So if I have a Postgres cluster in dev mode (single node), and I clone it into two regions:
- Will those new machines be read only and
- Will those new machines scale to zero
Yes, any machines in other regions will come up with ROLE=replica. (This is for PG Flex, at least—not sure about the older Stolon variety.)
$ fly status
ID STATE ROLE REGION IMAGE
b915aabbaa0453 started primary ewr flyio/postgres-flex:15.3
abcd091290fedc started replica bos flyio/postgres-flex:15.3
This part I would shy away from… The metadata for that does get preserved by clone, but stopped replicas will cause WALs to accumulate back on the primary, as I understand it:
https://community.fly.io/t/runaway-pg-wal-disk-usage/17438/2
When dabbling in the past, I did eventually see…
Warning: Inactive replication slot repmgr_slot_1545901680 is retaining 283 MB of WAL
And the size of /data/postgresql/pg_wal/ really was increasing (until the replica was hastily brought back online).
Ok good to know. I don’t really want to have them scale to 0, but I was curious how that worked since with “regular” machines I just set minimum machines to 1, and the other regions scale down. I didn’t know what happened for the PG instances behind the scene.
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.