On Fri, Jun 3, 2016 at 2:44 PM, Mike Pontillo <email address hidden>
wrote:
> Thanks for the review; some replies below.
>
> Diff comments:
>
> > === modified file 'curtin/net/__init__.py'
> > --- curtin/net/__init__.py 2016-06-03 00:35:50 +0000
> > +++ curtin/net/__init__.py 2016-06-03 00:35:50 +0000
> > @@ -331,7 +331,12 @@
> >
> >
> > # TODO: switch to valid_map for attrs
> > -def iface_add_attrs(iface):
> > +def iface_add_attrs(iface, index):
> > + # If the index is non-zero, this is an alias interface. Alias
> interfaces
> > + # represent additional interface addresses, and cannot have
> additional
> > + # attributes. So if this is an alias, stop right here.
>
> Well, ifupdown doesn't do much in the way of model validation. ;-) An
> alias is really just a (deprecated, I think) way to add more addresses to
> an interface. (that is to say, ifupdown may well allow them, I can't think
> of any cases where that isn't arguably a bug in ifupdown.)
>
> https://bugs.launchpad.net/maas/+bug/1588547/comments/5
>
>
It's probably worth supplying what the preferred eni should look like;
I'm certainly no bonding expert but attempted to apply formatting as
suggested
by the ifenslave documentation.
On Fri, Jun 3, 2016 at 2:44 PM, Mike Pontillo <email address hidden>
wrote:
> Thanks for the review; some replies below. net/__init_ _.py' net/__init_ _.py 2016-06-03 00:35:50 +0000 net/__init_ _.py 2016-06-03 00:35:50 +0000 attrs(iface) : attrs(iface, index): /bugs.launchpad .net/maas/ +bug/1588547/ comments/ 5
>
> Diff comments:
>
> > === modified file 'curtin/
> > --- curtin/
> > +++ curtin/
> > @@ -331,7 +331,12 @@
> >
> >
> > # TODO: switch to valid_map for attrs
> > -def iface_add_
> > +def iface_add_
> > + # If the index is non-zero, this is an alias interface. Alias
> interfaces
> > + # represent additional interface addresses, and cannot have
> additional
> > + # attributes. So if this is an alias, stop right here.
>
> Well, ifupdown doesn't do much in the way of model validation. ;-) An
> alias is really just a (deprecated, I think) way to add more addresses to
> an interface. (that is to say, ifupdown may well allow them, I can't think
> of any cases where that isn't arguably a bug in ifupdown.)
>
> https:/
>
>
It's probably worth supplying what the preferred eni should look like;
I'm certainly no bonding expert but attempted to apply formatting as
suggested
by the ifenslave documentation.
Ryan