[nexus] Add stubs for snapshot APIs

[smklein]

First part of #735

[smklein]

FYI, there is a minor divergence from the WIP doc you sent me: this change allows users to add a "name + description" of the snapshot object, and refer to it by a project-scoped name.

This fits the existing scheme for all other objects exposed by the API currently.

[davepacheco]

Cool! What's the plan for automated tests for the Nexus parts (like the CRUD stuff)?

[smklein]

Cool! What's the plan for automated tests for the Nexus parts (like the CRUD stuff)?

I can't do this quite yet, as a Nexus-level integration test would require a "real DB implementation" of this stuff. However, when that arrives, that'll be up next.

I've added:

[ ] Add support in simulated sled agent / crucible
[ ] Add integration tests in Nexus

to both #734 and #735

[jmpesp]

I have nothing to add beyond the other comments here.

[benjaminleonard]

Are there any plans to add state?

GCP has a ready state, and AWS uses completed with a progress percentage.

Same question for images (and any others in that class of resource that might similarly have a ready state).

Also, the snapshot_id on a disk – I gather that's the snapshot used to create the disk, since it's a one-to-one relationship, not a list of a snapshots. And it's fair to see we'll be able to filter snapshots by source disk right?

[smklein]

@benjaminleonard We could definitely add something like that - as snapshots haven't been implemented yet, I'm not 100% how long we expect they'll take - but adding "state" indicating progress to image creation seems totally reasonable.

You're right, the snapshot_id field on disks is the "snapshot used to create the disk", but I wouldn't describe the relationship as "one-to-one". Each disk will be created from either zero or one snapshot, and a snapshot may be used to create many disks.

And it's fair to see we'll be able to filter snapshots by source disk right?

We definitely could - this would be for an operation like "see all the snapshots of this disk that exist"?

I think we'd want to add an index for this operation (currently, we only have an index to lookup snapshots by (project_id, name)) but that should be possible.

[benjaminleonard]

Great thanks for clarifying Sean!

[leftwo]

One little thing to add is that a "source snapshot" can and probably will go away. Once a disk has been created from a snapshot, that disk will have a life of its own and it's possible the original snapshot will be deleted.