From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:56182) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VLGkZ-0006Wr-Bz for qemu-devel@nongnu.org; Sun, 15 Sep 2013 14:08:53 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1VLGkT-0006TK-Bf for qemu-devel@nongnu.org; Sun, 15 Sep 2013 14:08:47 -0400 Received: from nodalink.pck.nerim.net ([62.212.105.220]:35595 helo=paradis.irqsave.net) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VLGkS-0006SW-R0 for qemu-devel@nongnu.org; Sun, 15 Sep 2013 14:08:41 -0400 Date: Sun, 15 Sep 2013 20:10:21 +0200 From: =?iso-8859-1?Q?Beno=EEt?= Canet Message-ID: <20130915181021.GA5868@irqsave.net> References: <20130903162449.GF5285@irqsave.net> <20130906075606.GD4814@T430s.nay.redhat.com> <20130906084513.GE2588@dhcp-200-207.str.redhat.com> <20130906091820.GA24154@T430s.nay.redhat.com> <20130906095538.GF2588@dhcp-200-207.str.redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline In-Reply-To: <20130906095538.GF2588@dhcp-200-207.str.redhat.com> Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] Block Filters List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Kevin Wolf Cc: =?iso-8859-1?Q?Beno=EEt?= Canet , Fam Zheng , jcody@redhat.com, armbru@redhat.com, qemu-devel@nongnu.org, stefanha@redhat.com, pbonzini@redhat.com, mreitz@redhat.com Le Friday 06 Sep 2013 =E0 11:55:38 (+0200), Kevin Wolf a =E9crit : > Am 06.09.2013 um 11:18 hat Fam Zheng geschrieben: > > On Fri, 09/06 10:45, Kevin Wolf wrote: > > > Am 06.09.2013 um 09:56 hat Fam Zheng geschrieben: > > > > Since BlockDriver.bdrv_snapshot_create() is an optional operation= , blockdev.c > > > > can navigate down the tree from top node, until hitting some laye= r where the op > > > > is implemented (the QCow2 bs), so we get rid of this top_node_bel= ow_filter > > > > pointer. > > >=20 > > > Is it even inherent to a block driver (like a filter), if a snapsho= t is > > > to be taken at its level? Or is it rather a policy decision that sh= ould > > > be made by the user? > > >=20 > > OK, getting the point that user should have full flexibility and fine= operation > > granularity. It also stands against block_backend->top_node_below_fil= ter. Do we > > really have the assumption that all the filters are on top of the tre= e and linear? > > Shouldn't this be possible? > >=20 > > Block Backend > > | > > | > > Quodrum BDS > > / | \ > > throttle filter | \ > > / | \ > > qcow2 qcow2 qcow2 > >=20 > > So we throttle only a particular image, not the whole device. But thi= s will > > make a top_node_below_filter pointer impossible. >=20 > I was assuming that Beno=EEt's model works for the special case of > snapshotting in one predefined way, but this is actually a very good > example of why it doesn't. >=20 > The approach relies on snapshotting siblings together, and in this case > the siblings would be throttle/qcow2/qcow2, while throttle is still a f= ilter. This > would mean that either throttle needs to be top_node_below_filter and > throttling doesn't stay on top, or the left qcow2 is > top_node_below_filter and the other Quorum images aren't snapshotted. >=20 > > > In our example, the quorum driver, it's not at all clear to me that= you > > > want to snapshot all children. In order to roll back to a previous > > > state, one snapshot is enough, you don't need multiple copies of th= e > > > same one. Perhaps you want two so that we can still compare them fo= r > > > verification. Or all of them because you can afford the disk space = and > > > want ultimate safety. I don't think qemu can know which one is true= . > > >=20 > > Only if quorum ever knows about and operates on snapshots, it should = be > > considered specifically, but no. So we need to achieve this in the ge= neral > > design: allow user to take snapshot, or set throttle limits on partic= ular > > BDSes, as above graph. > >=20 > > > In the same way, in a typical case you may want to keep I/O throttl= ing > > > for the whole drive, including the new snapshot. But what if the > > > throttling was used in order to not overload the network where the = image > > > is stored, and you're now doing a local snapshot, to which you want= to > > > stream the image? The I/O throttling should apply only to the backi= ng > > > file, not the new snapshot. > > >=20 > > Yes, and OTOH, throttling really suits to be a filter only if it can = be a non > > top one, otherwise it's no better than what we have now. >=20 > Well, it would be a cleaner architecture in any case, but having it in > the middle of the stack feels useful indeed, so we should support it. >=20 > > > So perhaps what we really need is a more flexible snapshot/BDS tree > > > manipulation command that describes in detail which structure you w= ant > > > to have in the end. >=20 > Designing the corresponding QMP command is the hard part, I guess. During my vacation I though about the fact that JSON is pretty good to bu= ild a tree. QMP, HMP and the command line could take a "block-tree" argument which wo= uld look like the following. block-tree =3D { 'quorum': [ { 'throttle' : { 'qcow2' : { 'filename': "img1.qcow2= " } 'snapshotable': true, }, 'throttle-iops' : 150, 'throttle-iops-max' : 1000, }, { 'qcow2' : { 'filename': "img2.qcow2" }, 'snapshotable': true, }, { 'qcow2' : { 'filename': "img3.qcow2" } 'snapshotable': false, } ] }; This would be passed to QEMU in a compact form without carriage return an= d spaces. The block layer would convert this to C structs like the QMP code would d= o for a QMP command and the bs tree would be recursively build from top to bottom= by the Block Backend and each Block driver in the path using the C structs. Each level would instanciate the lower level until a raw or protocol driv= er is reached. What about this ? Best regards Beno=EEt