Comment 16 for bug 1936972

I've run into a similar problem: If a node was commissioned with a kernel that detects a particular network device, but the system is then deployed with a kernel that does not detect that device, the deployment fails with the same message as noted in the original bug report. I've encountered this most recently with a Redfish device on blubi (an Ericsson AB CRU 0201 server), but ISTR running into something similar with a "real" NIC in the past. I understand curtin complaining about this, but when the result is that the post-install reboot, and therefore the entire deployment, fails, it can be unclear to the user why the deployment failed -- MAAS logs show a successful installation followed by no output from the node. I had to use the system's remote KVM and watch the entire deployment to figure it out. I worked around it by deleting the Redfish device, which isn't important for us or MAAS; but if it were a real device, that solution would leave the device unconfigured, even when the node was deployed using a kernel that could use it. A more robust solution would enable the deployment to succeed, even if the device was inactive.