Multicast on container with Proxmox VE 2.3

[darkned]

Hi,

Noting that a VPS didnt receive every network transmissions, i investigated about multicast forwarding capabilities in bridged network interface.
I was surprised to note simultaneous tcpdump outputs were not identical between vmbr1 and veth<CTID> interfaces on Proxmox 2.3.

On the HN:

root@proxmox1:~#tcpdump -i vmbr1 src node031.cluster.lbt
[...]
23:41:01.505764 ARP, Request who-has 10.0.0.15 tell node031.cluster.lbt, length 46
23:41:12.092888 IP node031.cluster.lbt.8662 > 10.0.0.15.54726: Flags [R.], seq 0, ack 2477665065, win 0, length 0
23:41:12.093098 IP node031.cluster.lbt.8662 > 10.0.0.15.54727: Flags [R.], seq 0, ack 3111031119, win 0, length 0
23:41:14.935252 IP node031.cluster.lbt.58001 > 239.2.11.71.8662: UDP, length 60
23:41:26.436087 IP node031.cluster.lbt.8662 > 10.0.0.15.54742: Flags [R.], seq 0, ack 3691320116, win 0, length 0
23:41:26.436294 IP node031.cluster.lbt.8662 > 10.0.0.15.54743: Flags [R.], seq 0, ack 3584061608, win 0, length 0
23:41:30.560402 IP node031.cluster.lbt.1023 > 10.0.0.14.15001: UDP, length 364
23:41:31.435810 ARP, Reply node031.cluster.lbt is-at 00:25:90:74:e0:ee (oui Unknown), length 46
23:41:34.902150 IP node031.cluster.lbt.58001 > 239.2.11.71.8662: UDP, length 204
23:41:34.902167 IP node031.cluster.lbt.58001 > 239.2.11.71.8662: UDP, length 56
23:41:34.902173 IP node031.cluster.lbt.58001 > 239.2.11.71.8662: UDP, length 196
23:41:34.902178 IP node031.cluster.lbt.58001 > 239.2.11.71.8662: UDP, length 60
23:41:34.935645 IP node031.cluster.lbt.58001 > 239.2.11.71.8662: UDP, length 172
23:41:34.935672 IP node031.cluster.lbt.58001 > 239.2.11.71.8662: UDP, length 60

and

root@proxmox1:~#tcpdump -i veth108.0 src node031.cluster.lbt
[...]
23:41:01.505787 ARP, Request who-has 10.0.0.15 tell node031.cluster.lbt, length 46
23:41:12.092896 IP node031.cluster.lbt.8662 > 10.0.0.15.54726: Flags [R.], seq 0, ack 2477665065, win 0, length 0
23:41:12.093119 IP node031.cluster.lbt.8662 > 10.0.0.15.54727: Flags [R.], seq 0, ack 3111031119, win 0, length 0
23:41:26.436096 IP node031.cluster.lbt.8662 > 10.0.0.15.54742: Flags [R.], seq 0, ack 3691320116, win 0, length 0
23:41:26.436304 IP node031.cluster.lbt.8662 > 10.0.0.15.54743: Flags [R.], seq 0, ack 3584061608, win 0, length 0
23:41:31.435826 ARP, Reply node031.cluster.lbt is-at 00:25:90:74:e0:ee (oui Unknown), length 46

PS: node031 is a physical machine from my HPC cluster.

As you can see, all udp packets aren't received by my container but are availables on my HN bridged interface .

Someone know why, and above all, how to fix it to get all packets?

On Proxmox 1.9, with exactly the same proxmox node settings, these troubles are not: tcpdump outputs are identical and VPS receive all communications.

Thanks in advance.

 

[darkned]

Just for info, i just upgraded my proxmox server from 2.3 version to 3.1 without positive effect; i meet the same issue: no multicast udp packets comes into VPS.

 

[dietmar]

Maybe a problem with multicast snooping. Try

# echo 0 > /sys/devices/virtual/net/vmbr0/bridge/multicast_snooping

 

[darkned]

Maybe a problem with multicast snooping. Try

# echo 0 > /sys/devices/virtual/net/vmbr0/bridge/multicast_snooping

First of all, thanks for your reply. But.. without effect, unfortunately.

I justly thinked this may be concern IGMP snooping, network switch side...

See you.