From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43) id 1L0hGD-0002e8-1b for qemu-devel@nongnu.org; Thu, 13 Nov 2008 13:49:45 -0500 Received: from exim by lists.gnu.org with spam-scanned (Exim 4.43) id 1L0hGB-0002da-Cu for qemu-devel@nongnu.org; Thu, 13 Nov 2008 13:49:44 -0500 Received: from [199.232.76.173] (port=48680 helo=monty-python.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1L0hGB-0002dX-An for qemu-devel@nongnu.org; Thu, 13 Nov 2008 13:49:43 -0500 Received: from qw-out-1920.google.com ([74.125.92.145]:20946) by monty-python.gnu.org with esmtp (Exim 4.60) (envelope-from ) id 1L0hGB-0007ed-0U for qemu-devel@nongnu.org; Thu, 13 Nov 2008 13:49:43 -0500 Received: by qw-out-1920.google.com with SMTP id 5so798977qwc.4 for ; Thu, 13 Nov 2008 10:49:42 -0800 (PST) Message-ID: <491C76C2.9060409@codemonkey.ws> Date: Thu, 13 Nov 2008 12:49:38 -0600 From: Anthony Liguori MIME-Version: 1.0 Subject: Re: [Qemu-devel] [PATCH] add file: migration support (r2) References: <200811131844.11491.paul@codesourcery.com> In-Reply-To: <200811131844.11491.paul@codesourcery.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Reply-To: qemu-devel@nongnu.org List-Id: qemu-devel.nongnu.org List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: Charles Duffy Paul Brook wrote: > On Thursday 13 November 2008, Charles Duffy wrote: > >> This patch adds support for migration to and from file: targets, moves >> common helpers between exec: and file: use cases from migration-exec.c >> to migration.c, and adds a qemu_fdopen() helper parallel to qemu_fopen(). >> >> Compared to the previous version submitted, it removes some cruft which >> was accidentally included, and documents the limitations of O_NONBLOCK >> on Linux with local filesystem access. >> > > I don't see any documentation. > > The documentation should make it clear that this is not intended as a > substitute for savvm/loadvm. A "live" migration doesn't make sense if you're > writing to an actual file (rather than a FIFO) as you'll end up with > unecessarily large files. > Forget the large file, it's not "live" which makes it not useful compared with savevm/loadvm. Using rate limiting as a crutch to avoid this is wrong because it requires the user to know how much bandwidth QEMU can use before blocking. To do live migration to a file, you need to tie into something like posix-aio. You could actually use a bdrv_raw and expand it like we do for a qcow file although that's sort of an abuse of the block driver API. Regards, Anthony Liguori > Paul > > >