Moving HA Postgres cluster to new region

1

I’d like to move my cluster between two regions. What I just unsuccessfully tried:

  1. Scale from 2 to 4
  2. Add two new pg_data volumes in new region
  3. Dump config, change PRIMARY_REGION to new region
  4. fly deploy using correct image and new config
  5. Scale down to 2

This just resulted in deploy instability and my app being unavailable. I’ve rolled back to my original config (HA in original region) without data loss and my app can connect, although I now have a constant cmd/sentinel.go:276 no keeper info available {"db": "…", "keeper": "…"} on my primary.

What is the correct approach (and could it be documented if it isn’t?)

2

Hey there,

Here’s a short guide on how to achieve this: What is the correct process to change the postgres leader region? - #2 by shaun

cmd/sentinel.go:276 no keeper info available {“db”: “…”, “keeper”: “…”}

This just means that keepers that were once registered with Stolon are no longer available. You’ll see this if you have a node that’s down, or if you’ve scaled down. Generally, this does not cause issues and they will typically be removed automatically after 48 hours, however, there are some edge cases that can be read about here:

1 Like