On Sat, Jun 11, 2016 at 04:57:29PM -0700, Justin Pettit wrote: > > > On Jun 10, 2016, at 7:07 PM, Ben Pfaff <b...@ovn.org> wrote: > > > > On Fri, Jun 10, 2016 at 05:03:31PM -0700, Justin Pettit wrote: > >> > >>> On Jun 9, 2016, at 5:09 PM, Ben Pfaff <b...@ovn.org> wrote: > >>> > >>> I think that lrp-add with --may-exist should report an error if the MAC > >>> or NETWORK or [PEER] differ from the existing port's configuration. > >> > >> Currently, adding the port fails if the port belongs to a different > >> router. I assume you are suggesting that we report an error if those > >> fields are changing when re-adding to the same router, correct? > > > > Yes. > > > > The idea of --may-exist is to make the command idempotent: if the same > > command has already been executed, then it turns it into a no-op. But > > if a different command has been executed (with different MAC or NETWORK > > or PEER), it's not what the user has asked for. > > > > This follows the precedent set by the add-port command in ovs-vsctl, > > which with --may-exist complains if, for example, the interfaces > > specified for a bond are different from the ones that actually exist. > > > > I think there is at least one other precedent, but that's the one that > > comes to mind. > > Got it. I've added that check and a couple of checks to the unit > tests. I've attached a diff below.
Thanks! _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
