Currently introspection fails when a node has no disks. We need a new API/CLI flag to mark a node as diskless, to allow introspection on it to pass. We probably also need a global option to treat all nodes as diskless. Then we need Ironic inspection interface to provide this flag based on presence of a valid storage_interface field (to be introduced as part of BFV).
Upstream RFE: https://bugs.launchpad.net/ironic/+bug/1554243
patch merged upstream; https://review.openstack.org/#/c/326620/
Update on the upstream status: code merged, present in Ocata (inspector version 5.0) Changing status to ON_QA;
This was tested in a kvm virtual environment successfully. Steps: install osp 12 undercloud with several nodes (i.e 3 controller, 2 compute, 1 <spare>) On hypervisor, virsh detach-disk --domain <spare> --target vda --persistent openstack baremetal import --json ~/instackenv.json # contains openstack baremetal configure boot openstack baremetal introspection bulk start openstack baremetal list # look for the uuid of <spare> used for 0 disk openstack baremetal node show <uuid> | grep properties As expected result: The introspection succeeded now with one bare metal and property "u'local_gb': u'0'"
Additional testing to show that we can still detect and fail node with local_gb drive =0 by using device hints This will set the device hints to seek out this case. for i in `ironic node-list | awk '/None/ {print $2}'`; do ironic node-update $i add properties/root_device='{"size": "> 1"}';done follow with introspection
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHEA-2017:3462