question about clustering and seperate 10GB network

[ProxmoxBoy1]

We are installing new servers and for this we have made a small 10GB network for the cluster communication.
The normal management is going over the "normal" network witch is reachable from the whole network. In the new PVE these are VMBR0
This works.

Now to make a 10GBE cluster we made the new part with 2 10GBe switches with seperate connection to the other building
This works
On all PVE the 10GB is connected to VMBR1 on each PVE

(Not on this picture but in ech bulding we have a runing NAS connected also to the 10GBE network.)

Now the big question, is we setup the cluster how do we tel the cluster to use the 10GBE network ?

Or should we change the setup and connect all management also to the 10GBE network.
The management VLAN 102 is subnet 10.11.20.0/24
The 10GBE valan 172 is subnet 10.11.30.0/24

The VMBR3 for the VM's connection to the process

 

[VictorSTS]

What do you want to use that 10G interface for? Ceph? VM traffic? ZFS replication? Corosync does not need 10G

 

[ProxmoxBoy1]

VictorSTS

I have added the proxmox servers as the picture drawn before​

Now
If i pull out 1 off the VMBR0 (management connection) the system stays stable.
If i pul out 1 of the 10GBE links the Corosync is broken (as planned)

Now if i migrate a VM to antoher member it looks that the transfer is done by the managment VMBR0 and not over the fast lane VMBR1 10GB.

Now the question can this be changed with bond or bridge
(If this is not posible then i must remove the cluster and start over with new network design)

 

[sb-jw]

In the cluster options (Datacenter > Options, at least I think) you can specify the migration network.

 

[VictorSTS]

If i pul out 1 of the 10GBE links the Corosync is broken (as planned)

Why would you want to use a single Corosync link having two nics/networks?

https://pve.proxmox.com/wiki/Cluster_Manager#pvecm_redundancy