Cluster failures

[odesport]

Hello,

I'm working on a 6 nodes cluster (Proxmox 3.1). Last week the cluster failed. I've tried to delete the cluster config files but I was unable to create it again. So I re-installed the nodes but I can't build the cluster. The nodes I try to add show "Waiting for quorum". I've checked multicast :

ping 239.192.55.50

response: nothing

omping node1 node2

node1 : waiting for response msg
node1 : joined (S,G) = (*, 232.43.211.234), pinging
node1 : unicast, seq=1, size=69 bytes, dist=0, time=0.202ms
node1 : unicast, seq=2, size=69 bytes, dist=0, time=0.278ms
node1 : unicast, seq=3, size=69 bytes, dist=0, time=0.291ms
node1 : unicast, seq=4, size=69 bytes, dist=0, time=0.302ms
node1 : unicast, seq=5, size=69 bytes, dist=0, time=0.295ms
node1 : unicast, seq=6, size=69 bytes, dist=0, time=0.279ms
node1 : unicast, seq=7, size=69 bytes, dist=0, time=0.267ms
node1 : unicast, seq=8, size=69 bytes, dist=0, time=0.297ms

node1 : ssmpingd
node2 : asmping 239.192.55.50 172.16.0.10

asmping joined (S,G) = (*,239.192.55.234)
pinging 172.16.0.11 from 172.16.0.12
unicast from 172.16.0.10, seq=1 dist=0 time=0.769 ms
unicast from 172.16.0.10, seq=2 dist=0 time=0.290 ms
unicast from 172.16.0.10, seq=3 dist=0 time=0.296 ms
unicast from 172.16.0.10, seq=4 dist=0 time=0.288 ms
unicast from 172.16.0.10, seq=5 dist=0 time=0.293 ms
unicast from 172.16.0.10, seq=6 dist=0 time=0.286 ms
unicast from 172.16.0.10, seq=7 dist=0 time=0.289 ms
unicast from 172.16.0.10, seq=8 dist=0 time=0.294 ms

There is no multicast response, as I see in http://pve.proxmox.com/wiki/Multicast_notes

Do you think it can explain the problem of cluster ?

 

[dietmar]

Do you think it can explain the problem of cluster ?

Cluster does not work without multicast.

 

[odesport]

I know. I've checked the switches and multicast is allowed.

 

[odesport]

I've tried with two debian Wheezy servers on the same network and multicast packets are send. What can I check on my Proxmox servers ?

 

[odesport]

I've found the problem. Multicast snooping is enabled on bridge interfaces. To disable it :

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

 

[mir]

This is apparently switch depended. This is how it looks on my servers and I have no problems with multicast:
$ cat /sys/devices/virtual/net/vmbr0/bridge/multicast_snooping
1

 

[odesport]

You're right. The switch have igmp snooping enabled.

 

[mir]

My switch also have igmp snoping enabled and this does not break multicast even if multicast_snooping is enabled on virtual bridges. So again, this is switch depended and most be treated as such.

 

[odesport]

Thanks. I will try to find whet setting on the switches I have to change.