From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:55525) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1d8r1r-0006XO-5o for qemu-devel@nongnu.org; Thu, 11 May 2017 12:33:27 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1d8r1o-0006t5-Hk for qemu-devel@nongnu.org; Thu, 11 May 2017 12:33:27 -0400 Received: from mx1.redhat.com ([209.132.183.28]:43180) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1d8r1o-0006st-Bz for qemu-devel@nongnu.org; Thu, 11 May 2017 12:33:24 -0400 From: Juan Quintela In-Reply-To: <590AF10B.9020207@huawei.com> (Hailiang Zhang's message of "Thu, 4 May 2017 17:14:51 +0800") References: <20170425103049.4073-1-quintela@redhat.com> <20170425103049.4073-3-quintela@redhat.com> <20170428165522.GG3276@work-vm> <87vapht37i.fsf@secure.mitica> <590AF10B.9020207@huawei.com> Reply-To: quintela@redhat.com Date: Thu, 11 May 2017 18:33:20 +0200 Message-ID: <87efvvgxpr.fsf@secure.mitica> MIME-Version: 1.0 Content-Type: text/plain Subject: Re: [Qemu-devel] [PATCH 2/3] migration: Remove use of old MigrationParams List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Hailiang Zhang Cc: "Dr. David Alan Gilbert" , qemu-devel@nongnu.org, lvivier@redhat.com, peterx@redhat.com Hailiang Zhang wrote: > Hi, > >>> Hmm you don't seem to have replaced this with anything. >>> I think that's a behavioural change; the trick COLO did (I'm not sure if this >>> is still the way it works) is that they initiate the first migration >>> with block migration enabled so that the two hosts (with non-shared storage) >>> get sync'd storage, and then at the completion of that first migration >>> they then switch into the checkpointing mode where they're only >>> doing updates - that's why it gets switched off at this point >>> prior to the 1st checkpoint. >> >> Weird, really. >> >> I did't catch that. >> >> Will investigate. > > Yes, Dave is right, for non-shared disk, we need to enable block > migration for first cycle, > to sync the disks of two sides. After that, qemu will go into COLO > state which we need to > disable block migration. v2 posted. My understanding is that it maintains the sematic, please test/comment. Thanks, Juan.