Comment 37 for bug 1834875

@daniel-thewatkins, I'm not convinced that this bug is invalid for cloud-init. After reading through all of this again, I still don't understand, what guarantee there is that when `udevadm settle` is called, all relevant events have already been queued.

Copying udevadm over, and with that suppressing the error, suggests that maybe the event queue handling is spurious, but on the other hand, it might just be that previous versions were slower at something and the 10ms window discussed above is always exceeded because of that anyway.

I'm not saying there cannot also be a bug somewhere else. But if there is no specification that says there cannot under any circumstances be a race condition in what cloud-init is doing, then cloud-init should handle this more robustly.

I'm not an expert on that level, but somehow in a world of multi-core CPUs and fancy schedulers, invoking command line tools in a certain order does not seem to preclude the possibility of a race in how the event is submitted, routed and queued without there being an explicit locking mechanism.