If set have the max_fsm_pages set correctly, that makes VACUUM faster and removes the need for VACUUM FULL or REINDEX. Should be slightly more than the total number of data pages which will be touched by updates and deletes between vacuums.
I have a followup question. What if the set of dead tuples is too big and I need to VACUUM FULL, as opposed to VACUUM. (The update size varies greatly from day to day.) Will the clustering effectively do a VACUUM FULL, or just a VACUUM?
Thanks again for your help, Mark
---------------------------(end of broadcast)--------------------------- TIP 3: Have you checked our extensive FAQ?