Build Zabbix Server HA Cluster in 10 minutes by Kaspars Mednis/ Zabbix Summit Online 2021

Deploying a Zabbix server HA cluster has never been as simple as it is in Zabbix 6.0 LTS with the native Zabbix server HA cluster option. Kaspars Mednis, Zabbix Senior Trainer shows that you can achieve redundancy for your Zabbix server and prevent downtime during hardware failures or planned maintenance. Watch the video to learn more!
More videos from #ZabbixSummitOnline2021 via the link: www.zabbix.com/events/zabbix_...
#zabbix #opensource #monitoring #zabbixmonitoring #zabbixsummitonline

[@HistorianTrevi]

Besides the fact that, as mentioned further below in the comments, the db becomes the single point of failure and this should be stressed out more (regardless of the db being part of not of Zabbix per se since it's nonetheless a key component) also a quick review of two or three examples of architectures (i.e. standard install vs each zabbix component in its own machine vs something else useful i can't think of right now) and one or two simulations of a node switch in order to understand what exactly zabbix does in a real scenario would have been much appreciated (but perhaps this is training material)
Thank you for sharing though, keep up the good work!

[@Kulfaangaren]

Something that I find strange is that HANodeName is pushed in most of the talk but when it comes to removing the node (zabbix_server -R ha_remove_node=) a "node ID" has to be used.

[@dasgoll]

A semicolon is used in passive check.

[@MichaRutkowskiEngineering]

what about trappers? Do i need to send a trap to both instances and only active will receive and the other one will return error?

[@zabbix]

Currently, you may run into a connection error whenever you explicitly specify a standby node as the Zabbix sender destination (-z option)
I would instead suggest pointing the Zabbix sender to a Zabbix agent configuration file instead ( -c option)
Note that the Zabbix agent ServerActive parameter syntax now supports specifying multiple cluster nodes:
www.zabbix.com/documentation/6.0/en/manual/appendix/config/zabbix_agentd
Once you prepare the configuration file, Zabbix sender should be able to pick out the active node from the node list in the Zabbix agent configuration file.

[@kcsj172011]

The db is single point of failure

[@Kulfaangaren]

Yes, it is a single point of failure in your environment, but it is not part of Zabbix release. A Db is a requirement to run Zabbix but is not distributed as part of Zabbix which makes it out of scope for "Zabbix Server HA".

[@samukarbrj]

@@Kulfaangaren yesterday I had a problem in the storage. This HA mode wont have solve it :/

[@Kulfaangaren]

@@samukarbrj What's you point here ? Is your storage part of Zabbix release, no it is not.
What I tried to point out in my above post is that every system has to have a scope and for Zabbix the scope of the HA is to make sure that the parts distributed as part of Zabbix doesn't have a single point of failure. Just as there is HA solution for Zabbix, there are HA solutions for DBs and storage.
With storage you have the free options like GlusterFS and CEPH, and the non-free ones like Quantastor or VMWare.
For DB HA, you can either build a simpler, yet effective, master & slave setup or you could go for something more elaborate like for example use Percona XtraDB to make your MariaDB/MySQL HA.

[@HistorianTrevi]

thank you for the clarification, this fact was troubling me

[@HistorianTrevi]

@Kulfangaaren
thank you for the clarification on Zabbix view of the database but I disagree on it being outside of the scope of ha since, on a practical level, ha is established through the db itself