Problem with MariaDb Galera Cluster : Deadlock: wsrep aborted transaction

[Arcocide]

Hi,

I would like to get a Zabbix server set up, and we're running into a transient problem with deadlocks coming from the database.

Here's an example of what sorts of things show up in the front-end:

query failed: [1213] Deadlock: wsrep aborted transaction [update item_discovery set lastcheck=1543314650 w.....

I use mariadb galera cluster with 3 nodes and a front load balancing IP.

And if there's any additional information that I can provide that would be helpful, please let me know.

Regads,

[kloczek]

Galera it is more or less only kind of toy.
Required by zabbix performance especially on low latency read and write DB operations makes quite quickly useless active-active DB backends.

[Arcocide]

Galera it is more or less only kind of toy.
Required by zabbix performance especially on low latency read and write DB operations makes quite quickly useless active-active DB backends.

And so it's normal ?

[kloczek]

And so it's normal ?

Unsuccessful roll back some DB updates of the multiple active-active DBs nodes causing some issues like deadlocks or desync when such type of DB is used?
Yes.
More intensively DB backend is used (by write queries like inserts, update and delete queries) than higher probability that such state will be reached. Synchronization overhead and issues caused by desync sometimes are really pain in the a*s.