Re: [PERFORM] Best way to load test a postgresql server

2009-06-11 Thread Erik Aronesty
Technically you can then use pgbench on that set of statements, but I usually just use perl's "Benchmark" module (i'm sure ruby or java or whatever has a similar tool) (First, I log statements by loading the application or web server with statement logging turned on so I'm not "guessing" w

Re: [PERFORM] degenerate performance on one server of 3

2009-06-05 Thread Erik Aronesty
> See ALTER TABLE and CREATE TABLE  (and the Index variants). > > ALTER TABLE foo SET (fillfactor=90); I'll try that. > This will leave on average, 10% of every 8k block empty and allow updates to > columns to more likely live within the same block. Good for the items table. Probably bad for th

Re: [PERFORM] degenerate performance on one server of 3

2009-06-04 Thread Erik Aronesty
> read the entry on pg_stat_all_tables yeah, it's running ... vacuum'ed last night it's odd, to me, that the performance would degrade so extremely (noticeably) over the course of one year on a table which has few insertions, no deletions,and daily updates of an integer non null column (stock lev

Re: [PERFORM] degenerate performance on one server of 3

2009-06-03 Thread Erik Aronesty
estination = 'syslog' client_min_messages = notice should be enought to get it going and for me to see it right? not sure which setting controls logging of autovac, nor am i sure of a way to *ask* the server if autovac is running. On Mon, Jun 1, 2009 at 10:06 AM, Tom Lane wrote: >

Re: [PERFORM] degenerate performance on one server of 3

2009-05-31 Thread Erik Aronesty
it was all vacuum full...thanks the other 2 servers truncate and reload that table from time to time ... IE: they are always vacuumed as the "master" ... that server never does it... hence the bloat but why wasn't autovac enough to reclaim at least *most* of the space? that table *does* get up

[PERFORM] degenerate performance on one server of 3

2009-05-31 Thread Erik Aronesty
I have 3 servers, all with identical databases, and each performing very differently for the same queries. www3 is my fastest, www2 is the worst, and www1 is in the middle... even though www2 has more ram, faster CPU and faster drives (by far), and is running a newer version of postgres. I have be

Re: [PERFORM] Hosted servers with good DB disk performance?

2009-05-26 Thread Erik Aronesty
On Tue, May 26, 2009 at 6:28 PM, Craig James wrote: > Greg Smith wrote: >> >> What I'd love to have is a way to rent a fairly serious piece of dedicated >> hardware, ideally with multiple (at least 4) hard drives in a RAID >> configuration and a battery-backed write cache.  The cache is negotiable