On Friday 01 October 2004 12:26 pm, Shane | SkinnyCorp wrote: > Funny... > > I vacuum full EVERY night @ midnight... > > And yes, that's great about your similar machine with more RAM... > only... does YOUR table have 60+ users @ 120-some queries per > second at any given moment? > > ... > > I didn't need a friggin' tutorial on vacuum man... I need some > advice. I don't need a class in database design either, and I know > quite well how postgresql works, but aside from this I opted to > drop whatever I *think* I know about pgsql, and ask some of the > 'experts' on this list... > > And all I'm getting is shit straight out of the damn manual. Which > I've read. Many times... > > Got any other suggestions? > > - Shane
Um, a new attitude would help - if you knew all the answers you wouldn't be asking the questions. Before biting off the heads of those trying to help you, try reading http://linuxmafia.com/faq/Essays/smart-questions.html by Eric S. Raymond and Rick Moen. I assume from your reading list you also read: http://www.varlena.com/varlena/GeneralBits/Tidbits/annotated_conf_e.html Some of your postgresql.conf settings look _way_ out of the normal range. A couple that jump out: ... > wal_buffers = 8192 # min 4, 8KB each 64 MB for wal_buffers? What analysis led you to set this to over 2000 times higher than the default? Mine is set to 64 buffers which seems fine but of course ymmv. At least it is shared rather than per-process memory so it's not completely killing you. > vacuum_mem = 127072 Also looks way too high. This is settable per connection. Bump it up when running your vacuum fulls if you need to. You only showed pg processes from top. What is the swapping activity on your machine? Do you have memory left over for caching? Cheers, Steve ---------------------------(end of broadcast)--------------------------- TIP 9: the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
