I went through the test case and found to have the following problems with the test case:
1) The XML configuration information with the existing test case is not completely correct for *creating* a KVM/XenFV guest. It does not set the test_disk information correctly, but this does not matter when we just *define *the guest, but surely the XML config was not useful in *creating* the KVM/XenFV guest. The test_disk was set to 'xvdb' for all Xen/XenFV/KVM. 2) The network pool information that is used for creating the guest and the networkpool used in the rest of the test case are different. 3) The tc was using create_diskpool_file() function to create a DiskPool, this should have been create_diskpool_conf() which calls either creating of diskconf file or using the diskpool depending on the version of the libvirt on the machine. 4) Also, the networkpool information that was initialized using the pool_init_list() test case was getting overwritten in the case where there were more than one networkpool existed. And this could be one of the potential reason for the test case to be intermittently fail, since the information in the list would not have been set appropriately.
Yep, I agree here. I was seeing this fail when I had more than one network pool defined because the guest was using one pool but the test was comparing against a different pool. Thanks for looking at this test - looks like a lot of issues to fix. -- Kaitlin Rupert IBM Linux Technology Center kaitlin@linux.vnet.ibm.com