pg_rewind failing to detect different system in docker container

Question

I have two docker containers running on two different docker hosts.

docker host1: 192.168.10.3

docker host2: 192.168.10.4

Container 1 is running on host1 and container 2 is running on host2. Port 5432 is published on both the nodes. wal_log_hints is on on both the container running PostgreSQL.

Now when i try pg_rewind first time, from container2:

su - postgres -c "/usr/pgsql-14/bin/pg_rewind -D /var/lib/pgsql/14/data --source-server=\"port=5432 user=cysql dbname=test host=container1.myorg.com\" --dry-run"

It fails with below error:

pg_rewind: source and target cluster are on the same timeline
pg_rewind: no rewind required

If i run same things on plain machine RockyLinux 8.6, i get below error for pg_rewind:

pg_rewind: fatal: source and target clusters are from different systems

I know pg_rewind is supposed to use when you sync back to cluster again, but my case is different, i want to first check if pg_rewind can do diff sync, if yes, then go for it, else use pg_basebackup. Not sure why pg_rewind behave different in case of containers and normal machine. Any help please?

Answer 1

The error means that the second database was not created from a file system level backup of the first one. You can only use pg_rewind on two clusters that started out as clones of each other, not on clusters created with separate initdb invocations.

• Niraj Nandane: We ship PostgreSQL image and spin two containers. Please note that initdb is already ran and its part of docker image. We do not run initdb while spinning up container. How i can change the system identifier on already ran initdb?
• Laurenz Albe: Whatever you are spinning, the two database clusters were created by different initdb runs. You cannot change the system identifier. I cannot tell you where you went wrong, because I don't know what exactly you did.