Re: Assessing performance of fetches - Mailing list pgsql-performance

From Tom Lane
Subject Re: Assessing performance of fetches
Date
Msg-id 15055.1303155679@sss.pgh.pa.us
Whole thread Raw
In response to Assessing performance of fetches  (John Rouillard <rouilj@renesys.com>)
List pgsql-performance
John Rouillard <rouilj@renesys.com> writes:
> I am seeing:

>   2011-04-16 00:55:33 UTC user@database(3516): LOG:  duration:
>      371954.811 ms  statement: FETCH FORWARD 1 FROM c_2aaaaaaeea50_a08

> While I obviously have a problem here, is there any way to log the
> actual select associated with the cursor other than logging all
> statements?

Can't think of one :-(

> Also once I have the select statement, does the fact that is is
> associated with a fetch/cursor change the steps I should take in
> tuning it compared to somebody just issuing a normal select?

The planner does treat cursor queries a bit different from plain
queries, putting more emphasis on getting the first rows sooner.
If you want to be sure you're getting the truth about what's happening,
try
    EXPLAIN [ANALYZE] DECLARE c CURSOR FOR SELECT ...
rather than just
    EXPLAIN [ANALYZE] SELECT ...
Other than that, it's the same as tuning a regular query.

            regards, tom lane

pgsql-performance by date:

Previous
From: John Rouillard
Date:
Subject: Assessing performance of fetches
Next
From: "Pierre C"
Date:
Subject: Re: Background fsck