Scott Whitney <scott(at)journyx(dot)com> wrote:
> vacuumdb -f -v -z -F template0
VACUUM FULL (the -f option) is almost always a bad idea, for many
reasons. I wouldn't be surprised if it somehow messed you up.
I would schedule a database REINDEX on any databases where you used
the -f option, and then re-evaluate, if you feel compelled to push
this issue. I still doubt that this effort will buy you anything in
recoverability. Between the hint bits and the frozen xmin in all
the tuples, the clog won't be consulted anyway.
If the pursuit of this is causing you to do direct updates of system
tables and to resort to VACUUM FULL, you're putting your data at
risk for no apparent benefit.
-Kevin