What I'm about to say may be total garbage, but I'm just telling it how
it happened.
I'm using Postges to log certain customer activities on our web site -
logins, things like that. I vacuum once a night. Now the database has
grown quite large, the main table alone is 180 Mb. It contains lots and
lots of small rows.
I've been finding recently that my postmaster processes have been eating
up more and more CPU, and I've been wondering what to do about it (the
connections to the database are pooled and kept open for long
periods). The other (perhaps strange) thing is, we hardly ever do
queries on this thing. Gathering stats is still on my to-do list.
Anyway, a few minutes ago I ran a single query. After it finished, I
noticed that the CPU usage had dropped dramatically. Like from a load
average of about 5 to about 0.8. Now I'm wondering if the query I ran
was the cause of this. Will postgres slow down if the only operations
being performed are inserts?
It may just be that activity on the site dropped off at the same time a
ran the query, but the number of clients seemed to be pretty normal.
Is there any validity to this idea?
--
Lachlan O'Dea <mailto:lodea@vet.com.au> Computer Associates Pty Ltd
Webmaster Vet - Anti-Virus Software
http://www.vet.com.au/
"It is better to light a candle than to curse the darkness."
- Carl Sagan