On Tue, 2009-02-03 at 20:46 +0800, Phoenix Kiula wrote: > This is PG 8.2.9
VACUUM spoils the cache in 8.2 What happens is that VACUUM reads through the whole table, knocking other blocks out of cache. These then need to be read in again by other processes, so there is some I/O thrashing. If your bgwriter settings are ineffective then normal users will also need to write the dirty blocks left by VACUUM and probably flush WAL as well while doing it, using even more I/O. We fixed this in 8.3 so that VACUUM uses at most 256KB of memory as it goes, which makes it both faster because of CPU L2 cache effects and hardly spoils shared_buffer cache at all. Bgwriter is also better tuned so it will handle dirty blocks better. -- Simon Riggs www.2ndQuadrant.com PostgreSQL Training, Services and Support -- Sent via pgsql-general mailing list (pgsql-general@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
