GitHub user joolli added a comment to the discussion: Host based routing - EVPN-VXLAN to the host - Single VXLAN Device (SVD)
Yes, the first part shows how you would create a vlan-aware bridge and a vxlan interface (that is also vlan-aware) and make it a member of the bridge, add a vlan and map a VNI to it. The VXLAN interface will use the VNI info (tunnel_info). All the rest is just for demonstration purposes. Yes, I wanted to distinguish between the two numbers in the demonstration. They can be the same. We use a libvirt hook that reads the xml, where we have some extra info about the vlan each interface is a member of, and that script sets the vlan of the bridge member. I'm not sure what you mean by knowing the VNI before creating the VM. You can add a VLAN to VNI mapping (create a VNI) later even if there are VM's already a member of a VLAN on the bridge, if that is what you mean? I don't think there is a performance benefit. It just scales more as it uses less memory, so you're able to create more VNI's. Much less information for the kernel to manage as you only have the two interfaces, the bridge and vxlan interface and those two hold all the information, rather than having a bridge (with a bunch of attributes) and a vxlan interface (also with a bunch of attributes) for every VNI. GitHub link: https://github.com/apache/cloudstack/discussions/10786#discussioncomment-12991753 ---- This is an automatically sent email for users@cloudstack.apache.org. To unsubscribe, please send an email to: users-unsubscr...@cloudstack.apache.org
