Comment 21 for bug 1902960

OK, I've managed to reproduce this (in a non-Juju launched VM). The ordering of these journal lines look suspicious to me:

Nov 09 17:41:51.091033 ubuntu systemd[1]: Starting udev Coldplug all Devices...
Nov 09 17:41:51.236309 ubuntu systemd[1]: Finished Load Kernel Modules.
Nov 09 17:41:51.363482 ubuntu systemd[1]: Finished udev Coldplug all Devices.

Because, you guessed it, hv_netvsc is shipped as a kernel module:

$ lsmod | grep hv_netvsc
hv_netvsc 81920 0

So my assumption is that udev coldplugging of the network device is happening before the driver is loaded, and so (unsurprisingly :) it doesn't find the driver.

I suspect that adding an `After=systemd-modules-load.service` to systemd-udev-trigger-service addresses the issue, but as this only reproduces on first boot this is difficult to test.

Given the above (and the fact that cloud-init doesn't run for another ~5s after these lines), I _think_ this is a systemd/kernel interface issue, not a cloud-init issue.