From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from cuda.sgi.com (cuda2.sgi.com [192.48.176.25]) by oss.sgi.com (8.14.3/8.14.3/SuSE Linux 0.8) with ESMTP id o2G1CVlI105399 for ; Mon, 15 Mar 2010 20:12:31 -0500 Received: from mail.internode.on.net (localhost [127.0.0.1]) by cuda.sgi.com (Spam Firewall) with ESMTP id 314F624C945 for ; Mon, 15 Mar 2010 18:14:05 -0700 (PDT) Received: from mail.internode.on.net (bld-mail15.adl6.internode.on.net [150.101.137.100]) by cuda.sgi.com with ESMTP id ipFOEeNKEPHZBHjx for ; Mon, 15 Mar 2010 18:14:05 -0700 (PDT) Date: Tue, 16 Mar 2010 12:14:03 +1100 From: Dave Chinner Subject: Re: [RFC] Delayed logging Message-ID: <20100316011403.GB12369@dastard> References: <20100315043000.GK4732@dastard> <20100315160801.GB10406@infradead.org> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <20100315160801.GB10406@infradead.org> List-Id: XFS Filesystem from SGI List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Sender: xfs-bounces@oss.sgi.com Errors-To: xfs-bounces@oss.sgi.com To: Christoph Hellwig Cc: xfs@oss.sgi.com On Mon, Mar 15, 2010 at 12:08:01PM -0400, Christoph Hellwig wrote: > Some comments on the patches: > > - "xfs: cluster fsync transaction" seems like something that can stand > on it's own and go into the tree now. Despite the comments it > currently only clusters in fsync and not write_inode which might > lead to higher benefits, btw. As I said, I'll have a bit more of a think about this one... > - the busy extent tracking might be worth to be reordered before the > delayed logging series. In itself it might also want some reordering > as there's a lot churn in the patches there. Making > "XFS: Simplify transaction busy extent tracking" first in that > subseries might help quite a bit to reduce that churn Agreed. They were ordered this way because I only added the extent tracking after doing the delayed logging and tracking down one of the sources of log forces that was limiting performance was overflowing the per-ag busy extent array. I'll rework it.... > - The actual CIL implementation seems to be split into too small > patches IMHO. E.g. "xfs: extend the log item to support delayed > logging" and "xfs: Introduce the Committed Item List" are two > sides of the same coin and splitting it might not make too much > sense. True, it might be a bit fine grained. Really, what I wanted to do is split the changes that affected the non-delayed logging from those that are only executed when delayed logging is active. That way I could confirm that the non-delayed path was still operating correctly before adding all the delayed logging code that used it. I can combine them together again if you want. Cheers, Dave. -- Dave Chinner david@fromorbit.com _______________________________________________ xfs mailing list xfs@oss.sgi.com http://oss.sgi.com/mailman/listinfo/xfs