OK, so Ryan and I just kicked this back and forth a bit in a meeting. Looking at the failure, I believe that we successfully resize the disks/partitions, but when we look up the partition by partuuid to double-check what changed we fail. (Ryan's hypothesis is that the symlink is in the process of being recreated by udev when we go to check it.)
(The reason I believe we successfully resize the disk is that the traceback occurs after we've successfully run the growpart command which we can see in the cloud-init log and in dmesg.)
Access to a failing instance would be really helpful, so I can double-check our thinking here.
OK, so Ryan and I just kicked this back and forth a bit in a meeting. Looking at the failure, I believe that we successfully resize the disks/partitions, but when we look up the partition by partuuid to double-check what changed we fail. (Ryan's hypothesis is that the symlink is in the process of being recreated by udev when we go to check it.)
(The reason I believe we successfully resize the disk is that the traceback occurs after we've successfully run the growpart command which we can see in the cloud-init log and in dmesg.)
Access to a failing instance would be really helpful, so I can double-check our thinking here.