On Wed, Mar 30, 2005 at 13:09:33 -0500,
Adam Siegel <adam@sycamore.us> wrote:
>
> We perform a vacuum full after each mass delete. This cycle can happen
> many times during over a couple of weeks. We are in a test lab
> environment and are generating a lot of data.
>
> One of the problems we have is that the vacuum on the table can take up
> to 10 hours. We also expect to see the physical disk space go down, but
> this does not happen. If we accidently fill up the disk, then all bets
> are off and we are unable to recover. A vacuum never seems to finish
> (several days).
This may mean that there are open transactions pinning the records you
have deleted so that they aren't being removed by the vacuum.
Also, under some circumstances a CLUSTER can be faster than a VACUUM FULL.
