Well, the providers should be creating <source network='something'>, so I need to check and make sure we're doing that.
I checked whether the provider adds the <source network='something'> and it does for KVM. I tried creating the KVM guest with two interface types:
1) Network Type :
It needs the following fields in the xml file: <interface type='network'> <mac address='11:22:33:aa:bb:cc'/> <source network='default-net'/> </interface>
We are handling this in the provider code.
2) Bridge Type : It needs the following fields in the xml file:
<interface type='bridge'> <mac address='11:22:33:aa:bb:cc'/> <source bridge='testbridge'/> </interface> The provider code in the xmlgen.c file does not support the KVM guest to be created with the "bridge" type.
Excellent find Deepti. Thanks for looking into this - I think this is an issue with the provider. You're correct - this is an issue in xmlgen.c What is happening here is the following: 1) Guest is created using virsh using network type bridge. 2) xen_net_to_xml() in xmlgen.c gets called instead of kvm_net_to_xml(). 3) Since xend is fine with using a bridge interface without the <source bridge=''/> tag defined, we don't specify one in the xmlgen code. QEMU borks when it attempts to create the guest with the new XML.
1) Also, I noticed that for creating the Xen guest we do not require the <source > field in the <interface > section for both Network and Bridge interface Types. But for the KVM guest its compulsory to have the <source > field in the <interface > section for both Network and Bridge interface Types. 2) Any specific reason for not supporting the KVM guest to be created for the bridge type interface, and supporting Xen guest with only bridge type ?
If you use the network interface type, the libvirt gets all the necessary network information from the virtual network. We don't have to worry about whether the bridge exists, etc.
3) Also, I did not understand the need for the script field in the interface section. We can create the Xen, KVM guests without this.
This is a good point. Actually, I thought there was some talk of having both Xen and KVM guests use the network interface, but I might be mistaken.
KR> Dan - should the provider return an error if the guest doesn't use KR> a network type interface?
Meaning an existing guest created outside the providers with a network type we don't support? I dunno, I'll have to think on that one.
Right - that's the issue here. We have a KVM guest with a bridge network type. The AddResource() call fails because it generates the new XML incorrectly. -- Kaitlin Rupert IBM Linux Technology Center kaitlin@linux.vnet.ibm.com