<mallah@trade-india.com> writes:
> Hmm i do the below daily to this table. And this is pgsql 7.3
> TRUNCATE TABLE public.site_search;
Drat. Now that I look at it, 7.3 TRUNCATE doesn't send any kind of
relcache inval message, so it's got the same kind of problem that the
nailed-in-cache catalogs do: other backends don't reset their
which-page-to-insert-on pointer.
This is already fixed for 7.4, because in 7.4 TRUNCATE updates the
pg_class entry.
We decided not to backpatch any fix for the system-catalog version
of the problem, because it is purely cosmetic (there is no problem
other than the warning messages). So I guess I think the same about
this version of the problem.
regards, tom lane