Re: DB sessions 100 times of DB connections - Mailing list pgsql-performance

From Steve Crawford
Subject Re: DB sessions 100 times of DB connections
Date
Msg-id 53BC1C3F.1000407@pinpointresearch.com
Whole thread Raw
In response to Re: DB sessions 100 times of DB connections  ("Huang, Suya" <Suya.Huang@au.experian.com>)
List pgsql-performance
On 07/03/2014 06:59 PM, Huang, Suya wrote:

BTW, I’m using the pgbadger report to check for peak connections/sessions.

 

From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Huang, Suya
Sent: Friday, July 04, 2014 11:44 AM
To: pgsql-performance@postgresql.org
Subject: [PERFORM] DB sessions 100 times of DB connections

 

Hi,

 

We’ve experienced a DB issue yesterday and after checked the log found that the peak sessions is 3000 while the peak DB connections is only around 30. The application is having problem of pulling data but no warnings in DB log as it doesn’t exceed max_connections.

 

 

How could this happen? How does sessions/connections work in Postgres?

 

As handy as pgbadger is, I have found that its max-connections values don't pass the "sniff test" as it generally shows peak values that exceed the configured number of connections. I haven't dug in to find out why but could conjecture that the fact that log entries are generally truncated to the nearest second could cause this sort of thing.

Unexpected connection buildup is often a side-effect of something else like a large resource-intensive query, a query holding locks that prevent the other connections' queries from completing or a variety of other things.

If you are looking to solve/prevent the undescribed "issue", please provide more detail.

-Steve

pgsql-performance by date:

Previous
From: Sékine Coulibaly
Date:
Subject: Re: PGSQL 9.3 - billion rows
Next
From: Josh Berkus
Date:
Subject: Re: PGSQL 9.3 - billion rows