[SOLVED] KVM on Hetzner with bridged network

torusJKL · 11-04-2015, 03:53 AM

Hi

I have a root server at Hetzner.
I received 2 additional dedicated IPs.
Main IP: 5.5.15.132
Second IP: 5.5.15.151
Third IP: 5.5.11.226

I used the following post to set the first IP up with br0:
http://blog.alexdpsg.net/898/

This works very well for the first vm.

But I don't understand how I connect the second IP to the second VM.

I duplicated the br0.xml to br1.xml:

Code:

<network>
  <name>br1</name>
  <uuid>66ae947c-bfb3-4c7b-8d66-af3ea3cc3ba2</uuid>
  <forward dev='eth0' mode='route'>
    <interface dev='eth0'/>
  </forward>
  <bridge name='br1' stp='on' delay='0'/>
  <mac address='52:54:00:65:11:23'/>
  <ip address='5.5.11.226' netmask='255.255.255.254'>
  </ip>
</network>

and I used virsh to define the bridge

Code:

virsh net-define br0.xml
virsh net-start br0

brctl show gives me the following

Code:

bridge name     bridge id               STP enabled     interfaces
br0             8000.5404a6f16f10       yes             eth0
                                                        vnet0
                                                        vnet1
br1             8000.000000000000       yes

From what I understand both vnet go over the same bridge.
How can I configure vnet1 to connect over br1?

Or do I have to add the second IP to the same br0 interface?
And if yes how do I do that?

Thanks!

berndbausch · 11-04-2015, 05:28 AM

Quote:

How can I configure vnet1 to connect over br1?

What you really want, I think, is not plugging a virtual network interface into the br1 bridge, but rather a virtual machine into the br1 network. The brctl output indicates that this hasn't happened yet.

If you really need vnet1 to be connected to br1, create a VM including this XML snippet (based on the libvirt documentation):

Code:

    <interface type='network'>
      <source network='br1' />
      <target dev='vnet1'/>
      <mac address="00:11:22:33:44:55"/>
      ....

However, vnet1 is already in use, it would seem, so I don't think this will succeed.

Edit: Using the same names for your libvirt networks as for the bridges is a bit confusing. Consider naming the networks something like net_br0 etc.

torusJKL · 11-04-2015, 07:11 AM

Thanks for your reply.
I know have vnet1 connected to br1

Code:

bridge name     bridge id               STP enabled     interfaces
br0             8000.5404a6f16f10       yes             eth0
                                                        vnet0
br1             8000.fe540065cc44       yes             vnet1

But htere is still no network connection on the second vm.
Maybe because br1 is not connected to eth0? How would I create this connection?

berndbausch · 11-04-2015, 07:19 PM

To be honest, I am not quite sure how libvirt implements your configuration, but if eth0 is plugged into br0, I doubt it can also be plugged into br1. Thus no connectivity.

It's not quite clear what you want to achieve. You have a physical server and a few IP addresses. Do you want to assign those IP addresses to the VMs you run on the physical server?

This should work with a single bridged (not routed) libvirt network. Just plug both VMs into that network and configure them to use the other two IP addresses. This will basically extend the network outside of your physical server into the server, so that the VMs are directly plugged into the outside network.
See http://libvirt.org/formatnetwork.html#examplesBridge.

torusJKL · 11-05-2015, 12:16 AM

I didn't get the the bridge and the eth device have a 1:1 connection.

Could it be that the second IP does not work because it is in a different subnet?
This is how the hosts interfaces is configured:

Code:

# Loopback device:
auto lo
iface lo inet loopback

# device: eth0
auto  br0
iface br0 inet static
  address   5.5.15.132
  netmask   255.255.255.224
  gateway   5.5.15.129
   bridge_ports eth0
   bridge_stp on
   bridge_fd 0
iface br0 inet6 static
  address 2a01:2f4:222:101d::2
  netmask 64
  gateway fe80::1

Would I need to change the netmask?

Thanks!

torusJKL · 11-05-2015, 01:58 AM

I changed the netmask on the host interfaces configuration to

Code:

netmask 255.255.0.0

and this seems to help.
I can now ping the host and the first vm and vice versa.

But I can't ping a domain name and I can't access the vm from the outside (using ssh).

Somehow the ping results also look different.
on the working vm1 I see the ping coming from the non working vm2

Code:

ping 5.5.11.226
PING 5.5.11.226 (5.5.11.226) 56(84) bytes of data.
64 bytes from 5.5.11.226: icmp_seq=1 ttl=63 time=0.366 ms

But if I ping from the vm2 I get the result from the main IP:

Code:

ping 5.5.15.151
PING 5.5.15.151 (5.5.15.151) 56(84) bytes of data.
From 5.5.15.132: icmp_seq=1 Redirect Host(New nexthop: 5.5.15.132)
64 bytes from 5.5.11.226: icmp_seq=1 ttl=63 time=0.382 ms

Is this redirection the problem?

torusJKL · 11-05-2015, 04:39 AM

ip addr show br0

Code:

3: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default
    link/ether 54:04:a6:f1:6f:10 brd ff:ff:ff:ff:ff:ff
    inet 5.5.15.132/16 brd 5.5.255.255 scope global br0
       valid_lft forever preferred_lft forever

ip route

Code:

default via 5.5.15.129 dev br0
5.5.0.0/16 dev br0  proto kernel  scope link  src 5.5.15.132

torusJKL · 11-06-2015, 12:51 PM

I forgot to define the correct mac address in the vm.xml
It now works. :-)