On Wed, Feb 2, 2011 at 10:45 AM, Chris Browne <cbbro...@acm.org> wrote: > peter.geoghega...@gmail.com (Peter Geoghegan) writes: >> I'm not sure why you'd advocate CLUSTER as a way to reclaim disk space. > > Because it works pretty well; it reorganizes the table on the basis of > the order indicated by one index, and simultaneously: > a) Shortens the table, removing all dead space; > b) Regenerates all indices, so they too have no dead space.
It's important at this point to set fill factor before the cluster if something besides the default 100% makes sense. any randomly updated table full of small records will usually benefit from a fill fact even as high as 95% which is very little "wasted" space for a gain in HOT updates starting in 8.3. HOT saved our bacon at work. They really lowered the requirements for disk access / index update a LOT. I wish I'd have saved the pg_stat_index from 8.1 versus 8.3. And IO numbers. Our load dropped by a power of ten more or less. -- Sent via pgsql-general mailing list (pgsql-general@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
