From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alasdair G Kergon Subject: Re: [PATCH 2/2] mirror throttling Date: Fri, 15 Feb 2013 00:53:21 +0000 Message-ID: <20130215005321.GK24130@agk-dp.fab.redhat.com> References: <1357552933-26126-1-git-send-email-gzhao@suse.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org To: Mikulas Patocka Cc: Guangliang Zhao , linux-kernel@vger.kernel.org, dm-devel@redhat.com, lucienchao@gmail.com, "Alasdair G. Kergon" List-Id: dm-devel.ids On Wed, Jan 09, 2013 at 12:44:38AM -0500, Mikulas Patocka wrote: > We maintain a history of kcopyd usage in variables io_period and > total_period. The actual kcopyd activity is "(100 * io_period / > total_period)" percent of time. If we exceed user-defined percentage > threshold, we sleep. Well, I'm going to take this pair of patches for now. Some people do need this throttling today and this seems to me to be a decent and simple way to give them a lever to deal with the problem. I'm not proposing we expose this through LVM or other userspace tools at this stage: people who need it should tune it directly through sysfs. If need be, we can revisit this in future either by refining the algorithm or making it adjustable per-device rather than per-module. (Or by re-vamping kcopyd itself...) The current version is: http://people.redhat.com/agk/patches/linux/editing/dm-kcopyd-introduce-configurable-throttling.patch Alasdair