On Wed, Oct 23, 2013 at 12:11:39PM -0500, andy wrote:
- On 10/23/2013 11:07 AM, David Kerr wrote:
- >On Tue, Oct 22, 2013 at 12:41:58PM -0500, andy wrote:
- >- Hi all.
- >-
- >- My website is about to get a little more popular. I'm trying to add in
- >- some measurements to determine an upper limit of how many concurrent
- >- database connections I'm currently using.
- >-
- >- I've started running this:
- >-
- >- SELECT sum(numbackends) AS count, sum(xact_commit) as ttlcommit FROM
- >- pg_stat_database;
- >-
- >- Every 10 seconds or so. I don't think its what I want though. It seems
- >- way too small. I'm guessing that its not a measure of the previous 10
- >- seconds. Its a count of how many backends are in use at the exact
- >- moment I run the sql.
- >-
- >- Is there a cumulative count someplace?
- >-
- >- Thanks for your time,
- >-
- >- -Andy
- >
- >You've gotten good info from the other guys on how to scale your're DB
- >but to answer you're original question. I usually use this query:
- >
- >select count(*) from pg_stat_activity where state <> 'idle';
- >
- >That gives you the # of "active" connections to your database and is
- >something you want to try to get good metrics on.
- >
- >Idle connections have some overhead but if Active > # of CPUs your
- >performance
- >starts to degrade. Now, really that's pretty normal but, ideally, you need
- >to
- >know what the ratio of Active Connections to # CPUs still gives you
- >acceptable
- >performance. And that's really based on your app and hardware.
- >
- >
-
- How often do you run that? Once a second? And graph it? I was doing
- it every 10 seconds, but it doesn't give me a good view of the system.
I actually have it as a munin module so it runs every few minutes.
If I'm actually doing a performance test or something I would run it
every second or every 5 / 10 seconds.
The knowledge is only really useful if you have a good trend built up so you
know how your app responds to having a certian # of active connections.