I posted this on StackOverflow and it was suggested this query was better suited here.
I'm trying to encourage the use and monitoring of autovacuum in some PostgreSQL 8.3 databases.
One objection I hit often is that people don't "trust" autovacuum or there are bugs in autovacuum in 8.3 which mean that it's ignored in preference to scheduling vacuuming. Mostly our tables are small and this approach appears to work. However, with our larger (& also heavily updated tables) this really doesn't work (dead tuple counts increase, exceed max_fsm_pages, and the tables don't get cleaned up etc etc).
I'm just wondering if anyone has a reference for autovacuum in 8.3 being buggy or not working. My own experience has shown that autovac works fine and, where necessary, adding entries to the pg_autovacuum table does the trick.
I'd like to understand the problem with autovacuum (if one exists).