qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH RFC] block: Tolerate existing writers on read on


From: Kevin Wolf
Subject: Re: [Qemu-devel] [PATCH RFC] block: Tolerate existing writers on read only BdrvChild
Date: Thu, 2 Mar 2017 15:23:18 +0100
User-agent: Mutt/1.5.21 (2010-09-15)

Am 02.03.2017 um 12:21 hat Fam Zheng geschrieben:
> On Wed, 03/01 17:22, Kevin Wolf wrote:
> > Am 01.03.2017 um 17:10 hat Fam Zheng geschrieben:
> > > On Wed, 03/01 16:16, Kevin Wolf wrote:
> > > > > I'm not sure about this because: 1) this is intrusive from a user 
> > > > > PoV, many
> > > > > scripts and upper layer tools will stop working;
> > > > 
> > > > Are you sure? I don't expect user scripts or even proper management
> > > > tools to use qemu-io on running VMs. I do expect that some users are
> > > > using 'convert -s' with running VMs despite our recommendation against
> > > > it.
> > > > 
> > > > If they are aware that they're doing something that works only in the
> > > > right circumstances, then breaking it isn't nice. But my gut feeling is
> > > > that most of them don't know about the implications of accessing a live
> > > > image. In this case breaking their use case and making them think about
> > > > whether they want to add something like a --force option sounds like a
> > > > good thing because they aren't caught by surprise when things go wrong
> > > > eventually.
> > > 
> > > Yes, the use case is poor for qcow2, and your points stand there. But 
> > > image
> > > locking will be at the posix level, which has a wider range of users. I 
> > > cannot
> > > draw the same conclusion on raw images as easily.
> > 
> > Well, with raw, I'm even less concerned about breaking the commands
> > related to internal snapshots. :-)
> 
> Yes, I'm agree with a --force there. For qemu-img map and qemu-io, personally 
> I
> think it's better to keep the default working. qemu-io is a expert mode tool,
> whoever using it at all should already know what he's doing, --force doesn't 
> add
> much protection for the innocent.

Being an expert doesn't protect you from stupid mistakes like forgetting
that a VM is still running. --force doesn't prevent you from performing
your evil action, but it prevents accidents where you didn't even intend
to be evil for a change.

I think qemu-io and qemu-img map are tools that only human users should
be using while a VM is running, so breaking command line syntax
compatibility by requiring a --force option wouldn't hurt too much.

Kevin



reply via email to

[Prev in Thread] Current Thread [Next in Thread]