From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([140.186.70.92]:45982)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <lcapitulino@redhat.com>) id 1QZMnB-0000IJ-66
	for qemu-devel@nongnu.org; Wed, 22 Jun 2011 08:44:26 -0400
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <lcapitulino@redhat.com>) id 1QZMn9-0005vJ-N6
	for qemu-devel@nongnu.org; Wed, 22 Jun 2011 08:44:24 -0400
Received: from mx1.redhat.com ([209.132.183.28]:56714)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <lcapitulino@redhat.com>) id 1QZMn9-0005v8-5J
	for qemu-devel@nongnu.org; Wed, 22 Jun 2011 08:44:23 -0400
Date: Wed, 22 Jun 2011 09:44:19 -0300
From: Luiz Capitulino <lcapitulino@redhat.com>
Message-ID: <20110622094419.2ee2d06a@doriath>
In-Reply-To: <4E01AE27.1040407@redhat.com>
References: <20110621142926.7bc4df99@doriath>
	<4E01AE27.1040407@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Subject: Re: [Qemu-devel] block: regression: savevm/delvm too slow
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Kevin Wolf <kwolf@redhat.com>
Cc: Stefan Hajnoczi <stefanha@gmail.com>, qemu-devel <qemu-devel@nongnu.org>

On Wed, 22 Jun 2011 10:56:07 +0200
Kevin Wolf <kwolf@redhat.com> wrote:

> Am 21.06.2011 19:29, schrieb Luiz Capitulino:
> > I'm getting the following times when doing a savevm and delvm in current
> > HEAD eb47d7c5d (time in minutes, each time corresponds to a savevm/delvm run):
> > 
> >  savevm: 5:28m, 11:00m, 11:10m
> >  delvm: 4:30m, 4:40m, > 15m
> > 
> > Now, trying with qemu 0.13.0 I get:
> > 
> >  savevm: < 1:00m, 4:00m, 4:34m
> >  delvm: few seconds for all cases
> > 
> > Yes, you read it correctly, I tried with 0.13.0 because 0.14.0 also has the
> > bug. This is the pattern I see when I run strace against HEAD while running
> > the savevm command:
> > 
> >  pwrite(7, "\0\1\0\1\0\1\0\1\0\2\0\2\0\2\0\2\0\2\0\2\0\2\0\2\0\2\0\2\0\2\0\2"..., 65536, 196608) = 65536
> >  fdatasync(7)                            = 0
> > 
> > Ie. a fdatasync() follows every single pwrite(). Something similar also happens
> > with delvm. I don't see this pattern with 0.13.0.
> > 
> > The good news is that I've tracked it down and Mr. git bisect says that:
> > 
> > 29c1a7301af752de6721e031d31faa48887204bd is the first bad commit
> > commit 29c1a7301af752de6721e031d31faa48887204bd
> > Author: Kevin Wolf <kwolf@redhat.com>
> > Date:   Mon Jan 10 17:17:28 2011 +0100
> > 
> >     qcow2: Use QcowCache
> >     
> >     Use the new functions of qcow2-cache.c for everything that works on refcount
> >     block and L2 tables.
> >     
> >     Signed-off-by: Kevin Wolf <kwolf@redhat.com>
> > 
> > :040000 040000 83e364185d37845bb27f1dccd1249d14cc7a9a1e 0c91964a52b5869333d4fb2cb0fa83104151359e M	block
> 
> Let me guess... You're using cache=writethrough?

I'm using default options. To do the tests I started qemu with:

 # qemu -hda disks/test2.img -enable-kvm -m 1G -monitor stdio

> 
> Previously, qcow2_update_snapshot_refcount() had implemented its own
> kind of writeback cache that would be used even with cache=writethrough.
> Now we're using the generic Qcow2Cache, which implements a write-through
> behaviour for cache=writethrough.
> 
> We could fix this by temporarily switching the cache to writeback mode.
> 
> Kevin
>