Re: 7 hrs for a pg_restore?

[Erik Jones <erik(at)myemma(dot)com>]

On Feb 19, 2008, at 2:55 PM, Douglas J Hunley wrote:

> On Tuesday 19 February 2008 15:07:30 Jeff wrote:
> > On Feb 19, 2008, at 1:22 PM, Tom Lane wrote:
> > > maintenance_work_mem, to be more specific.  If that's too small it
> > > will
> > > definitely cripple restore speed.  I'm not sure fsync would make  
> > > much
> > > difference, but checkpoint_segments would.  See
> > > http://www.postgresql.org/docs/8.3/static/populate.html#POPULATE-PG-
> > > DUMP
> >
> > I wonder if it would be worthwhile if pg_restore could emit a warning
> > if maint_work_mem is "low" (start flamewar on what "low" is).
> >
> > And as an addition to that - allow a cmd line arg to have pg_restore
> > bump it before doing its work?  On several occasions I was moving a
> > largish table and the COPY part went plenty fast, but when it hit
> > index creation it slowed down to a crawl due to low maint_work_mem..
>
> fwiw, I +1 this
>
> now that I have a (minor) understanding of what's going on, I'd  
> love to do
> something like:
> pg_restore -WM $large_value <normal options>

pg_restore is a postgres client app that uses libpq to connect and,  
thus, will pick up anything in your $PGOPTIONS env variable.  So,

PGOPTONS="-c maintenance_work_mem=512MB" && pg_restore ....

Erik Jones

DBA | Emma®
erik(at)myemma(dot)com
800.595.4401 or 615.292.5888
615.292.0777 (fax)

Emma helps organizations everywhere communicate & market in style.
Visit us online at http://www.myemma.com