Re: atrocious update performance - Mailing list pgsql-performance

From Rod Taylor
Subject Re: atrocious update performance
Date
Msg-id 1079387674.33643.12.camel@jester
Whole thread Raw
In response to Re: atrocious update performance  ("Rosser Schwarz" <rschwarz@totalcardinc.com>)
List pgsql-performance
On Mon, 2004-03-15 at 16:15, Rosser Schwarz wrote:
> > > # explain update account.cust set prodid = tempprod.prodid
> > >     where tempprod.did = origid;
>
> > >  Merge Join  (cost=0.00..232764.69 rows=4731410 width=252)
> > >    Merge Cond: (("outer".origid)::text = ("inner".did)::text)
> > >    ->  Index Scan using ix_origid on cust  (cost=0.00..94876.83
> > >        rows=4731410 width=244)
> > >    ->  Index Scan using ix_did on tempprod  (cost=0.00..66916.71
> > >        rows=4731410 width=18)
>
> > I'm going to hazard a guess and say you have a number of foreign keys
> > that refer to account.cust.prodid? This is probably the time consuming
> > part -- perhaps even a missing index on one of those keys
> > that refers to
> > this field.
>
> Actually, there are no foreign keys to those columns.  Once they're
> populated, I'll apply a foreign key constraint and they'll refer to the
> appropriate row in the prod and subprod tables, but nothing will
> reference account.cust.[sub]prodid.  There are, of course, several foreign
> keys referencing account.cust.custid.

If there are no feign keys to it, I wouldn't expect it to take more than
10 minutes on slow hardware.

Fresh out of ideas here.



pgsql-performance by date:

Previous
From: Tom Lane
Date:
Subject: Re: High CPU with 7.4.1 after running for about 2 weeks
Next
From: Stephen Robert Norris
Date:
Subject: Re: Scaling further up