So I tried to install TimescaleDB extension to my existing postgresql app by deploying a changed version of this repo. Of course, it didn’t work, because I am a Docker noob.
Then I decided that I will “rollback” it by removing all my changes and doing fly deploy from that repo. That didn’t work either.
Then I decided that I will create another postgresql app from a snapshot from time when my original postgresql app was still working.
I did LOG_LEVEL=debug fly pg create --verbose --snapshot-id <SNAPSHOT_ID> but the response was:
Our PG restore feature doesn’t support restores for apps running a custom image. We should have a better error for this.
Do you know what version of postgres you were running before? You can revert to our standard postgres build by running fly deploy -i flyio/postgres:<version> --strategy immediate.
Thank you for your reply @kurt
After doing that I can see cmd/keeper.go:719 cannot get configured pg parameters {"error": "dial unix /tmp/.s.PGSQL.5433: connect: no such file or directory"} in my logs.
I am not sure what changed, because I tried quite a lot of things, but now it’s working. Even the original DB app is OK, so I didn’t lost any data.
I am bit confused, but happy.
@stefan.luptak Sorry for not getting back to you earlier! The issue with your original DB was that the PRIMARY_REGION environment variable was set to the incorrect region. I tweaked that for you and things came back to life. However, i’m still a little puzzled by the panic… I’ll keep investigating and let you know if I find anything.
Thanks a lot for your help @shaun. Probably better error message would help me to solve it by myself. I know you know there’s a room for improvement on the Fly.io side, I just had to mention it.