On Tue, Mar 25, 2014 at 05:45:12PM +0000, Anoob Soman wrote:
> 
> On 25/03/14 16:07, Ben Pfaff wrote:
> >On Tue, Mar 25, 2014 at 03:44:44PM +0000, Anoob Soman wrote:
> >>On 25/03/14 15:29, Gurucharan Shetty wrote:
> >>>On Tue, Mar 25, 2014 at 5:32 AM, Anoob Soman <anoob.so...@citrix.com> 
> >>>wrote:
> >>>>Hi,
> >>>>
> >>>>Does rhel openvswitch deployments required fake bond interfaces to be
> >>>>created. rhel/etc_sysconfig_network-scripts_ifup-ovs uses --fake-iface for
> >>>>creating bonds (add-bond). I have sent a patch "[PATCH] rhel: Remove
> >>>>creation of fake bond interface" to remove --fake-iface and add 
> >>>>--may-exist
> >>>>instead. Ignore my patch if fake bond interfaces are required.
> >>>The "--fake-iface" as an option in the RHEL ifup script came with its
> >>>first version. Does its presence hurt any of your use cases?
> >>No, it doesn't. But, AFAIK --fake-iface was add to support older
> >>version of XenServer requiring a bond interface for some legacy
> >>reasons, newer version of XenSever doesn't use --fake-iface. This
> >>means --fake-iface is not used by anyone (and hoping rhel doesn't
> >>need it either), and can be removed from code.
> >I think OVS still supports XenServer old enough that --fake-iface is
> >needed.
> >
> Yes. XenSever 6.0.2 still uses --fake-iface, but openvswitch in
> XS6.0.2 is taken from vlan-maint branch. XS6.1 onwards doesn't use
> --fake-iface. Removal of --fake-iface from the code would happen in
> master and might end up in OVS 2.x branch and this should not affect
> vlan-maint or any other LTS branch, unless we have intention of back
> porting this change to vlan-maint.

I understand that Citrix ships a vlan-maint release with 6.0.2, but I
also know that some users at least used to install newer OVS on top of
it.  If that is still an important use case, then I'd like to continue
supporting it.  I'm going to ask around and see if anyone around here
knows of that still going on.

> My intention, of removing --fake-iface from
> rhel/etc_sysconfig_ifup-ovs is, if rhel doesn't want a fake bond
> interface why should it create one. --fake-iface was created for
> compatibility reasons and should only be used if it is required to.

I agree that removing --fake-iface from rhel/* is a separate issue,
that doesn't have to be tied to removing it globally (which really I'd
like to do, I'm just worried about break compatibility).
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Reply via email to