I''m trying to move from Fedora Core 5 to Fedora 7. I''ve run into an issue where the xen guests, which need several bridged interfaces bound to dummy devices so they can communicate with each other (similar to host-only networking with VMWare), won''t boot. I think I''ve traced the problem to a lack of the dummy1, dummy2, and dummy3 devices, and I''m not sure how to fix it. What I''ll describe below is the setup that worked in Fedora Core 5: I used the following custom network-bridge script: #!/bin/sh dir=$(dirname "$0") "$dir/network-bridge" "$@" vifnum=0 netdev=dummy0 "$dir/network-bridge" "$@" vifnum=1 netdev=dummy1 "$dir/network-bridge" "$@" vifnum=2 netdev=dummy2 "$dir/network-bridge" "$@" vifnum=3 netdev=dummy3 And the following line was used in the .xen files passed to xm create: vif = [''mac=2E:22:22:22:22:20,bridge=xenbr1'', ''mac=2E:22:22:22:22:21,bridge=xenbr2'', ''mac=2E:22:22:22:22:22,bridge=xenbr3''] Running the script resulted in the following: Using config file "./guest1.xen". Error: Device 0 (vif) could not be connected. Could not find bridge device xenbr1 If I try to run the custom network-bridge script manually passing the "start" command I get several errors: Device "dummy1" does not exist. usage: ifdown <device name> Device "dummy1" does not exist. Cannot find device "dummy1" SIOCGIFFLAGS: No such device Device "dummy2" does not exist. usage: ifdown <device name> Device "dummy2" does not exist. Cannot find device "dummy2" SIOCGIFFLAGS: No such device Device "dummy3" does not exist. usage: ifdown <device name> Device "dummy3" does not exist. Cannot find device "dummy3" SIOCGIFFLAGS: No such device Any ideas on how to get this working would be greatly appreciated. - David
David, The configuration has changed quite a bit between F5 and F7. I''ve found that there is not much you can do unless you go "virsh dumpxml ...". In my experimentation, I made the most progress when I defined the device with dummy data in the GUI version. I could then dump the result out and make changes in the xml template I got. Finally I would reload by "virsh define ...". Hope this helps. -- Per -----Original Message----- From: fedora-xen-bounces@redhat.com [mailto:fedora-xen-bounces@redhat.com] On Behalf Of David Mueller Sent: Tuesday, August 21, 2007 10:07 AM To: fedora-xen@redhat.com Subject: [Fedora-xen] Fedora 7 - missing dummy interfaces I''m trying to move from Fedora Core 5 to Fedora 7. I''ve run into an issue where the xen guests, which need several bridged interfaces bound to dummy devices so they can communicate with each other (similar to host-only networking with VMWare), won''t boot. I think I''ve traced the problem to a lack of the dummy1, dummy2, and dummy3 devices, and I''m not sure how to fix it. What I''ll describe below is the setup that worked in Fedora Core 5: I used the following custom network-bridge script: #!/bin/sh dir=$(dirname "$0") "$dir/network-bridge" "$@" vifnum=0 netdev=dummy0 "$dir/network-bridge" "$@" vifnum=1 netdev=dummy1 "$dir/network-bridge" "$@" vifnum=2 netdev=dummy2 "$dir/network-bridge" "$@" vifnum=3 netdev=dummy3 And the following line was used in the .xen files passed to xm create: vif = [''mac=2E:22:22:22:22:20,bridge=xenbr1'', ''mac=2E:22:22:22:22:21,bridge=xenbr2'', ''mac=2E:22:22:22:22:22,bridge=xenbr3''] Running the script resulted in the following: Using config file "./guest1.xen". Error: Device 0 (vif) could not be connected. Could not find bridge device xenbr1 If I try to run the custom network-bridge script manually passing the "start" command I get several errors: Device "dummy1" does not exist. usage: ifdown <device name> Device "dummy1" does not exist. Cannot find device "dummy1" SIOCGIFFLAGS: No such device Device "dummy2" does not exist. usage: ifdown <device name> Device "dummy2" does not exist. Cannot find device "dummy2" SIOCGIFFLAGS: No such device Device "dummy3" does not exist. usage: ifdown <device name> Device "dummy3" does not exist. Cannot find device "dummy3" SIOCGIFFLAGS: No such device Any ideas on how to get this working would be greatly appreciated. - David -- Fedora-xen mailing list Fedora-xen@redhat.com https://www.redhat.com/mailman/listinfo/fedora-xen
