GitHub user bradh352 added a comment to the discussion: Host based routing - EVPN-VXLAN to the host - Single VXLAN Device (SVD)
I can probably test out your script tonight to see if it breaks the current implementation. Regarding the new SVD mode, did you have to make changes elsewhere in cloudstack for this to work? Or just make sure the vlan-aware bridge and vxlan interface are up before cloudstack tries creating interfaces? The main problem I see here with the new SVD approach is the VNI and VLAN appear to be required to match. In my deployments I specifically make the VXLAN VNI range outside the VLAN range. I think you're going to need to enumerate all VNI/VLAN mappings to see if the VNI is already mapped to a vlan, and if not, generate a new mapping based on the next available VLAN id. If so, use the existing mapping. Not so sure on shutdown how hard it might be to unmap as multiple VMs might be using the same mapping, it may not be worth it, hard to say the likelihood of having used up all possible VLANs over time as VMs come and go. Finally, what method do you use to get the proper vxlan and bridge interface configured at boot? Do you have any netplan, networkd, nor NetworkManager example? I wouldn't mind testing it out. I'm currently using Ubuntu 24.04 for most of my tests, but looking at the linked libvirt news that's too old to pull in libvirt 11. GitHub link: https://github.com/apache/cloudstack/discussions/10786#discussioncomment-13347756 ---- This is an automatically sent email for users@cloudstack.apache.org. To unsubscribe, please send an email to: users-unsubscr...@cloudstack.apache.org
