Re: using a lot of maintenance_work_mem

Devrim G�ND�Z wrote:
> On Wed, 2011-02-16 at 23:24 +0200, Peter Eisentraut wrote:
> > 
> > > But before expending time on that, I'd want to see some evidence
> > that
> > > it's actually helpful for production situations.  I'm a bit dubious
> > > that you're going to gain much here.
> > 
> > If you want to build an index on a 500GB table and you have 1TB RAM,
> > then being able to use >>1GB maintenance_work_mem can only be good,
> > no? 
> 
> That would also probably speed up Slony (or similar) replication engines
> in initial replication phase. I know that I had to wait a lot while
> creating big indexes on a machine which had enough ram.

Well, I figure it will be hard to allow larger maximums, but can we make
the GUC variable maximums be more realistic?  Right now it is
MAX_KILOBYTES (INT_MAX).

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com
 + It's impossible for everything to be true. +