Comment 9 for bug 1776888

Hi Tone,

I tested the partlabel in the virtual machine can be generated immediately, but running docker on a virtual machine, the above phenomenon will appear.

Because my virtual machine is our company's cloud virtual machine, our workflow is to verify the deployment on the virtual machine first, and then deploy the cluster on the physical machine to verify the performance. I think this scenario should be very common, so I think we should use the most common deployment method. As you said, ceph supports partuuid and partlabel, but partuuid is a more common one.

I'm glad you have implemented partlabel's bluestore, but the process is somewhat different from that of the filestore, I think the filestore implementation is better, with less code and easier to read.

So I've made some optimizations based on your implementation,my patch is more similar to the way filestore handles disk information, and I refer to the initialization process of ceph-disk, I hope you can review it.

Thanks very much!^^