Josh Berkus <josh@agliodbs.com> writes:
> So, I'll ask again (because I didn't see a reply): is there any way
> users can *check* if they've been corrupted? Short of waiting for PK/FK
> violations?
I think it would work to do a REINDEX on each table (doesn't matter which
index you select) and see if it complains about not being able to find any
parent tuples. This would definitely catch the case as long as the
orphaned HOT tuple is still live. If it's not, the immediate problem is
gone ... but it's still possible you have follow-on corruption.
regards, tom lane