Re: Better HINT message for "unexpected data beyond EOF"

On Wed, Mar 26, 2025 at 4:59 AM Jakub Wartak
<jakub(dot)wartak(at)enterprisedb(dot)com> wrote:
> ERROR: unexpected data beyond EOF in block 1472 of relation base/5/16387
> HINT: This has been seen to occur with buggy kernels; consider
> updating your system.
>
> to something more generic and less confusing. It is coming from
> ffae5cc5a602 (2006), and we are probably not running those "buggy"
> kernels anywhere. I've seen this error multiple times, but it is
> usually due to some external influence overwriting/replacing the files
> in PGDATA and some (potentially new) backends open()ing those "new"
> files and finding unexpected file layout. In the real world this means
> usually:
> a. files being potentially accidentally replaced/overwritten, please
> see attached file for reproducer
> b. some obscure bugs (e.g. in EPAS - PG fork - we have on-demand
> automatic partition creation and we had bug/race conditions where
> multiple backends end up writing to the same relfilenode oid file)
>
> so how about:
> -HINT: This has been seen to occur with buggy kernels; consider
> updating your system.
> +HINT: This has been observed with files being overwritten, buggy
> kernels and potentially other external file system influence.

I agree that we should emphasize the possibility of files being
overwritten. I'm not sure we should even mention buggy kernels -- is
there any evidence that's still a thing on still-running hardware? I
don't really like "other external file system influence" because that
sounds like vague weasel-wording.

--
Robert Haas
EDB: http://www.enterprisedb.com