Lynn Kerby wrote:

On Jul 20, 2008, at 10:07 PM, David Mair wrote:

Paolo Pedaletti wrote:
Ciao Uri:
Does using a different ifname help ?
PROXY:  ifname=tap2 and dmz2
WEBAPP: ifname=tap1 and dmz1
DB:     ifname=tap0 and dmz0
no, it doesn't.
always
destination host unreachable
Also check route on guests.
yes, they are correct, the default to external network eth0 while all
eth1 10.0.2.0/24 internal
nothing. :-(
someone has tried a working setup for 2 network connections, one
"public" and one "private" ?

Yes. On this host I have equivalents for all the VMware default networks (vmnet0, vmnet1 and vmnet8). Each of them is built with host tools (brctl, tunctl, host IP stack, ISC dhcpd).

It looks like the -net user and socket listen/connect options do not create shared virtual cabling between more than one and two guests respectively. The documentation implies that because it says that the -net socket mcast option does create a virtual cabling bus you can connect more than two guests to. This might work for you:

kvm -name PROXY
-net nic,vlan=0,macaddr=00:18:BE:EF:17:2A,model=rtl8139
    -net tap,vlan=0,ifname=tap0,script=./qemu-ifup.sh
-net nic,vlan=1,macaddr=00:18:BE:EF:17:2B,model=rtl8139
    -net socket,vlan=1,mcast=230.0.0.1:1234
-drive index=0,media=disk,if=scsi,file=./ubuntu-server.PROXY.root,boot=on
-drive index=1,media=disk,if=scsi,file=./ubuntu-server.PROXY.home
-drive index=2,media=disk,if=scsi,file=./ubuntu-server.PROXY.swap

kvm -name WEBAPP
-net nic,vlan=0,macaddr=00:18:BE:EF:17:1A,model=rtl8139
    -net tap,vlan=0,ifname=tap0,script=./qemu-ifup.sh
-net nic,vlan=1,macaddr=00:18:BE:EF:17:1B,model=rtl8139
    -net socket,vlan=1,mcast=230.0.0.1:1234
-drive index=0,media=disk,if=scsi,file=./ubuntu-server.WEB.root,boot=on
-drive index=1,media=disk,if=scsi,file=./ubuntu-server.WEB.home
-drive index=2,media=disk,if=scsi,file=./ubuntu-server.WEB.swap

kvm -name DB
-net nic,vlan=0,macaddr=00:18:BE:EF:17:0A,model=rtl8139
    -net tap,vlan=0,ifname=tap0,script=./qemu-ifup.sh
-net nic,vlan=1,macaddr=00:18:BE:EF:17:0B,model=rtl8139
    -net socket,vlan=1,mcast=230.0.0.1:1234
-drive index=0,media=disk,if=scsi,file=./ubuntu-server.DB.root,boot=on
-drive index=1,media=disk,if=scsi,file=./ubuntu-server.DB.home
-drive index=2,media=disk,if=scsi,file=./ubuntu-server.DB.swap

That's suitable for testing but not for production use IMO. The guest to guest traffic on the vlan=1 NICs may be visible on host public interfaces (tunneled in UDP multicast traffic of course). There will be no DHCP server on the vlan=1 guest NICs unless you install one. For a production environment you'll have to do more work.

I just noticed in David's post (and went back to yours also) that you are specifying ifname=tap0 for each of the guests. You will need to assign each guest it's *own* tap interface and attach that to the appropriate bridge(s) for this to work.

I'm not familiar with the '-net socket' stuff here, so I don't know what that does for you.

It should allow for private guest to guest [virtual] cabling without needing any host configuration. FWIW, I do it the way you do...with a host bridge and a tap for each guest.

--
David.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to