Per, Ok. I was hoping there was something realatively straightforward to change. I''ve also been working on a new approach to use KVM with Fedora 7. I''ve been able to use Revisor to create a custom Install CD image and used that with virt-manager to create a virtual machine. However, I''ve run into a wall with virt-manager''s insistance on using only private IP address space. Your suggestion to dump the result to xml and modify that seems like it might work, except I can''t get virsh to work: # virsh dumpxml test_vm error: failed to connect to the hypervisor error: no valid connection Where test_vm is the name of my virtual machine in the Name column of the virt-manager GUI. It doesn''t make a difference whether or not the virtual machine is running at the time. - David On 8/22/07, Per Hjartoy <per@actius.com> wrote:> David, The configuration has changed quite a bit between F5 and F7. I''ve > found that there is not much you can do unless you go "virsh dumpxml ...". > In my experimentation, I made the most progress when I defined the device > with dummy data in the GUI version. I could then dump the result out and > make changes in the xml template I got. Finally I would reload by "virsh > define ...". Hope this helps. -- Per > > -----Original Message----- > From: fedora-xen-bounces@redhat.com [mailto:fedora-xen-bounces@redhat.com] > On Behalf Of David Mueller > Sent: Tuesday, August 21, 2007 10:07 AM > To: fedora-xen@redhat.com > Subject: [Fedora-xen] Fedora 7 - missing dummy interfaces > > I''m trying to move from Fedora Core 5 to Fedora 7. I''ve run into an > issue where the xen guests, which need several bridged interfaces > bound to dummy devices so they can communicate with each other > (similar to host-only networking with VMWare), won''t boot. I think > I''ve traced the problem to a lack of the dummy1, dummy2, and dummy3 > devices, and I''m not sure how to fix it. What I''ll describe below is > the setup that worked in Fedora Core 5: > > I used the following custom network-bridge script: > > #!/bin/sh > dir=$(dirname "$0") > "$dir/network-bridge" "$@" vifnum=0 netdev=dummy0 > "$dir/network-bridge" "$@" vifnum=1 netdev=dummy1 > "$dir/network-bridge" "$@" vifnum=2 netdev=dummy2 > "$dir/network-bridge" "$@" vifnum=3 netdev=dummy3 > > And the following line was used in the .xen files passed to xm create: > > vif = [''mac=2E:22:22:22:22:20,bridge=xenbr1'', > ''mac=2E:22:22:22:22:21,bridge=xenbr2'', > ''mac=2E:22:22:22:22:22,bridge=xenbr3''] > > Running the script resulted in the following: > > Using config file "./guest1.xen". > Error: Device 0 (vif) could not be connected. Could not find bridge > device xenbr1 > > If I try to run the custom network-bridge script manually passing the > "start" command I get several errors: > > Device "dummy1" does not exist. > usage: ifdown <device name> > Device "dummy1" does not exist. > Cannot find device "dummy1" > SIOCGIFFLAGS: No such device > Device "dummy2" does not exist. > usage: ifdown <device name> > Device "dummy2" does not exist. > Cannot find device "dummy2" > SIOCGIFFLAGS: No such device > Device "dummy3" does not exist. > usage: ifdown <device name> > Device "dummy3" does not exist. > Cannot find device "dummy3" > SIOCGIFFLAGS: No such device > > Any ideas on how to get this working would be greatly appreciated. > > - David > > -- > Fedora-xen mailing list > Fedora-xen@redhat.com > https://www.redhat.com/mailman/listinfo/fedora-xen > > >
Is "test_vm" your guest? If not, that is what you need to use as an argument to virsh and not the Domain-0 (vm container). -- Per -----Original Message----- From: David Mueller [mailto:dsm717@gmail.com] Sent: Wednesday, August 22, 2007 2:03 PM To: Per Hjartoy Cc: fedora-xen@redhat.com Subject: Re: [Fedora-xen] Fedora 7 - missing dummy interfaces Per, Ok. I was hoping there was something realatively straightforward to change. I''ve also been working on a new approach to use KVM with Fedora 7. I''ve been able to use Revisor to create a custom Install CD image and used that with virt-manager to create a virtual machine. However, I''ve run into a wall with virt-manager''s insistance on using only