Alvaro Herrera <alvherre@commandprompt.com> writes:
> Excerpts from Brad Nicholson's message of vie ago 06 12:01:27 -0400 2010:
>> It found 45878 dead tuples in 396 pages for the index authors_archive_pkey.
>> It found 16558 dead tuples in 492 pages for the table authors_archive.
> But why did it choose to skip the rest of the pages in authors_archive,
> if there certainly are a lot of vacuumable tuples in (some of) them?
I think the discrepancy is probably explained here:
/*
* DEAD item pointers are to be vacuumed normally; but we don't
* count them in tups_vacuumed, else we'd be double-counting (at
* least in the common case where heap_page_prune() just freed up
* a non-HOT tuple).
*/
That last message prints tups_vacuumed, but those other ones are counting
all the removed item pointers. So apparently Gordon had a whole lot of
pre-existing DEAD item pointers. I wonder why ...
regards, tom lane
