Re: BUG #17268: Possible corruption in toast index after reindex index concurrently

On Wed, Nov 3, 2021 at 3:26 AM PG Bug reporting form
<noreply@postgresql.org> wrote:
> it turns out one entry in toast index (?) was corrupted:
> select  md5(body) from zz where id = ...;
> ERROR:  missing chunk number 0 for toast value 4040061139 in
> pg_toast_2624976286
>
> According to "created_at" column in linked table row was created at
> "2021-11-02 13:04:22.192125", i.e. during reindex concurrently.

I wonder if it's a coincidence that that number (~4.04 billion) is not
that far from 2^32-1 (~4.294 billion).

Can you run amcheck? Perhaps the output of the following will be interesting:

create extension amcheck;
set client_min_messages=debug1;
select bt_index_check('pg_toast.pg_toast_2624976286_index', true);

(Couldn't hurt to try it, at least.)

> I'm wondering if it's known bug and how risky could it be to reindex toast's
> indexes. It was done automatically with tool which monitors indexes' bloat
> and index size reduced several times in this case.

If I had to guess, I'd guess that this is a new and unknown bug.

-- 
Peter Geoghegan