Hi, Matthew,
Matthew T. O'Connor wrote:
> I'm sure he did, but just because you don't have enough FSM space to
> capture all everything from your "burst", that doesn't mean that space
> can't be reclaimed. The next time a regular vacuum is run, it will once
> again try to fill the FSM with any remaining free space it finds in the
> table. What normally happens is that your table will never bee 100%
> free of dead space, normally it will settle at some steady state size
> that is small percentage bigger than the table will be after a full
> vacuum. As long as that percentage is small enough, the effect on
> performance is negligible.
This will work if you've a steady stream of inserts / updates, but not
if you happen to have update bulks that exhaust the FSM capacity. The
update first fills up all the FSM, and then allocates new pages for the
rest. Then VACUUM comes and refills the FSM, however, the FSM does not
contain enough free space for the next large bulk update. The same is
for deletes and large bulk inserts, btw.
So your table keeps growing steadily, until VACUUM FULL or CLUSTER comes
along to clean up the mess.
Markus
--
Markus Schaber | Logical Tracking&Tracing International AG
Dipl. Inf. | Software Development GIS
Fight against software patents in EU! www.ffii.org www.nosoftwarepatents.org
