Lonni J Friedman <lfriedman@vasoftware.com> writes:
>> Oh well. Have you checked that the reportedly-broken index is in fact
>> the index for the TOAST table associated with the artifact_file table?
> i'm not sure that i'm following you 100% here. I'm able to dig this up:
Yeah, it looks like 302323 is the OID of artifact_file, so those are the
right TOAST tables for it.
> so i tried:
> insert into pg_toast_302323 values (7685119, 1, '');
> but that fails:
> ERROR: You can't change toast relation pg_toast_302323
Ooops, I was thinking that you could bypass that check as superuser,
but evidently not.
You're probably going to have to do this the hard way: figure out by
elimination which field(s) of which row(s) reference the broken toast
entries, and reset those fields to null (or some valid value, anyway).
It disturbs me quite a bit that the problem seems to be moving around.
That suggests you've got ongoing corruption. Might be a good idea to
run disk and memory diagnostics to find out if there are hardware
problems.
regards, tom lane