Kieren Scott <kierenscott(at)hotmail(dot)com> wrote:
> I'm trying to understand what is going on internally when doing a
> VACUUM FULL on a table in 8.3.
>
> I have a table that is 1GB in size, 500M is used, and 500M is free
> space. When I do a vacuum full on this table, will it either: -
>
> 1) Compact all of the used tuples into free space within the
> existing disk file and then shrink the file to 500M. Therefore
> simply freeing up 500M in the disk file.
On 8.3, this is what it will do, although it can take a very long
time. I've given up on this before completion (sometimes after
leaving it cranking away for a couple days) every time I've tried it
on a table with more than a few GB and any significant bloat. I
don't know that I've ever tried it on a table as small as you
describe, but I would bet that CLUSTER is going to be much faster if
you have the half a GB free space. Another issue with VACUUM FULL
is that it bloats indexes; so you generally need to follow it with a
REINDEX on the table.
-Kevin