Microk8s CLI mayastor-pools command should use recommended way to refer to disks

[adamfowleruk]

Summary

The Mayastor documentation recommends using disk name references that persist between system reboots rather than references such as /dev/nvme1n1. Currently the Microk8s mayastor-pools add command only supports these names rather than the by-id names. This fails because specifying a full path such as --device /dev/disks/by-id/nvme-SomeDiscReference-54321 attempts to use the whole name as part of the pool name when it generates the DiskPool YAML, which is not a valid K8S name, and so fails.

What Should Happen Instead?

As a bare minimum, a sensible name in the YAML should be generated from the last part of the device path rather than full path.

Ideally this command should take a local name such as nvme1n1 and use linux syscalls to determine its by-id (E.g. via the udevadm command/API). It should then generate a sensible persistent name that is valid, and use the by-id name in the YAML's 'disks: ['NAME']' element.

Even better, modifying the mayastor-pools command to include something like add --thisnode --all-free-nvme or --all-free-sata, or even better a cluster wide version of this, would be much more user friendly. (One pool per disk)

Reproduction Steps

1. Enable the Mayastor addon
2. Attempt to run microk8s mayastor-pools add --node mynode --device /dev/disks/by-id/nvme-SOMEID
3. Observe that this command fails

Introspection Report

Unable to share - airgapped customer system. Was on Ubuntu 23.10.1 and microk8s v1.29.2-strict and that version of the Mayastor addon.

Can you suggest a fix?

As above - several options.

Are you interested in contributing with a fix?

Not sure where to start on a reliable way via Linux system calls if I'm honest. My udevadm and mount approach I'm using with sed feels a bit hacky.