Horton

Re: 7 hrs for a pg_restore?

Jeff Davis <pgsql@j-davis.com>

View thread
From: Jeff Davis <pgsql@j-davis.com>
To: Dave Cramer <pg@fastcrypt.com>
Cc: Douglas J Hunley <doug@hunley.homeip.net>, "Joshua D. Drake" <jd@commandprompt.com>, pgsql-performance@postgresql.org
Date: 2008-02-19T19:51:19Z
Lists: pgsql-performance 
On Tue, 2008-02-19 at 14:28 -0500, Dave Cramer wrote:
> shared buffers is *way* too small as is effective cache
> set them to 2G/6G respectively.

They are way too small, but I don't think that explains the index
creation time.

Effective_cache_size is only used by the planner, and this problem is
not caused by a poorly chosen plan.

It's important to set shared_buffers higher as well, but he has so much
RAM compared with his dataset that he's certainly not going to disk. I
don't think this explains it either. 

I think it's just the result of building a lot of indexes on localized
text using only one core at a time.

Regards,
	Jeff Davis