Hello,

Our one real requirement for the network operation for this project is basic 
network filtering so that if for example someone just assigns the IP address of 
the gateway for the VLAN's subnet to their VM's vNIC it doesn't break 
everything.

The problems we noticed with our original PoC:

We were planning on using firewalld to protect the management IPs on the 
hypervisors but we noticed that the only way we could get the guest VM 
networking to function is to disable firewalld.
If we tried pinging an IP address that was assigned to a guest VM which is on a 
completely different network entirely the management NIC IP would just reply 
back with "filtered" until firewalld was disabled.

We also noticed that the SystemVMs by default create a "defacto bridge" between 
the public/guest network by having one NIC in each of these networks. This is 
undesirable for us so I am trying to find an alternative.

The questions are:

Is it expected that having firewalld running would make it impossible to route 
traffic to the guestVM network or did I just not find the right kernel 
tuning/routing/firewalld setting to make it work?

If you can use firewalld to protect the management NIC IP address does that 
interfere with the basic network filtering I mentioned as a requirement for 
this?

Is there a way to put the "public" vNIC for the system VMs in another 
VLAN/subnet other than the pool that the guest VMs use if so does anyone know 
how to facilitate this?

Our new PoC we are planning will use local storage on the HVs only and there 
will be 2x1G NICs [bond0] and 2x10G NICs [bond1]

Cloudbr0 [bond0] native VLAN will be the actual "management" network, and then 
if it's possible the "public network for the System VMs" will be just a trunked 
vlan to the bond.
Cloudbr1 [bond1] will just have the guest VM network trunked to the guestVM 
network

Does this seem like a workable configuration?

Thanks for your time.
-Drew











Reply via email to