There are two basic configurations for pg_statsinfo in warm-standby.
- An independent repository server from active and standby servers
- Each repository servers in active and standby servers
pg_statsinfo will run even on the standby server and parse server logs written in standby. However, it doesn't take any snapshots because the standby server doesn't accept connections.
1. An independent repository server from active and standby servers
You can put the repository to the third server independing from the active and the standby servers. This configuration enables to access the same repository always no matter which server is the active.
In warm-standby
After failover
2. Each repository servers in active and standby servers
If you don't have no more servers other than the active and the standby servers, you can put the repositories together with them. Snapshot from the active server will be stored in the server. On the other hand, the standby server cannot store any statistics, but it will not a problem because there are no checkpoints nor autovacuums in the standby server.
Since snapshots for the active server will be also replicated, you can analyze the snapshots after failover from the previously-standby server (in active then).
In warm-standby
After failover