private IP address space. Your suggestion to dump the result to xml and modify that seems like it might work, except I can''t get virsh to work: # virsh dumpxml test_vm error: failed to connect to the hypervisor error: no valid connection Where test_vm is the name of my virtual machine in the Name column of the virt-manager GUI. It doesn''t make a difference whether or not the virtual machine is running at the time. - David On 8/22/07, Per Hjartoy <per@actius.com> wrote:> David, The configuration has changed quite a bit between F5 and F7. I''ve > found that there is not much you can do unless you go "virsh dumpxml ...". > In my experimentation, I made the most progress when I defined the device > with dummy data in the GUI version. I could then dump the result out and > make changes in the xml template I got. Finally I would reload by "virsh > define ...". Hope this helps. -- Per > > -----Original Message----- > From: fedora-xen-bounces@redhat.com [mailto:fedora-xen-bounces@redhat.com] > On Behalf Of David Mueller > Sent: Tuesday, August 21, 2007 10:07 AM > To: fedora-xen@redhat.com > Subject: [Fedora-xen] Fedora 7 - missing dummy interfaces > > I''m trying to move from Fedora Core 5 to Fedora 7. I''ve run into an > issue where the xen guests, which need several bridged interfaces > bound to dummy devices so they can communicate with each other > (similar to host-only networking with VMWare), won''t boot. I think > I''ve traced the problem to a lack of the dummy1, dummy2, and dummy3 > devices, and I''m not sure how to fix it. What I''ll describe below is > the setup that worked in Fedora Core 5: > > I used the following custom network-bridge script: > > #!/bin/sh > dir=$(dirname "$0") > "$dir/network-bridge" "$@" vifnum=0 netdev=dummy0 > "$dir/network-bridge" "$@" vifnum=1 netdev=dummy1 > "$dir/network-bridge" "$@" vifnum=2 netdev=dummy2 > "$dir/network-bridge" "$@" vifnum=3 netdev=dummy3 > > And the following line was used in the .xen files passed to xm create: > > vif = [''mac=2E:22:22:22:22:20,bridge=xenbr1'', > ''mac=2E:22:22:22:22:21,bridge=xenbr2'', > ''mac=2E:22:22:22:22:22,bridge=xenbr3''] > > Running the script resulted in the following: > > Using config file "./guest1.xen". > Error: Device 0 (vif) could not be connected. Could not find bridge > device xenbr1 > > If I try to run the custom network-bridge script manually passing the > "start" command I get several errors: > > Device "dummy1" does not exist. > usage: ifdown <device name> > Device "dummy1" does not exist. > Cannot find device "dummy1" > SIOCGIFFLAGS: No such device > Device "dummy2" does not exist. > usage: ifdown <device name> > Device "dummy2" does not exist. > Cannot find device "dummy2" > SIOCGIFFLAGS: No such device > Device "dummy3" does not exist. > usage: ifdown <device name> > Device "dummy3" does not exist. > Cannot find device "dummy3" > SIOCGIFFLAGS: No such device > > Any ideas on how to get this working would be greatly appreciated. > > - David > > -- > Fedora-xen mailing list > Fedora-xen@redhat.com > https://www.redhat.com/mailman/listinfo/fedora-xen > > >
Yes, test_vm is the name of the virtual machine in the second column
of the list in virt-manager. If I just try to launch virsh I get an
error too:
# virsh
error: failed to connect to the hypervisor
Welcome to virsh, the virtualization interactive terminal.
Type: ''help'' for help with commands
''quit'' to quit
virsh #
Note that I am using the standard kernel 2.6.22.1-41.fc7 and not a
xen0 kernel, so it should be using KVM rather than xen. The computer
has an Intel Core Duo processor with the Virtualization instructions
enabled in BIOS -- the guest VM runs just fine.
- David
On 8/22/07, Per Hjartoy <per@actius.com> wrote:> Is "test_vm" your guest? If not, that is what you need to use as
an
> argument to virsh and not the Domain-0 (vm container). -- Per
>
> -----Original Message-----
> From: David Mueller [mailto:dsm717@gmail.com]
> Sent: Wednesday, August 22, 2007 2:03 PM
> To: Per Hjartoy
> Cc: fedora-xen@redhat.com
> Subject: Re: [Fedora-xen] Fedora 7 - missing dummy interfaces
>
> Per,
>
> Ok. I was hoping there was something realatively straightforward to
change.
