Zabbix Server with Patroni Cluster and Load Balancer ends up in read only transaction

[lapniklas]

Hello,

we are running a Zabbix-HA Cluster (2 Zabbix-Servers) which connect to a Patroni (Postgres) cluster through a Load Balancer.

But when a fail over happens on the postgres cluster the primary Zabbix Server ends up in a read-only transaction which generates a lot of Errors in the Serverlog. Is there any way to detect this behavior and failover the cluster to the secondary node?

Note that we did not noticed this error since the zabbix server was unable to send Notifications since they also end up in the readonly transaction and so they were never written to the database and also never send. We just randomly found huge log files on the zabbix server.

Zabbix Server Version: zabbix_server (Zabbix) 6.0.8

[freddareg45]

As i know check the status of your Patroni cluster and verify that all the nodes are healthy and in sync. You can use the Patroni command-line tool or the Patroni REST API to check the cluster status and look for any errors or warnings. And ensure that the connection between the Zabbix Server and the database cluster is stable and not experiencing any network issues. Test the connectivity by connecting directly to the database nodes using the appropriate client tool.

[cyber]

Looks like your loadbalancer still sends queries to "old" db node? Does not understand of switchover quickly enough? I don't think there would be any point in switching zabbix server to other host...

[lapniklas]

I am using a managed Load Balancer in the hetzner cloud. I already adjusted the configuration of the load balancer. I just want some kind of backup solution that restarts the server when it is not switched to the new node correctly.