Fedora 7 - missing dummy interfaces

[David Mueller]

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

[Per Hjartoy]

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(a)redhat.com [mailto:[e-mail]]
On Behalf Of David Mueller
Sent: Tuesday, August 21, 2007 10:07 AM
To: fedora-xen(a)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(a)redhat.com
https://www.redhat.com/mailman/listinfo/fedora-xen

[David Mueller]

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(a)actius.com&gt; 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(a)redhat.com [mailto:[e-mail]]
>  On Behalf Of David Mueller
>  Sent: Tuesday, August 21, 2007 10:07 AM
>  To: fedora-xen(a)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(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-xen
>
>
>

[Per Hjartoy]

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:[e-mail]] 
Sent: Wednesday, August 22, 2007 2:03 PM
To: Per Hjartoy
Cc: fedora-xen(a)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(a)actius.com&gt; 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(a)redhat.com [mailto:[e-mail]]
>  On Behalf Of David Mueller
>  Sent: Tuesday, August 21, 2007 10:07 AM
>  To: fedora-xen(a)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(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-xen
>
>
>

[David Mueller]

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(a)actius.com&gt; 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:[e-mail]]
>  Sent: Wednesday, August 22, 2007 2:03 PM
>  To: Per Hjartoy
>  Cc: fedora-xen(a)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(a)actius.com&gt; 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(a)redhat.com [mailto:[e-mail]]
>  > On Behalf Of David Mueller
>  > Sent: Tuesday, August 21, 2007 10:07 AM
>  > To: fedora-xen(a)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(a)redhat.com
>  > https://www.redhat.com/mailman/listinfo/fedora-xen
>  >
>  >
>  >
>
>
>

[Daniel P. Berrange]

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.  [phone] -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=|

[David Mueller]

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

[Andrew Cathrow]

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(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-xen

[Daniel P. Berrange]

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_lib... 

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.  [phone] -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=|