Adding Additional Databases

A PostgreSQL cluster usually hosts more than one logical database. The hosted onboarding auto-discovers them once you authenticate with admin credentials — you pick which to monitor, subject to your plan's database quota. The one you're connecting to during the wizard is just the entry point; nothing about that initial pick is special, you can flip databases on and off from Settings → Servers later without re-running setup.

Plan limits

Free is 1 server with 3 databases on it. Pro is 10 servers and up to 50 databases total. Business is 50 servers and 500 databases. Enterprise removes the cap entirely (contact sales). Going over the database quota blocks new adds — existing ones keep collecting, you don't lose data, but you can't add a sixteenth database on a Pro plan that has fifty enabled across other servers.

Adding a second cluster

Settings → Servers → + Add server. The onboarding wizard opens fresh; MonPG treats each server as an independent monitoring target with its own monpg_monitor role. There's no shared state across clusters by design — that way an issue with one cluster's role doesn't cascade.

Multi-DB fan-out

MonPG collects metrics per logical database, not just per cluster. If you're monitoring 16 databases on one cluster, the collector opens 16 short-lived connections per cycle to pull the per-database stats: pg_stat_statements, pg_stat_user_tables, pg_stat_user_indexes. The Databases table on the server's settings page shows per-DB query count and size so you can see the fan-out worked.

This was a real bug we hit in 2026-04: the snapshot was being written with one database's worth of data even when 16 were configured. The fix (SD-032) made fan-out mandatory in every collector path, which is why the Databases table looks the way it does now — it's the verification surface.

Aurora and RDS clusters

For Aurora, connect to the writer endpoint. Read replicas are auto-detected via pg_stat_replication and show up on the Replication page; you don't need to add them as separate servers unless you want collection parallelism (in which case treat each reader as its own server). For non-Aurora RDS clusters with replicas, same pattern — connect to the primary, read replicas show up automatically.