I queried pg_stat_activity a bunch of times and was able to tell what tables were being queried. I finally honed in on the section of code causing the problem and fix it and all is good on the server now - over 90% idle.
I think the reason some of the processes were showing high CPU usage was that they are not reporting real time usage but some kind of average. My guess is that some of the processes that showed high usage were not the sources of the CPU problems I was currently having.
Thanks all.
On 11/22/06,
adey <
adey11@gmail.com> wrote:
The following query with the stats_command_string parameter turned on will give you some of the SQL per transaction:-
SELECT
datid,
datname as "DB Name",
substr(procpid,1,6) as "Procpid",
substr(usesysid,1,5) as "UseSysid",
usename,
current_query as "SQL",
query_start
FROM
pg_stat_activity
order by
procpid;
On 11/22/06, Aaron Bono <postgresql@aranya.com > wrote:I have a couple processes/connections to one of our databases that appears to be eating up most of the CPU and we are trying to determine what these processes are doing that is taking so much CPU time.
Is there a way to monitor the SQL being run for a specific connection/process?
We are using PostgreSQL 8.1.5 on CentOS Linux.
--
==================================================================
Aaron Bono
Aranya Software Technologies, Inc.
http://www.aranya.com http://codeelixir.com==================================================================