Hi Wido, I'm not at a point to add the zones yet. I'm currently stuck with EVPN (as before unfortunately). For my hypervisors, of course they're running FRR. Meanwhile, my switches are running Mellanox Onyx. My issue is that the type 2 route is not being accepted because of different route targets.
On the switch, I can of course manually set the import/export RT just for the VNI I'm using for CS management. But for the hypervisors, is there any config that can be done on FRR so that every VNI will follow the same RTs? For sure we don't want to configure the RTs for every single VNI (say we allocated over 10k VNI per zone). Oh ya, the hypervisors are running EBGP between them and the switches. Current RTs look like this for VNI 10027 (management): 1. Switches (manually configured for both import/export) <OUR_IRR_ASN>:10027 2. Hypervisors FRR (auto) First = 38933:10027 Second = 38932:10027 After looking at this blog post from ipspace <https://www.ipspace.net/Data_Center_BGP/EVPN_Route_Target_Considerations>, it looks like it's a bit hard to integrate the switches that we had with FRR. Because different vendors seem to generate the auto RTs differently. Thanks for any recommendations! On Mon, 15 Jul 2024 at 18:54, Wido den Hollander <w...@widodh.nl> wrote: > See my reply inline > > Op 12/07/2024 om 04:33 schreef Muhammad Hanis Irfan Mohd Zaid: > > 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? > > > > No, you can isolate all traffic. Public/Management traffic just goes via > the loopback address of the hypervisor. > > > > https://docs.google.com/drawings/d/1oPWU5p_wUd9UPhXGZg7j4acu5XYPbkLWzmihi6Qbwl8/edit > < > 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. > > > > There is no need to create a bond0 device. Via each uplink (p1p1 and > p1p2 in this example) you create a BGP session with the ToR. MLAG is not > required there. > > On the hypervisor you then have p1p1 and p1p2 over which you create a > BGP (Unnumbered) session with the ToR. > > We have then created only one bridge, cloudbr1: > > root@hv-138-a13-37:~# ip addr show dev cloudbr1 > 6: cloudbr1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue > state UP group default qlen 1000 > link/ether 12:5b:28:2a:a7:f4 brd ff:ff:ff:ff:ff:ff > inet 10.100.2.108/20 brd 10.100.15.255 scope global cloudbr1 > valid_lft forever preferred_lft forever > root@hv-138-a13-37:~# > > root@hv-138-a13-37:~# bridge link|grep cloudbr1 > 7: vxlan100: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 master cloudbr1 > state forwarding priority 32 cost 100 > root@hv-138-a13-37:~# > > We have used VNI 100 as our internal POD network which is also carried > by Frr: > > hv-138-a13-37.ams02.cldin.net# sh evpn vni 100 > VNI: 100 > Type: L2 > Tenant VRF: default > VxLAN interface: vxlan > VxLAN ifIndex: 7 > Local VTEP IP: 10.255.255.108 > Mcast group: 0.0.0.0 > Remote VTEPs for this VNI: > 10.255.255.81 flood: HER > 10.255.255.123 flood: HER > 10.255.255.11 flood: HER > 10.255.255.99 flood: HER > ... > ... > 10.255.255.36 flood: HER > 10.255.255.35 flood: HER > 10.255.255.33 flood: HER > 10.255.255.2 flood: HER > 10.255.255.106 flood: HER > 10.255.255.100 flood: HER > 10.255.255.105 flood: HER > 10.255.255.124 flood: HER > 10.255.255.127 flood: HER > 10.255.255.107 flood: HER > 10.255.255.126 flood: HER > 10.255.255.104 flood: HER > Number of MACs (local and remote) known for this VNI: 109 > Number of ARPs (IPv4 and IPv6, local and remote) known for this VNI: 107 > Advertise-gw-macip: No > hv-138-a13-37.ams02.cldin.net# > > The management server is a VM running somewhere on a Proxmox environment > (not really relevant) and has a NIC in VNI 100. This is how the Agents > communicate with the Management server. > > The hypervisors have a /128 and /32 on their loopback which they use to > communicate to their storage (Ceph and TrueNAS NFS): > > root@hv-138-a13-37:~# ip addr show dev lo > 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN > group default qlen 1000 > link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 > inet 127.0.0.1/8 scope host lo > valid_lft forever preferred_lft forever > inet 10.255.255.108/32 brd 10.255.255.108 scope global lo > valid_lft forever preferred_lft forever > inet6 2a05:xxxx:xxxx:2::108/128 scope global > valid_lft forever preferred_lft forever > inet6 ::1/128 scope host > valid_lft forever preferred_lft forever > root@hv-138-a13-37:~# > > Wido > > > May I know how you achieve this Wido? > > > > On Sat, 29 Jun 2024 at 12:53, Wido den Hollander <w...@widodh.nl > > <mailto: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> > > > <mailto: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> > < > 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> > > > > > < > 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> > > > <mailto: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> > > > <mailto:users@cloudstack.apache.org > > <mailto:users@cloudstack.apache.org>> > > > >>> Cc: w...@widodh.nl <mailto:w...@widodh.nl> > > <mailto: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 > > > >> > > > > > > > > > >
