Re: [GENERAL] Performance while loading data and indexing - Mailing list pgsql-hackers

From Doug cNaught
Subject Re: [GENERAL] Performance while loading data and indexing
Date
Msg-id m34rcck6x7.fsf@varsoon.wireboard.com
Whole thread Raw
In response to Performance while loading data and indexing  ("Shridhar Daithankar" <shridhar_daithankar@persistent.co.in>)
List pgsql-hackers
Greg Copeland <greg@CopelandConsulting.Net> writes:

> I'm not sure about reiserfs or ext3 but with XFS, you can create your
> log on another disk.  Also worth noting is that you can also configure
> the size and number of log buffers.  There are also some other
> performance type enhancements you can fiddle with if you don't mind
> risking time stamp consistency in the event of a crash.  If your setup
> allows for it, you might want to consider using XFS in this
> configuration.

You can definitely put the ext3 log on a different disk with 2.4
kernels.

Also, if you put the WAL logs on a different disk from the main
database, and mount that partition with 'data=writeback' (ie
metadata-only journaling) ext3 should be pretty fast, since WAL files
are preallocated and there will therefore be almost no metadata
updates.

You should be able to mount the main database with "data=ordered" (the
default) for good performance and reasonable safety.

I think putting WAL on its own disk(s) is one of the keys here.

-Doug

pgsql-hackers by date:

Previous
From: Masaru Sugawara
Date:
Subject: About connectby() again
Next
From: Masaru Sugawara
Date:
Subject: Re: About connectby() again