The information that libvirt virsh pool-list is returning is not completely correct: For ex:
virsh -c qemu:///system vol-list cimtest-diskpoolName Path ----------------------------------------- .X0-lock /tmp/.X0-lock cimtest.uuid /tmp/cimtest.uuid default-kvm-dimage /tmp/default-kvm-dimage tmpnsIv8q /tmp/tmpnsIv8q
The last pool information tmpnsIv8q does not seem to exist on the machine, hence when I ran the HostSystem/03_hs_to_settdefcap.py it failed with the following error:
-------------------------------------------------------------------------------------------------------------------------------
ERROR - 'KVM_SettingsDefineCapabilities' returned 0 RASD objects instead of 16 CIM_ERR_INVALID_CLASS: Linux_ComputerSystem CIM_ERR_FAILED: Unable to get volume information: cannot open volume '/tmp/tmpnsIv8q': No such file or directory -------------------------------------------------------------------------------------------------------------------------------
Are you running with recent sources? This issue should have been fixed with Richard's patch: "This patch removes the error throwed when volume info cannot be extracted in the disk template generation" rev: 842, changeset: 2f4943568299
The test case passed after I created the file /tmp/tmpnsIv8q.
Items written to /tmp aren't permanent. Programs often write temporary files here during execution, so we cannot count on all the files in /tmp to exist. Actually, we really shouldn't place our images in /tmp. We should place them in something like /var/spool/libvirt-cim? Some place more permanent, and a little more expected from a application. A good admin would make sure that a directory based pool (the kind of pool cimtest creates) would only be populated with images. So we should be doing the same with ours.
I am not sure if this is the limitation from libvirt side or if its a feature with libvirt. Anyways +1 for these changes.
-- Kaitlin Rupert IBM Linux Technology Center kaitlin@linux.vnet.ibm.com