Hi,
I have a weird problem here. Sometimes the IPv6 connectivity of VMs is broken.
In this case the Router send ICMP Neighbor Solicitations to the VM, but doesn't get a response back, so the neigbor entry for these VMs are in failed or incomplete state on the router.
After capturing a bit it turns out that these Neigbhor Solicitations from the router comes in the active slave interface of the bond (active/passive), but proxmox doesn't forward the packets towards the vm. Verified this by sniffing on the VMs tap interface.
Running a 3-node PVE-CEPH Cluster with Proxmox 6.4-13 and ifupdown2.
Network Configuration on the proxmox hosts:
These are the neighbor solicitations from the router to the VM, captured on the active slave interface of bond1:
Maybe an multicast issue?
After the vm is migrated to another host IPv6 connectivity is back again. Don't know how long it takes, but after a while it is lost again. Sometime a reboot of the VM helps, sometimes not. I observed this behaviour on all 3-nodes.
Anyone any idea?
Thanks in advance and kind regards,
Kokel
I have a weird problem here. Sometimes the IPv6 connectivity of VMs is broken.
In this case the Router send ICMP Neighbor Solicitations to the VM, but doesn't get a response back, so the neigbor entry for these VMs are in failed or incomplete state on the router.
After capturing a bit it turns out that these Neigbhor Solicitations from the router comes in the active slave interface of the bond (active/passive), but proxmox doesn't forward the packets towards the vm. Verified this by sniffing on the VMs tap interface.
Running a 3-node PVE-CEPH Cluster with Proxmox 6.4-13 and ifupdown2.
Network Configuration on the proxmox hosts:
Code:
auto bond1
iface bond1 inet manual
bond-slaves ens2f0 ens4f0
bond-miimon 100
bond-mode active-backup
bond-primary ens2f0
auto ens2f0
iface ens2f0 inet manual
auto ens4f0
iface ens4f0 inet manual
auto vmbr0
iface vmbr0 inet manual
bridge-ports bond1
bridge-stp off
bridge-fd 0
bridge-vlan-aware yes
bridge-vids 2-4094
These are the neighbor solicitations from the router to the VM, captured on the active slave interface of bond1:
Code:
10:09:04.314586 IP6 fe80::21a:8cff:fef0:4faa > ff02::1:ff00:251: ICMP6, neighbor solicitation, who has caffe::251, length 32
10:09:05.313507 IP6 fe80::21a:8cff:fef0:4faa > ff02::1:ff00:251: ICMP6, neighbor solicitation, who has caffe::251, length 32
Maybe an multicast issue?
After the vm is migrated to another host IPv6 connectivity is back again. Don't know how long it takes, but after a while it is lost again. Sometime a reboot of the VM helps, sometimes not. I observed this behaviour on all 3-nodes.
Anyone any idea?
Thanks in advance and kind regards,
Kokel