From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:39850) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1aMS21-0001EV-4L for qemu-devel@nongnu.org; Thu, 21 Jan 2016 22:05:02 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1aMS1w-0008Cn-9m for qemu-devel@nongnu.org; Thu, 21 Jan 2016 22:05:01 -0500 Received: from szxga01-in.huawei.com ([58.251.152.64]:18015) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1aMS1v-0008Bm-Gq for qemu-devel@nongnu.org; Thu, 21 Jan 2016 22:04:56 -0500 References: <20160121190843.GH2446@work-vm> From: Hailiang Zhang Message-ID: <56A19C32.6000606@huawei.com> Date: Fri, 22 Jan 2016 11:04:18 +0800 MIME-Version: 1.0 In-Reply-To: <20160121190843.GH2446@work-vm> Content-Type: text/plain; charset="windows-1252"; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] COLO hybrid mode and other changes List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: "Dr. David Alan Gilbert" , zhangchen.fnst@cn.fujitsu.com, lizhijian@cn.fujitsu.com, xiecl.fnst@cn.fujitsu.com, wency@cn.fujitsu.com Cc: peter.huangpeng@huawei.com, qemu-devel@nongnu.org Nice work :) On 2016/1/22 3:08, Dr. David Alan Gilbert wrote: > Hi, > I've got a world with a few COLO changes in that you might > like to look at; they're all experimental, but it seems > to be working for me. > > The world is at: > https://github.com/orbitfp7/qemu/commits/orbit-wp4-colo-jan16 > > It's based off: > The December/2.4 colo framework world with periodic checkpoints > + The december user space colo-proxy > + The integration code that was in the colo-proxy git repo > (those needed some updates to go into the 2.4-periodic world) > > And then I've added: > - An updated version of my hybrid mode code that switches between > COLO and periodic checkpoints depending on the frequency of > miscompares > - A minimum checkpoint time, so that on fast miscompares > we do get some runtime (as in the earlier colo series) > - I'm using a condition-variable to flag the miscompares from > the colo-proxy to the colo thread; this seems cleaner and it > avoids having to do short waits; it might be a bit tricky > to make portable. > - Add back some statistcs in 'info migrate' > - An HMP command equivalent to x-blockdev-change. > - RDMA transport for COLO > > I don't intend to post any of these changes in full to the list > until your main COLO code is in; although if any of my changes > are useful to you, then feel free to include them in sets you post. > Thanks for your effort. I will look at them later. Hailiang > Dave > -- > Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK > > . >