around 22ms ?
IP(DNS) fail-over will be using BIG-IP F5.

[kernbug]

Hi Zdenek
Do you have suggestion of HA over WAN ?
active-active MariaDB A and B sit in two D.C. across US East and central which has ping time around 22ms ?
IP(DNS) fail-over will be using BIG-IP F5.

WAN link reliability and latency makes the synchronous replication solution a big challenge. You can test this in test HA setup between two nodes with this command (Linux):

tc qdisc add dev <interface_name> root netem delay 25ms

[zdenek.spichal]

Ahoj

For HA solution accross two DCs you can setup HA proxy to use one DB node as primary and second only when the first is not accesible (no load balancing or just active-pasive DB solution).
Or better you can have on primary location two nodes of MariaDB Galera cluster with HA proxy (with load balancing or primary/secondary setup) and third node just sync to secondary location and set backup there for example.
I think you should have whole HA solution on your primary site (web+app+db) and just extend sync to your secondary site just for disaster recovery.

Best regards

Zdenek

[kloczek]

Ahoj

For HA solution accross two DCs you can setup HA proxy to use one DB node as primary and second only when the first is not accesible (no load balancing or just active-pasive DB solution).
Or better you can have on primary location two nodes of MariaDB Galera cluster with HA proxy (with load balancing or primary/secondary setup) and third node just sync to secondary location and set backup there for example.
I think you should have whole HA solution on your primary site (web+app+db) and just extend sync to your secondary site just for disaster recovery.

All what is needed here is just slave DB in second DC location.