Re: Exact same output - pg_stat_statements
On Tue, Jan 03, 2023 at 01:52:17PM +0800, Julien Rouhaud wrote: > Resetting the data adds some noticeable overhead as newly added entries will > need to generate a normalize query string and so on. What most people do is > taking regular snapshots of pg_stat_statements (and other stats) view and then > compare the snapshots. There are a few open source projects doing that > available. Why do you assume that the cost of normalization query string will be in any way measurable? Especially in comparison do extra cost of doin math on potentially thousands of rows? Best regards, depesz
Re: Exact same output - pg_stat_statements
On Mon, Jan 02, 2023 at 02:34:13PM +0100, hubert depesz lubaczewski wrote: > On Fri, Dec 30, 2022 at 11:04:59AM -0500, Rushikesh socha wrote: > > Hi, Whenever I am running the below query on one of my Azure PostgreSQL > > PaaS instances I am getting exact same output. I feel it shows old > > information but as far as i know pg_stat_statements only shows current > > information and not past right ? It may be a bug? > > pg_stat_statements has all the data since last reset of stats. > > So if you never reset stats, it accumulated data for howeve rlong you > are using pg. > > Not really surprising that top total-time uses are the same. > > If you want to sensibly use it you should call reset function every now > and then. Resetting the data adds some noticeable overhead as newly added entries will need to generate a normalize query string and so on. What most people do is taking regular snapshots of pg_stat_statements (and other stats) view and then compare the snapshots. There are a few open source projects doing that available.
Re: Exact same output - pg_stat_statements
On Fri, Dec 30, 2022 at 11:04:59AM -0500, Rushikesh socha wrote: > Hi, Whenever I am running the below query on one of my Azure PostgreSQL > PaaS instances I am getting exact same output. I feel it shows old > information but as far as i know pg_stat_statements only shows current > information and not past right ? It may be a bug? pg_stat_statements has all the data since last reset of stats. So if you never reset stats, it accumulated data for howeve rlong you are using pg. Not really surprising that top total-time uses are the same. If you want to sensibly use it you should call reset function every now and then. Best regards, depesz