Comment 86 for bug 1834875

Yes, this is my read on the issue as well. The trigger is related to the inotify watch that systemd-udevd puts on the disk. Something that might help that we could try per xnox's comment around use of flock.

if growpart were to flock /dev/sda (we need to sort out what flags are needed to prevent udev probe and rule execution) prior to running sgdisk, and not release this flock until after partx runs (to inform the kernel of the update) and a udevadm settle (and possibly a trigger of a CHANGE event).

This should prevent early reads of the modified but not update-to-date partition data in the kernel.

Lastly, another area of exploration is: why isn't a change event emitted when partx runs to update the kernel with new partition data?

If there is one, and the rules then generate the correct symlink; another approach for growpart is for it to block until the size of the new partition is correct. growpart calculates what the new size should be; so it could poll for this value thus blocking cloud-init's size check from running until growpart is confident that the kernel has updated the correct value (and the subsequent rules have completed execution)!