>
> I''ve also been working on a new approach to use KVM with Fedora 7.
> I''ve been able to use Revisor to create a custom Install CD image
and
> used that with virt-manager to create a virtual machine. However,
> I''ve run into a wall with virt-manager''s insistance on
using only
> private IP address space. Your suggestion to dump the result to xml
> and modify that seems like it might work, except I can''t get virsh
to
> work:
>
> # virsh dumpxml test_vm
> error: failed to connect to the hypervisor
> error: no valid connection
>
> Where test_vm is the name of my virtual machine in the Name column of
> the virt-manager GUI. It doesn''t make a difference whether or not
the
> virtual machine is running at the time.
>
> - David
>
> On 8/22/07, Per Hjartoy <per@actius.com> wrote:
> > David, The configuration has changed quite a bit between F5 and F7.
I''ve
> > found that there is not much you can do unless you go "virsh
dumpxml ...".
> > In my experimentation, I made the most progress when I defined the
device
> > with dummy data in the GUI version. I could then dump the result out
and
> > make changes in the xml template I got. Finally I would reload by
"virsh
> > define ...". Hope this helps. -- Per
> >
> > -----Original Message-----
> > From: fedora-xen-bounces@redhat.com
[mailto:fedora-xen-bounces@redhat.com]
> > On Behalf Of David Mueller
> > Sent: Tuesday, August 21, 2007 10:07 AM
> > To: fedora-xen@redhat.com
> > Subject: [Fedora-xen] Fedora 7 - missing dummy interfaces
> >
> > I''m trying to move from Fedora Core 5 to Fedora 7.
I''ve run into an
> > issue where the xen guests, which need several bridged interfaces
> > bound to dummy devices so they can communicate with each other
> > (similar to host-only networking with VMWare), won''t boot. I
think
> > I''ve traced the problem to a lack of the dummy1, dummy2, and
dummy3
> > devices, and I''m not sure how to fix it. What I''ll
describe below is
> > the setup that worked in Fedora Core 5:
> >
> > I used the following custom network-bridge script:
> >
> > #!/bin/sh
> > dir=$(dirname "$0")
> > "$dir/network-bridge" "$@" vifnum=0 netdev=dummy0
> > "$dir/network-bridge" "$@" vifnum=1 netdev=dummy1
> > "$dir/network-bridge" "$@" vifnum=2 netdev=dummy2
> > "$dir/network-bridge" "$@" vifnum=3 netdev=dummy3
> >
> > And the following line was used in the .xen files passed to xm create:
> >
> > vif = [''mac=2E:22:22:22:22:20,bridge=xenbr1'',
> > ''mac=2E:22:22:22:22:21,bridge=xenbr2'',
> > ''mac=2E:22:22:22:22:22,bridge=xenbr3'']
> >
> > Running the script resulted in the following:
> >
> > Using config file "./guest1.xen".
> > Error: Device 0 (vif) could not be connected. Could not find bridge
> > device xenbr1
> >
> > If I try to run the custom network-bridge script manually passing the
> > "start" command I get several errors:
> >
> > Device "dummy1" does not exist.
> > usage: ifdown <device name>
> > Device "dummy1" does not exist.
> > Cannot find device "dummy1"
> > SIOCGIFFLAGS: No such device
> > Device "dummy2" does not exist.
> > usage: ifdown <device name>
> > Device "dummy2" does not exist.
> > Cannot find device "dummy2"
> > SIOCGIFFLAGS: No such device
> > Device "dummy3" does not exist.
> > usage: ifdown <device name>
> > Device "dummy3" does not exist.
> > Cannot find device "dummy3"
> > SIOCGIFFLAGS: No such device
> >
> > Any ideas on how to get this working would be greatly appreciated.
