Using the subnet broadcast address for clustering instead of Multicast?

[Maxnet]

The Wiki mentions:

All nodes must be in the same network as it uses IP Multicast to communicate between nodes (See also Corosync Cluster Engine). Note: Some switches does not support IP multicast by default and must be manually enabled first.

Was wondering if I could get around having to mess with the switch configuration, by telling Corosync to use the broadcast address for its communication instead of multicast.
I recall it has an option for that.

Does the Proxmox cluster stuff generate a lot of traffic, or are there any other reasons why one wouldn't go the easy way and use the broadcast address?

 

[dietmar]

Does the Proxmox cluster stuff generate a lot of traffic,

What exactly is 'a lot'?

or are there any other reasons why one wouldn't go the easy way and use the broadcast address?

corosync/cman use multicast as default. If you want other settings you can simple set them in /etc/pve/cluster.conf

# man cman

Does that work for you?

 

[Maxnet]

What exactly is 'a lot'?

Was just wondering whether it sends a negligible amount, e.g. just a packet a second for each node to see if all nodes are still up.
If that's the case, I wouldn't mind if servers that are not part of the cluster receive the traffic as well.

"a lot" meaning things like fancy distributed file systems that would send a packet for each write inside a VM.

corosync/cman use multicast as default. If you want other settings you can simple set them in /etc/pve/cluster.conf

# man cman

Does that work for you?

Looks good, will try it out later.
I wonder why corosync uses multicast as default. Would think that a default that does not require any special switch configuration would be better.

 

[dietmar]

I wonder why corosync uses multicast as default. Would think that a default that does not require any special switch configuration would be better.

Would you mind to ask on the corosync list (I am curious too)?