Oleg Bartunov <oleg@sai.msu.su> writes:
> ERROR: No one parent tuple was found
> Everything seems work fine. How dangerous this error ?
Not very. This comes out of the VACUUM code that tries to deal with
maybe-not-quite-dead tuples (stuff that's been updated and the update
has committed, but there is at least one open transaction that's old
enough that it should see the pre-update tuple state if it decides to
look at the table). You can ensure that none of that logic runs if you
simply perform the VACUUM with no other transactions open.
I think there are probably still bugs in that area in current
sources :-( ... it's not exercised enough.
regards, tom lane