> >
> > - David
> >
> > --
> > Fedora-xen mailing list
> > Fedora-xen@redhat.com
> > https://www.redhat.com/mailman/listinfo/fedora-xen
> >
> >
> >
>
>
>
Daniel P. Berrange
2007-Aug-22 22:09 UTC
Re: [Fedora-xen] Fedora 7 - missing dummy interfaces
On Wed, Aug 22, 2007 at 02:02:48PM -0700, David Mueller wrote:> Per, > > Ok. I was hoping there was something realatively straightforward to change. > > I''ve also been working on a new approach to use KVM with Fedora 7. > I''ve been able to use Revisor to create a custom Install CD image and > used that with virt-manager to create a virtual machine. However, > I''ve run into a wall with virt-manager''s insistance on using only > private IP address space. Your suggestion to dump the result to xml > and modify that seems like it might work, except I can''t get virsh to > work:virt-manager will happily use bridging for KVM guests if you setup your host so that its physical interfaces are part of a bridge. There''s a guide here: http://watzmann.net/blog/index.php/2007/04/27/networking_with_kvm_and_libvirt Basically we recommend bridging (configured in /etc/sysconfig/network-scripts) for machines with permanent wired ethernet connections. We recommend the NAT based virtual networking for machines with dynamic connections (eg Networkmanager) or using wireless / VPN / dialup. Dan. -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|
Daniel P. Berrange
2007-Aug-23 02:16 UTC
Re: [Fedora-xen] Fedora 7 - missing dummy interfaces
On Wed, Aug 22, 2007 at 03:04:54PM -0700, David Mueller wrote:> Yes, test_vm is the name of the virtual machine in the second column > of the list in virt-manager. If I just try to launch virsh I get an > error too: > > # virsh > error: failed to connect to the hypervisor > Welcome to virsh, the virtualization interactive terminal. > > Type: ''help'' for help with commands > ''quit'' to quit > > virsh # > > Note that I am using the standard kernel 2.6.22.1-41.fc7 and not a > xen0 kernel, so it should be using KVM rather than xen.When connecting with virsh, you need to provide a URI to override the default use of Xen. For QEMU/KVM use ''virsh --connect qemu:///system'' Regards, Dan. -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|
Daniel P. Berrange wrote:> When connecting with virsh, you need to provide a URI to override the > default use of Xen. For QEMU/KVM use ''virsh --connect qemu:///system''Thanks, that fixed the issue. I''ll look at the rest tomorrow and see If I can get something working. Perhaps a future feature could attempt to detect whether or not a xen kernel is running and if not, connect to QEMU/KVM instead? - David
On Wed, 2007-08-22 at 19:24 -0700, David Mueller wrote:> Daniel P. Berrange wrote: > > When connecting with virsh, you need to provide a URI to override the > > default use of Xen. For QEMU/KVM use ''virsh --connect qemu:///system'' > > Thanks, that fixed the issue. I''ll look at the rest tomorrow and see If > I can get something working. Perhaps a future feature could attempt to > detect whether or not a xen kernel is running and if not, connect to > QEMU/KVM instead? >You could always set the VIRSH_DEFAULT_CONNECT_URI environmental variable. VIRSH_DEFAULT_CONNECT_URI=qemu:///system> - David > > -- > Fedora-xen mailing list > Fedora-xen@redhat.com > https://www.redhat.com/mailman/listinfo/fedora-xen
I''m making progress, but I''ve run into a couple issues.
I''ve created
the following three xml files for the three networks (I''m only
worrying about one of the virtual machines now; once I get it working
I should be able to replicate it for the other):
<network>
<name>emu0</name>
<uuid>8d18febd-e295-4e67-9373-a8b2a5855e60</uuid>
<bridge name=''emubr0'' stp=''on''
forwardDelay=''0'' />
<ip address=''128.10.0.253''
netmask=''255.255.255.0'' />
</network>
<network>
<name>emu2</name>
<uuid>8d18febd-e295-4e67-9373-a8b2a5855e62</uuid>
<bridge name=''emubr2'' stp=''on''
forwardDelay=''0'' />
</network>
<network>
<name>emu3</name>
<uuid>8d18febd-e295-4e67-9373-a8b2a5855e63</uuid>
<bridge name=''emubr3'' stp=''on''
forwardDelay=''0'' />
</network>
emu2 and emu3 networks I didn''t assign an IP address to since the host
doesn''t need one; they will be exclusively for VM to VM communcation.
emu0 (and in the future emu1) will be used for communication between
host and VM -- each will be a different subnet.
