Want better error on refusal to snapshot disk attached to stopped instance

I tried to create a snapshot of a disk attached to a stopped instance. This is not allowed, but the 503 response surprised me — I would expect a 400 with a more helpful message.

{
  "request_id": "f14e26d3-f122-4f12-b284-daf99c2386c3",
  "error_code": "ServiceNotAvailable",
  "message": "Service Unavailable"
}

I believe the error comes from here.

https://github.com/oxidecomputer/omicron/blob/915276d47096a9ce076453eb75aaa9de29477e80/nexus/src/app/sagas/snapshot_create.rs#L841-L851

From the Nexus log:

error_message_external = Service Unavailable
error_message_internal = saga ACTION error at node "disk_generation_number": disk is in state DiskState(Attached(33e6d4a4-6cc7-440f-8a9f-152d94136191))

Something odd might be going on, though, because in theory we are checking in the instance state before we even run the saga and giving precisely the 400 I would expect. Or maybe the problem is that we're hitting the None arm of the match and therefore not getting an error? I'll need to look into this further.

https://github.com/oxidecomputer/omicron/blob/915276d47096a9ce076453eb75aaa9de29477e80/nexus/src/app/snapshot.rs#L110-L129

Start with nexus/src/app/sagas/snapshot_create.rs around lines 841-851 and nexus/src/app/snapshot.rs around lines 110-129. Reproduce the snapshot request for a disk attached to a stopped instance, then trace the instance-state check and the None match arm against the logged saga error. Done means the request returns HTTP 400 with a helpful error instead of 503.