Closed jimnydev closed 4 months ago
Just specify the variables when deploying the cluster:
with_haproxy_load_balancing: true
pgbouncer_install: false
In this case, the pgbouncer will not be installed, and connections via HAProxy will be routed directly to the database.
I would still like to have pgbouncer for connection pooling for other applications that are doing regular PostgreSQL protocol.
Now I understand that you are interested in the possibility of launching "selected" connections without the participation of a puller in a scenario using pgbouncer.
This feature will require further development. Feel free to suggest a PR.
Resolved via #549. Thank you!
Prior to submitting a PR and considering a deployment with pgbouncer: I was wondering if you would find it useful to expose Primary and Replica nodes directly (skipping pgbouncer) in HA Proxy config (using two new ports)?
Use case would be for applications that need to connect using
replication
protocol (which currently isn't supported by pgbouncer, and is unlikely to be supported for a while).Thanks!