And here is the xml dump of the VM, taken while its running:
<domain type=''kvm'' id=''5''>
<name>emu0</name>
<uuid>9b517b2c-b315-62fc-626b-2525e2576217</uuid>
<memory>262144</memory>
<currentMemory>262144</currentMemory>
<vcpu>1</vcpu>
<os>
<type>hvm</type>
<boot dev=''hd''/>
</os>
<features>
<acpi/>
</features>
<clock offset=''utc''/>
<on_poweroff>destroy</on_poweroff>
<on_reboot>restart</on_reboot>
<on_crash>destroy</on_crash>
<devices>
<emulator>/usr/bin/qemu-kvm</emulator>
<disk type=''file'' device=''disk''>
<source file=''/home/david/emu/kvm/f7_emu0.img''/>
<target dev=''hda''/>
</disk>
<interface type=''network''>
<mac address=''1e:11:11:11:11:10''/>
<source network=''emu0''/>
<target dev=''vnet0''/>
</interface>
<interface type=''network''>
<mac address=''1e:11:11:11:11:11''/>
<source network=''emu2''/>
<target dev=''vnet1''/>
</interface>
<interface type=''network''>
<mac address=''1e:11:11:11:11:12''/>
<source network=''emu3''/>
<target dev=''vnet2''/>
</interface>
<input type=''mouse'' bus=''ps2''/>
<graphics type=''vnc'' port=''5900''
listen=''127.0.0.1''/>
</devices>
</domain>
The first problem is that I can''t get the guest VM to communicate with
the host. I assign an address to the interface with this command:
# /sbin/ifconfig eth0 128.10.0.1/24 up
Then I attempt to ping the host, but get an error message Destinamtion
Host Unreachable. The same happens if I try to ping the guest from
the host.
I did notice an additional oddity. While the three interfaces all
show their correct MAC addresses in the Hardware tab of the Virtual
Machine Details tab, if I run ifconfig within the guest, all three
show the same address, 1E:11:11:11:11:12 (which should be eth2''s MAC
address).
- David
On 8/22/07, Daniel P. Berrange <berrange@redhat.com>
wrote:> On Wed, Aug 22, 2007 at 07:04:55PM -0700, David Mueller wrote:
> > On 8/22/07 Daniel P. Berrange wrote:
> > >virt-manager will happily use bridging for KVM guests if you setup
> > >your host so that its physical interfaces are part of a bridge.
> >
> > Actually, that''s not quite what I need. The final setup will
have two
> > guest machines, each with an eth0 to communicate with the Domain 0
host.
> > Each also will have a virtual eth1 and and eth2 that is used for
> > communication between the two VMs. The virtual machines need to
> > commuincate only with each other and the host; they don''t
actually need
> > to connect to the network. Because this will run in an isolated
network
> > not connected to the Internet, we''re not restricted to
private address
> > space and it would be much more work to remap all the addresses we use
> > to be only in private address space.
>
> Ok, that makes sense. virt-manager restricts you to only allow private
> address spaces - we could relax that to a warning. In the mean time you
> could use ''virsh net-define'' to setup a virtual network
using a public
> IP address range, and virsh doesn''t enforce addresing policy. Use
the
> take a look at /usr/share/libvirt for an example XML file you can tweak.
> For an isolated network, simply remove the <forward/> tag.
>
> Dan.
> --
> |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496
> -=|
> |=- Perl modules: http://search.cpan.org/~danberr/
> -=|
> |=- Projects: http://freshmeat.net/~danielpb/
> -=|
> |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505
> -=|
>
Daniel P. Berrange
2007-Aug-23 23:22 UTC
Re: [Fedora-xen] Fedora 7 - missing dummy interfaces
On Thu, Aug 23, 2007 at 04:17:55PM -0700, David Mueller wrote:> I''m making progress, but I''ve run into a couple issues. I''ve created > the following three xml files for the three networks (I''m only > worrying about one of the virtual machines now; once I get it working > I should be able to replicate it for the other): > > <network> > <name>emu0</name> > <uuid>8d18febd-e295-4e67-9373-a8b2a5855e60</uuid> > <bridge name=''emubr0'' stp=''on'' forwardDelay=''0'' /> > <ip address=''128.10.0.253'' netmask=''255.255.255.0'' /> > </network> > > <network> > <name>emu2</name> > <uuid>8d18febd-e295-4e67-9373-a8b2a5855e62</uuid> > <bridge name=''emubr2'' stp=''on'' forwardDelay=''0'' /> > </network> > > <network> > <name>emu3</name> > <uuid>8d18febd-e295-4e67-9373-a8b2a5855e63</uuid> > <bridge name=''emubr3'' stp=''on'' forwardDelay=''0'' /> > </network> > > emu2 and emu3 networks I didn''t assign an IP address to since the host > doesn''t need one; they will be exclusively for VM to VM communcation. > emu0 (and in the future emu1) will be used for communication between > host and VM -- each will be a different subnet. > > > The first problem is that I can''t get the guest VM to communicate with > the host. I assign an address to the interface with this command: > > # /sbin/ifconfig eth0 128.10.0.1/24 up > > Then I attempt to ping the host, but get an error message Destinamtion > Host Unreachable. The same happens if I try to ping the guest from > the host.I suspect this is due to the next problem you mention...