fix: postgres health check reporting any db without checksums as unhealthy

[mertalev]

Description

This is a small change to make the health check respond as healthy if checksums are disabled and it gets a successful response from Postgres. It currently bogusly reports existing databases as unhealthy.

Fixes #10134

[mmomjian]

Ty. I should have known everyone would just copy paste the new changes haha

[reini12345]

I still have the issue with v1.106.4

[zackpollard]

I still have the issue with v1.106.4

That is actually reporting checksum errors it seems

[reini12345]

did not have this before the 1.106 update

[zackpollard]

did not have this before the 1.106 update

We didn't have health checks before the 1.106 update, so these issues wouldn't have been visible before then. This is not a bug, the recently added checksums are reporting an inconsistency in your database data

[mmomjian]

I still have the issue with v1.106.4

This looks fine to me. Please use the updated healthcheck command found in this PR that contains the COALESCE command

The failure_count is a docker internal counter that increments with successive failed health checks. If it was a checksum error, it should appear in the message itself. In this case the result is NULL which appears as empty (checksums disabled).

[zackpollard]

I still have the issue with v1.106.4

This looks fine to me. Please use the updated healthcheck command found in this PR that contains the COALESCE command

The failure_count is a docker internal counter that increments with successive failed health checks. If it was a checksum error, it should appear in the message itself. In this case the result is NULL which appears as empty (checksums disabled).

Ah, well that's not confusing at all 🤣 It did make me think though, do we have docs to help people recover if they do end up with a legitimate checksum error? Or at least somewhere to point them at in the postgres docs?

[mmomjian]

Practically speaking, the answer is to restore from backup. The checksums are only sufficient to detect corruption, not to restore it. I can add that to the FAQ entry about it though.

[reini12345]

thx. With the updated healthcheck command the conatiner is now "healthy"

For people like me who have no idea about postgres, more documentation would not be bad. I had already started looking for a way to repair the database.