Upgrade Postgres on persistent volumes

We have a small Stolon setup with 3 keepers, a Sentinel and a proxy. Keepers run PG11.
In order to upgrade to PG14 I am exploring a pg_dump/pg_restore upgrade path. Since we use AWS persistent volumes, the idea is to:

  • pg_dump on the stolon-proxy (or should it be the master?)
  • bring the cluster down
  • mount the 3 volumes in a PG14 container
  • replace the /postgres directory with a fresh initdb
  • make changes in pg_hba.conf since we have some custom entries there.
  • bring the cluster back up
  • pg_restore on the proxy (or should it be the master?)

Is this anywhere near an accepted way of upgrading?

stolon-proxy always connects to the master.