From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:42632) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1elaaR-0001ui-7z for qemu-devel@nongnu.org; Tue, 13 Feb 2018 08:25:35 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1elaaP-0006cn-Rm for qemu-devel@nongnu.org; Tue, 13 Feb 2018 08:25:31 -0500 Date: Tue, 13 Feb 2018 13:25:11 +0000 From: Daniel =?utf-8?B?UC4gQmVycmFuZ8Op?= Message-ID: <20180213132511.GL573@redhat.com> Reply-To: Daniel =?utf-8?B?UC4gQmVycmFuZ8Op?= References: <20180107122336.29333-1-richiejp@f-m.fm> <5cf19623-72ac-fb8b-2054-a60d42419ec6@redhat.com> <20180111130427.GG8326@redhat.com> <20180213105024.GC5083@localhost.localdomain> <20180213114355.GC2378@work-vm> <20180213115102.GF573@redhat.com> <20180213132000.GG5083@localhost.localdomain> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <20180213132000.GG5083@localhost.localdomain> Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [Qemu-block] [PATCH 1/2] Add save-snapshot, load-snapshot and delete-snapshot to QAPI List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Kevin Wolf Cc: "Dr. David Alan Gilbert" , Max Reitz , Richard Palethorpe , Qemu-block , quintela@redhat.com, armbru@redhat.com, qemu-devel@nongnu.org, rpalethorpe@suse.com On Tue, Feb 13, 2018 at 02:20:00PM +0100, Kevin Wolf wrote: > Am 13.02.2018 um 12:51 hat Daniel P. Berrang=C3=A9 geschrieben: > > On Tue, Feb 13, 2018 at 11:43:55AM +0000, Dr. David Alan Gilbert wrot= e: > > > * Kevin Wolf (kwolf@redhat.com) wrote: > > > > Am 11.01.2018 um 14:04 hat Daniel P. Berrange geschrieben: > > > > > On Thu, Jan 11, 2018 at 01:46:38PM +0100, Max Reitz wrote: > > > > > > On 2018-01-08 14:52, Eric Blake wrote: > > > > > > > On 01/07/2018 06:23 AM, Richard Palethorpe wrote: > > > > > > >> Add QAPI wrapper functions for the existing snapshot funct= ionality. These > > > > > > >> functions behave the same way as the HMP savevm, loadvm an= d delvm > > > > > > >> commands. This will allow applications, such as OpenQA, to= programmatically > > > > > > >> revert the VM to a previous state with no dependence on HM= P or qemu-img. > > > > > > >=20 > > > > > > > That's already possible; libvirt uses QMP's human-monitor-c= ommand to > > > > > > > access these HMP commands programmatically. > > > > > > >=20 > > > > > > > We've had discussions in the past about what it would take = to have > > > > > > > specific QMP commands for these operations; the biggest pro= blem is that > > > > > > > these commands promote the use of internal snapshots, and t= here are > > > > > > > enough performance and other issues with internal snapshots= that we are > > > > > > > not yet ready to commit to a long-term interface for making= their use > > > > > > > easier. At this point, our recommendation is to prefer ext= ernal snapshots. > > > > > >=20 > > > > > > We already have QMP commands for internal snapshots, though. = Isn't the > > > > > > biggest issue that savevm takes too much time to be a synchro= nous QMP > > > > > > command? > > > > >=20 > > > > > Ultimately savevm/loadvm are using much of the migration code i= nternally, > > > > > but are not exposed as URI schemes. Could we perhaps take advan= tage of > > > > > the internal common layer and define a migration URI scheme > > > > >=20 > > > > > snapshot: > > > > >=20 > > > > > where '' is the name of the internal snapshot in the qcow= 2 file. > > > >=20 > > > > Let's include a node-name there, please. QEMU automagically decid= ing > > > > where to store the VM state is one of the major problems of the H= MP > > > > interface. > > > >=20 > > > > And while we're at it, we can make it more future-proof by allowi= ng to > > > > specify arbitrary options: > > > >=20 > > > > snapshot:node=3D,name=3D > > > >=20 > > > > That would allow us to add something like compressed=3Don|off lat= er. > > > > Actually, compressed VM state sounds pretty nice. Why don't we ha= ve this > > > > yet? The qcow2 format already provides everything you need for it= . > > > >=20 > > > > > Then you could just use the regular migrate QMP commands for lo= ading > > > > > and saving snapshots. > > > >=20 > > > > Yes, you could. I think for a proper implementation you would wan= t to do > > > > better, though. Live migration provides just a stream, but that's= not > > > > really well suited for snapshots. When a RAM page is dirtied, you= just > > > > want to overwrite the old version of it in a snapshot, you don't = want to > > > > waste space by keeping both the old and the current version of th= e page > > > > content in the file. > > >=20 > > > The current snapshots are run with the CPU paused aren't they? The= y > > > share exactly the same RAM saving code as migration. > >=20 > > The commands block the monitor too, so I always assumed they were non= -live >=20 > Yes, they are non-live, so RAM simply doesn't change and we that's the > reason why we currently don't get duplicate pages in the snapshot. >=20 > But as I understand it, the whole point of migrate snapshot:... would b= e > to make it live, so the potential duplication is something that we need > to consider for it. Yeah, true. Re-writing ram pages in-place could be doable if I added a feature flag QIO_CHANNEL_FEATURE_SEEKABLE, so the migration code can tell if it will just append forever, or can seek back to rewrite. Regards, Daniel --=20 |: https://berrange.com -o- https://www.flickr.com/photos/dberran= ge :| |: https://libvirt.org -o- https://fstop138.berrange.c= om :| |: https://entangle-photo.org -o- https://www.instagram.com/dberran= ge :|