Database and API replicas

[phobie]

Hello,

i am currently working on how to scale 5.0 deployment on API level. With postgres & timescale, logical replication will not work as DDL will need to be replicated manually so only read-only streaming replication for DB side.
I need grafana-zabbix plugin working, so every DB replica can have its own API, but so far, as replicas will be read-only, there is at least login/session problem on read-only replica. Anyone has somehow overcome this?

Another solution might be to put pgpool-II instance infront of API and route all read-only requests to replicas, so metrics and session will be written to master, and rest of queries will go to slaves.

I consider timescaledb easier to maintain as trigger based partitioning and it might suit our size better (6k NVPS, 3500 hosts).

I will be gratefull for any comments on multiple API topic.

Thanks a lot.

[splitek]

Grafana (grafana-zabbix plugin) connects to Zabbix frontend, all API calls go throught Zabbix frontend. In plugin is option "direct db" but as far i known it's used only for reading data (probably only metrics).