Hope this is the right forum for asking this
question.
I am finding that my HVM guests are using tun
interfaces in the Dom0 bridge instead of vif interfaces, looks like this is a
QEMU thing. The vif interfaces are generated at startup but not used. There are
2 major problems with this.
1) All
of my guests are on 2 networks with 2 NICS. Although the correct vif interfaces
are being put in the correct bridge (for what it’s worth), the tun
interfaces all go in the first bridge and are not named in a way that makes
them identifiable for moving.
2) If
the vif interface is not used I cannot dictate MAC address from ‘vmname.hvm’
and therefore cannot allocate IP address from DHCP reservations.
In the following examples 2 guest OS’s
exist (vmid 10 and 11).
In the first case Dom0 and both vm’s
work on net1 but vm’s do not work on net2. This is default behaviour.
In the 2nd case some changes
have been made and all vm’s work on all networks.
[root@XenHost02 ~]# brctl show
bridge name
bridge
id
STP enabled interfaces
xenbr0
8000.3264131ec979
no
peth0
tun0
tun1
tun2
tun3
vif0.0
vif10.0
vif11.0
xenbr1
8000.d6a5a5f280b3
no
peth1
vif0.1
vif10.1
vif11.1
[root@XenHost02 ~]# brctl show
bridge name
bridge
id
STP enabled interfaces
xenbr0
8000.3264131ec979
no
peth0
tun0
tun2
vif0.0
xenbr1
8000.d6a5a5f280b3
no
peth1
tun1
tun3
vif0.1
Note that I removed the vm vifs in case2, this is
because they don’t do anything anyway, frustrating because they’re
in the correct bridges. I could write a cludgy script to automate this, but
identification of tun interfaces with 10+ vm’s is going to be a problem
and I still will be unable to allocate MAC-address which for me is a big issue.
Is this default behaviour for QEMU? It’s going
to cause problems for HVM guests in the future.
Thanks for listening.