Better VM/FCM Resource naming and k8s node naming

[eponerine]

Currently, when a VM is created, it has a name that looks like:

AKSCLUSTERNAME-USAGE-UNIQUEID

Where...

• AKSCLUSTERNAME = duh
• USAGE = ControlPlane, NodePool, LoadBalancer
• UNIQUEID = some UID that isn't a GUID

However, in k8s itself, the nodes take the format:

moc-UNIQUEID

Where...

• UNIQUEID = L{UID}, W{UID} (depending if Linux or Windows)

Some suggestions:

• Unify the UNIQUEID used between Hyper-V, FCM, and k8s so that it doesn't require looking in 5 different spots to correlate a k8s node and its VM/FCM resource
• Update the k8s UNIQUEID prefix of L or W to also include if ControlPlane or NodePool (this helps differentiating in Az Monitor Node view)

I understand that hostname size is probably limited to 15 chars because of the Windows worker nodes, so maybe trim the UID by a character or something?

[Elektronenvolt]

I understand that hostname size is probably limited to 15 chars because of the Windows worker nodes, so maybe trim the UID by a character or something?

One issue I had been running in at a preview release half a year before GA: To be able to configure gMSA for Windows containers we had to domain join the Windows worker nodes. The hostname had been set to workload clustername + UID. To stay within the 15 chars limit, the workload cluster name was limited to 4 chars, otherwise a domain join failed because of hostname too long.

Meanwhile, a domain join ist not required anymore: gMSA - non domain joined host. Possibly the 15 char limitation is not an issue anymore...