A network restart is leading missing IP address on vm-public

[osokin]

Hi,

I've figured our that service netif restart removes IP address from the vm-public interface, and there's no way (except the manual one) to assign IP address to the interface back, i.e.:

# ifconfig vm-public inet 10.0.0.1/24

Would that be good idea to move vm-public initialization to /etc/rc.conf and make that explicitly.

[akrivulya]

As a workaround one's can create bridge interface with /etc/rc.conf and make vm's switch of type manual.

[osokin]

As a workaround one's can create bridge interface with /etc/rc.conf and make vm's switch of type manual.

Just to make that a bit more clear, so /etc/rc.conf records should be something like the following:

cloned_interfaces="vm-public"
ifconfig_vm-public="addm wlan0" # change the interface's name on your taste
config_vm-public="inet 10.0.1.1/24 up"

Is there anything needs to add/change in /vm/.config/system.conf file?

[akrivulya]

Something like this in /etc/rc.conf:

cloned_interfaces="bridge0"
ifconfig_bridge0="addm wlan0 name vm-public 10.0.1.1/24" # change the interface's name on your taste
ifconfig_wlan0="up"

and then
# vm switch create -t manual -b vm-public vm-public

[osokin]

Something like this in /etc/rc.conf:

cloned_interfaces="bridge0"
ifconfig_bridge0="addm wlan0 name vm-public 10.0.1.1/24" # change the interface's name on your taste
ifconfig_wlan0="up"

and then # vm switch create -t manual -b vm-public vm-public

That doesn't work. Also, it would be great to avoid using - symbol in the network interfaces names. Would that be good idea to rename vm-public to vmpublic0 or so?

[akrivulya]

Also, it would be great to avoid using - symbol in the network interfaces names. Would that be good idea to rename vm-public to vmpublic0 or so?

Generally speaking if you need to configure vm-public0 network interface in /etc/rc.conf it must be present as vm_public0, for example:
ifconfig_vm_public0="DHCP"

This configuration works for me:

cloned_interfaces="bridge0"
create_args_bridge0="link 58:9c:fc:b1:dc:6d"
ifconfig_bridge0="192.168.254.1/27 name vm-LAN"

[osokin]

Also, it would be great to avoid using - symbol in the network interfaces names. Would that be good idea to rename vm-public to vmpublic0 or so?

Generally speaking if you need to configure vm-public0 network interface in /etc/rc.conf it must be present as vm_public0, for example: ifconfig_vm_public0="DHCP"

This configuration works for me:

cloned_interfaces="bridge0"
create_args_bridge0="link 58:9c:fc:b1:dc:6d"
ifconfig_bridge0="192.168.254.1/27 name vm-LAN"

Thanks. I've decided to use bridge0 interface name in the vm-bhyve configuration files.

[GogoFC]

I was just wondering why assign an IP to a switch?

[akrivulya]

I was just wondering why assign an IP to a switch?

For example if hypervisor is a gateway for VM's network.

[GogoFC]

I'm not an expert but I understand some. So do you have bunch of different networks? I've only figured out how to use the physical router with the switch and then some other virtual switches but for those I have VMs running as routers. I haven't figured out how to do pf to do routing on the host. It would be nice to have website dedicated to bhyve showcase of people's setup.

[akrivulya]

Yes, in some cases hardware router or router in VM may be a little overhead. pf nat on hypervisor also can do it. Even more, for example we have a lot of customers vm's running inside vnet jails with custom pf rules and ip addressing. Each jail has an internal bridge for vm's network with ip address assigned which acts as a gateway.