Well, it's possible to get things working with a veth and have redirection to IFB on the non-ovs-enslaved port. But I'm currently trying to get away with the use of veth and see if an ovs-enslaved port can be leveraged directly for redirection to IFB for QoS. Maybe it's not possible as such based on the information that you shared.
Thanks & Regards, RS On Tue, 16 Jan, 2024, 1:57 pm Ilya Maximets, <i.maxim...@ovn.org> wrote: > On 1/16/24 13:09, Reshma Sreekumar wrote: > > So generally the IFB netdev is used to enforce QoS on egress when the > > traffic is ingress in reality i.e. one can redirect ingressing traffic > > (or even egress for that matter) to an IFB netdev and it will hit the > > egress qdisc on IFB, where we can configure the egress QoS. However, > > another usecase is to have consolidated traffic arrive on IFB for QoS . > > i.e. ingress traffic from different netdevs redirected to IFB netdev for > > a combined QoS. And my use-case is similar, where I want traffic arriving > > on several ports, including an ovs-enslaved port to be redirected to IFB > > for QoS. So, in that case, I believe, enslaving IFB into ovs will not > work > > for me. If ovs supports some sort of ingress redirection, then that > might help! > > So, you don't want traffic to return to OVS? If so, why you need that > interface to be in OVS at all? > > You may create another veth pair and add one side of it to OVS, then > add an OpenFlow rule to direct all traffic from veth1 to that new port. > Traffic will appear on the other side of a veth. Though I'm not sure > why OVS is in the picture in the first place. > > > > > Thanks & Regards, > > RS > > > > On Mon, Jan 15, 2024 at 10:23 PM Ilya Maximets <i.maxim...@ovn.org > <mailto:i.maxim...@ovn.org>> wrote: > > > > On 1/15/24 21:05, Reshma Sreekumar via discuss wrote: > > > I see..thanks for the explanation. My use-case is to have a > traffic redirection > > > from an ovs-enslaved port to IFB netdevice on ingress Qdisc. > Considering the > > > given mode of operation,may be it's better if I opt for some other > methods of > > > redirection to IFB netdevice? > > > > I never worked with IFB interfaces, so I'm not 100% sure that will > work, > > but you may remove your port from OVS and add you IFB interface > instead. > > Egress qdisc on IFB will be in use, you may protect it from being > removed > > by using linux-noop QoS type. > > > > I suppose, the traffic will enter OVS after IFB. So, you'll have > your > > veth1 (ingress qdisc) --> ifb (egress qdisc) --> OVS. Should be > fine > > if you're going to redirect all the traffic to ifb. > > > > Let us know if that works. > > > > Best regards, Ilya Maximets. > > > >
_______________________________________________ discuss mailing list disc...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-discuss
