Re: stats collector process high CPU utilization - Mailing list pgsql-performance

From minshall@intellicon.biz
Subject Re: stats collector process high CPU utilization
Date
Msg-id 62089.75.14.32.120.1170994418.squirrel@intellicon.biz
Whole thread Raw
In response to Re: stats collector process high CPU utilization  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: stats collector process high CPU utilization  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-performance
> Benjamin Minshall <minshall@intellicon.biz> writes:
>> Tom Lane wrote:
>>> How many tables are visible in the pg_stats views?
>
>> There are about 15 databases in the cluster each with around 90 tables.
>>   A count of pg_stats yields between 500 and 800 rows in each database.
>
> Sorry, I was imprecise.  The view "pg_stats" doesn't have anything to do
> with the stats collector; what I was interested in was the contents of
> the "pg_stat_xxx" and "pg_statio_xxx" views.  It'd be enough to check
> pg_stat_all_indexes and pg_stat_all_tables, probably.  Also, do you have
> the 8.1 installation still available to get the comparable counts there?
>

I checked all 15 databases on both 8.1 and 8.2; they were all quite
consistent:

pg_stat_all_indexes has about 315 rows per database
pg_stat_all_tables has about 260 rows per database

The pg_statio_* views match in count to the pg_stat_* views as well.

While exploring this problem, I've noticed that one of the frequent insert
processes creates a few temporary tables to do post-processing.  Is it
possible that the stats collector is getting bloated with stats from these
short-lived temporary tables?  During periods of high activity it could be
creating temporary tables as often as two per second.

pgsql-performance by date:

Previous
From: Tom Lane
Date:
Subject: Re: stats collector process high CPU utilization
Next
From: Tom Lane
Date:
Subject: Re: stats collector process high CPU utilization