I'm still trying to figure out the physical network setup based on your
talks, Wido.

Just to clarify that VXLAN isolation only works for Guest traffic types,
correct? Then how would Management and Public traffic type be configured
with a setup running fully L3 (BGP unnumbered etc) between the leaf
switches and KVM hosts?

https://docs.google.com/drawings/d/1oPWU5p_wUd9UPhXGZg7j4acu5XYPbkLWzmihi6Qbwl8/edit

>From the diagram I found here, it looks like they create a VLAN interface
on top of a bond as their VTEP. For the public network, another VLAN
interface is created for it.

May I know how you achieve this Wido?

On Sat, 29 Jun 2024 at 12:53, Wido den Hollander <w...@widodh.nl> wrote:

>
>
> Op 24/06/2024 om 09:19 schreef Muhammad Hanis Irfan Mohd Zaid:
> > Then how would the other networks (guest and public) be configured
> > during the zone creation process?
> >
>
> Just use the defaults labels, because the rest will be done by the
> hypervisor.
>
> Make sure you have selected VXLAN and you are good to go!
>
> > For management, I've created a fixed VNI 10028. The management server
> > and the hosts can ping each other via this VNI. I'm using cloudbr0 tho
> > for the internal comm bridge name.
> >
> > I'm in need to understand how should physical network be configured with
> > the first zone creation wizard. I've added the modifyvxlan.sh script to
> > folder. And so far BGP EVPN works perfectly to the ToR switches.
> >
>
> Once the network is running, try to start the first VM and see what the
> agent does. VXLAN devices and bridges will be created.
>
> > I don't think I can join the conference because of financial reasons 😂
> >
>
> Remotely :-)
>
> > On Mon, 24 Jun 2024, 14:22 Wido den Hollander, <w...@widodh.nl
> > <mailto:w...@widodh.nl>> wrote:
> >
> >
> >
> >     Op 21/06/2024 om 05:22 schreef Muhammad Hanis Irfan Mohd Zaid:
> >      > On the web UI, when setting up a zone, how would I create the
> >     physical
> >      > networks and assign the correct traffic types and labels?
> >      >
> >
> >     You don't need to do that. No need for separate networks.
> >
> >     Just make sure you have a cloudbr1 for internal commmunication, this
> >     can
> >     be a fixed VNI where you create a network using systemd-networkd for
> >     example.
> >
> >      > You can refer to a simple diagram I drew below.
> >      >
> >
> https://drive.google.com/file/d/1_xGUxEu-U2mJltdIj94CMK0s4zAH-Ret/view?usp=drive_link
> <
> https://drive.google.com/file/d/1_xGUxEu-U2mJltdIj94CMK0s4zAH-Ret/view?usp=drive_link
> >
> >      >
> >      > So let's say based on the diagram, I create a physical network
> named
> >      > "Management", an isolation method "VLAN", with traffic type
> >     "MANAGEMENT"
> >      > and label of "cloudbr0". Next, I create another physical network
> >     named
> >      > "Public", an isolation method "VXLAN", with traffic type "PUBLIC"
> >     and label
> >      > of "lo". Lastly, I create another physical network named "Guest",
> an
> >      > isolation method "VXLAN", with traffic type "GUEST" and label of
> >     "lo".
> >      >
> >      > Will this work? Is my understanding of physical networks correct?
> >      >
> >
> >     No, again. Not needed. The modifyvxlan.sh script creates bridges on
> the
> >     fly and your Frrouting will see these bridges and start the VXLAN
> >     advertisement.
> >
> >     Anything specific in your config you need help with?
> >
> >     Might be good if I gave a VXLAN masterclass during the conference in
> >     November ;-)
> >
> >     Wido
> >
> >      >
> >      >
> >      > On Thu, 20 Jun 2024 at 20:21, Wido den Hollander
> >     <w...@widodh.nl.invalid>
> >      > wrote:
> >      >
> >      >>
> >      >>
> >      >> Op 20/06/2024 om 11:15 schreef Alex Mattioli:
> >      >>> Hi Muhammad,
> >      >>>
> >      >>> Are you planning on using VXLAN or bridges?  Those are mutually
> >      >> exclusive, with VXLAN you have a single VLAN interface with an
> >     IP which is
> >      >> the VTEP (Virtual Tunnel EndPoint) for your VXLAN encapsulated
> >     traffic.
> >      >>
> >      >> Yes, however, keep in mind that the script 'modifyvxlan.sh'
> creates
> >      >> Linux bridges on the fly, one for each VNI.
> >      >>
> >      >> The administrator doesn't need to do anything though, this is
> >     all done
> >      >> by CS.
> >      >>
> >      >> Just make sure you use this script:
> >      >>
> >     https://download.cloudstack.org/tools/scripts/vxlan/modifyvxlan.sh
> >     <https://download.cloudstack.org/tools/scripts/vxlan/modifyvxlan.sh>
> >      >>
> >      >> Wido
> >      >>
> >      >>>
> >      >>> Cheers
> >      >>> Alex
> >      >>>
> >      >>>
> >      >>>
> >      >>>
> >      >>> -----Original Message-----
> >      >>> From: Muhammad Hanis Irfan Mohd Zaid <hanisirfan.w...@gmail.com
> >     <mailto:hanisirfan.w...@gmail.com>>
> >      >>> Sent: Thursday, June 20, 2024 8:59 AM
> >      >>> To: users@cloudstack.apache.org
> >     <mailto:users@cloudstack.apache.org>
> >      >>> Cc: w...@widodh.nl <mailto:w...@widodh.nl>
> >      >>> Subject: Physical network labels when using VXLAN
> >      >>>
> >      >>> Hi. We're trying to deploy a POC environment with VXLAN EVPN.
> The
> >      >> underlay works perfectly and the overlay when creating a bridge
> >     for the
> >      >> management network (cloudbr0) can ping without any issues
> >     between the hosts
> >      >> and management server.
> >      >>>
> >      >>> Now I'm trying to figure out how the bridges should be
> >     configured for
> >      >> the guest and public network. The hosts are fully running L3
> >     towards our
> >      >> leaf switches. I'm clueless when trying to configure the
> >     physical networks
> >      >> of the zone in the web UI.
> >      >>>
> >      >>> Any suggestions? Thanks
> >      >>
> >      >
> >
>

Reply via email to