[SOLVED] VMs not communicating accross nodes on cluster with SDN VXLAN configuration

[abenoit-reeliant]

Hello. I'm trying to set up a Proxmox cluster and have VM on multiple nodes communicate with each other but it doesn't seem to work (cannot ping across multiple nodes)

I tried to follow the example https://pve.proxmox.com/pve-docs/chapter-pvesdn.html#pvesdn_setup_example_vxlan

Here is the details of the setup.

• PVE version is pve-manager/8.1.4/ec5affc9e41f1d79 (running kernel: 6.5.11-8-pve)
• 2 PVE nodes at the moment : 172.16.12.3 and 172.16.12.4
• PVE nodes are communicating through "Linux VLAN" interfaces (through a network set up by my hosting service provider)
  • Node 1 (172.16.12.3) has VLAN tag 3503
  • Node 2 (172.16.12.4) has VLAN tag 3309
• The 2 nodes communicate fine for cluster operations (administration for both nodes, migration, ...)
• SDN is setup this way:
  • 1 VXLAN zone:
    • ID: vxlanpca
    • Peer addresss list: 172.16.12.3,172.16.12.4
    • MTU: empty (auto)
    • IPAM: pve
  • 1 VNet :
    • Name: vxnetpca
    • Zone: vxlanpca
    • Tag: 99
    • VLAN Aware: checked
• VM 1 on node 1 has IP 192.168.100.4/27 and VM 2 on node 2 has IP 192.168.100.5/27
  • Both have mtu: 1450 in netplan configuration
• Both use:
  • Bridge: vxnetpca
  • VLAN Tag: empty
  • Firewall: unchecked
  • Model: VirtIO
  • MTU: empty

Both VM fail to ping the other one or view any TCP port as open.
What am I doing wrong?

 

[abenoit-reeliant]

OK, I was forgetting the firewall aspect. After the adding a rule to accept incoming trafic on the private VLAN interfaces (eno1.3503 and eno1.3309 respectively), ping is OK.