Message-ID: <4CC04F89.6020503@ca.afilias.info>
Date: Thu, 21 Oct 2010 10:34:49 -0400
From: Brad Nicholson <bnichols@ca.afilias.info>
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US;
	rv:1.9.2.11) Gecko/20101007 Lightning/1.0b2 Thunderbird/3.1.5
MIME-Version: 1.0
To: pgsql-performance@postgresql.org
Subject: Re: New wiki page on write reliability
References: <4CC04945.7030100@2ndquadrant.com>
In-Reply-To: <4CC04945.7030100@2ndquadrant.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 10-10-21 10:08 AM, Greg Smith wrote:
> Now that some of my recent writing has gone from NDA protected to 
> public sample, I've added a new page to the PostgreSQL wiki that 
> provides a good starting set of resources to learn about an ever 
> popular topic here, how write cache problems can lead to database 
> corruption:  http://wiki.postgresql.org/wiki/Reliable_Writes
>
> Bruce also has a presentation he's been working on that adds pictures 
> showing the flow of data through the various cache levels, to help 
> people visualize the whole thing, that should get added into there 
> once he's finished tweaking it.
>
> I'd like to get some feedback from the members of this list about 
> what's still missing after this expanded data dump.  Ultimately I'd 
> like to get this page to be an authoritative enough resource that the 
> "Reliability" section of the official documentation could point back 
> to this as a recommendation for additional information.  So much of 
> this material requires singling out specific vendors and staying up to 
> date with hardware changes, both things that the official docs are not 
> a good place for.
>

Looks like a good start.

I think a warning turning fsync off, the dangers of async_commit, and 
the potential problems with disabling full_page_writes might be worth 
mentioning on this page, unless you want to leave that buried in the 
attached references.

-- 
Brad Nicholson  416-673-4106
Database Administrator, Afilias Canada Corp.