With regards to other jobs on the server, there is a MySQL server on it as well, which from time to time has some multi-second queries generated from a webserver also on this host, but the MySQL is running with nice 10 (PostgreSQL isn't nice'd).
Do those MySQL queries hit disk hard?
I guess they may be able to do so - the MySQL database is 450 MB, and the server has 512 MB RAM, and some of the queries pretty summarizes everything in the database.
However, I just cross-referenced the access logs from the webserver with the duration logs, and although some of the spikes did happen, while there would have been some MySQL activity (I can't tell for sure, if it was simple queries or the long ones), other spikes happened without any website activity in the surrounding minutes.
I've never seen PostgreSQL have hicups like you describe when running on a machine by itself. I have experienced similar issues when another process (cron job in my case) caused brief swapping to occur.
OK. I may have to try to put the database on a separate server.
Regards, Anders K. Pedersen
---------------------------(end of broadcast)--------------------------- TIP 8: explain analyze is your friend