> I did notice an additional oddity. While the three interfaces all > show their correct MAC addresses in the Hardware tab of the Virtual > Machine Details tab, if I run ifconfig within the guest, all three > show the same address, 1E:11:11:11:11:12 (which should be eth2''s MAC > address).Is this with QEMU, or KVM ? Ie is it running /usr/bin/qemu-kvm, or the regular qemu binary. There was a bug where all NICs got the same MAC address https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=247641 IIRC we''ve only fixed KVM so far. Dan. -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|
/usr/bin/qemu-kvm While I''m at it, anyone know how to keep Fedora 7 from trying to do DHCP? It''s slowing down the guest VM boot time considerably as it has to wait for DHCP timeout to proceed. Three times since there are three interfaces. The interfaces will have their proper IP addresses assigned later. - David On 8/23/07, Daniel P. Berrange <berrange@redhat.com> wrote:> On Thu, Aug 23, 2007 at 04:17:55PM -0700, David Mueller wrote: > > I''m making progress, but I''ve run into a couple issues. I''ve created > > the following three xml files for the three networks (I''m only > > worrying about one of the virtual machines now; once I get it working > > I should be able to replicate it for the other): > > > > <network> > > <name>emu0</name> > > <uuid>8d18febd-e295-4e67-9373-a8b2a5855e60</uuid> > > <bridge name=''emubr0'' stp=''on'' forwardDelay=''0'' /> > > <ip address=''128.10.0.253'' netmask=''255.255.255.0'' /> > > </network> > > > > <network> > > <name>emu2</name> > > <uuid>8d18febd-e295-4e67-9373-a8b2a5855e62</uuid> > > <bridge name=''emubr2'' stp=''on'' forwardDelay=''0'' /> > > </network> > > > > <network> > > <name>emu3</name> > > <uuid>8d18febd-e295-4e67-9373-a8b2a5855e63</uuid> > > <bridge name=''emubr3'' stp=''on'' forwardDelay=''0'' /> > > </network> > > > > emu2 and emu3 networks I didn''t assign an IP address to since the host > > doesn''t need one; they will be exclusively for VM to VM communcation. > > emu0 (and in the future emu1) will be used for communication between > > host and VM -- each will be a different subnet. > > > > > > The first problem is that I can''t get the guest VM to communicate with > > the host. I assign an address to the interface with this command: > > > > # /sbin/ifconfig eth0 128.10.0.1/24 up > > > > Then I attempt to ping the host, but get an error message Destinamtion > > Host Unreachable. The same happens if I try to ping the guest from > > the host. > > I suspect this is due to the next problem you mention... > > > I did notice an additional oddity. While the three interfaces all > > show their correct MAC addresses in the Hardware tab of the Virtual > > Machine Details tab, if I run ifconfig within the guest, all three > > show the same address, 1E:11:11:11:11:12 (which should be eth2''s MAC > > address). > > Is this with QEMU, or KVM ? Ie is it running /usr/bin/qemu-kvm, or the > regular qemu binary. There was a bug where all NICs got the same MAC > address > > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=247641 > > IIRC we''ve only fixed KVM so far. > > Dan. > -- > |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 > -=| > |=- Perl modules: http://search.cpan.org/~danberr/ > -=| > |=- Projects: http://freshmeat.net/~danielpb/ > -=| > |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 > -=| >
On 24/08/07, David Mueller <dsm717@gmail.com> wrote:> /usr/bin/qemu-kvm > > While I''m at it, anyone know how to keep Fedora 7 from trying to do > DHCP? It''s slowing down the guest VM boot time considerably as it has > to wait for DHCP timeout to proceed. Three times since there are > three interfaces. The interfaces will have their proper IP addresses > assigned later./etc/sysconfig/networking/devices/ifcfg-eth[012] BOOTPROTO=NONE ?