Re: [Qemu-devel] [PATCH v12 2/3] quorum: implement bdrv_add_child() and bdrv_del_child()

["Dr. David Alan Gilbert" <dgilbert@redhat.com>]

* Alberto Garcia (berto@igalia.com) wrote:
> On Wed 30 Mar 2016 05:07:15 PM CEST, Max Reitz wrote:
> >> I also have another (not directly related) question: why not simply
> >> use the node name when removing children? I understood that the idea
> >> was that it's possible to have the same node attached twice to the
> >> same Quorum, but can you actually do that? And what's the use case?
> >
> > What I like about using the child role name is that it automatically
> > prevents you from specifying a node that is not a child of the given
> > parent.
> 
> Right, but checking if a node is not a child and returning an error is
> very simple. And it doesn't require the user to keep track of the node
> name *and* the child role name.
> 
> Unless I'm forgetting something this would be the first time we expose
> the child role name in the API, that's why I'm wondering if it's
> something worth doing.
> 
> > Which makes me notice that it might be a good idea to require the user
> > to specify the child's role when adding a new child. In this version
> > of this series (where only quorum is supported), the children are just
> > inserted in numerical order (first free slot is taken first), but
> > maybe the user wants to insert them in a different order.
> 
> For the Quorum case it totally makes sense to let the user choose the
> position of the new child.
> 
> But for creating a Quorum array in the first place we don't require
> that, the order is the one that the user provides, and the user does not
> need to know about the child role names at that point.

Actually in my setup I only add the local block device using the normal command
line, and use drive_add even at startup.  It solves a lot of the ordering
problems with getting COLO booted.

Dave

> 
> > And the "filling up quorum's children" topic then makes me notice that
> > (x-)blockdev-change should probably be transaction-able (so you can
> > restructure the whole BDS graph in a single transaction), but that's
> > something we can care about later on.
> 
> I agree.
> 
> Berto
--
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK