Support for new Rackspace instance types and Cloud Block Storage boot volumes

[chef-boneyard]

Support for new Rackspace instance types and Cloud Block Storage boot volumes

[jjasghar]

Thanks for the report, I'll dig into it ASAP.

[vidluther]

Any insights? Is this something we would need to get changed at the account level?

[maxlinc]

@smashwilson any idea? You made the Fog changes referenced above.

[smashwilson]

@maxlinc I did indeed add boot-from-volume support to Fog, yes.

The newer flavors only support using a remote cloud block device as the boot volume.

I was surprised to discover this, but it's true. If you use any of the compute or memory flavors, the only available boot source is boot-from-volume. This is also reflected in the Rackspace control panel:

Only boot-from-volume available:

Selectable boot sources for general-purpose flavors:

This isn't a Fog bug - it's a consequence of the way that the compute and I/O flavors are configured on Rackspace's end. To support these flavors, we'll need the ability to specify boot-from-volume from knife-rackspace.

As a result, knife rackspace can't create these, since it seems the API call to create the server now requires the UUID of a cloud block storage volume prepared with an image.

That's one way to do it, although there is an easier route: if you give Fog an image ID as :boot_image_id, the API will automatically create a volume bootstrapped from the chosen image and use it to launch the server in a single call. :image_id must still be provided, but left blank:

s = compute.servers.create name: 'non-bfv', image_id: '', boot_image_id: ubuntu_1404, flavor_id: 'compute1-4'

(The :image_id bit there is confusing and could be improved, though; ideally we should set :image_id automatically if one of the boot-from-volume parameters are provided.)

I think the answer here is to give knife-rackspace the ability to pass :boot_volume_id or :boot_image_id to the server creation call. I could have a look into making a PR if you'd like me to, although this codebase would be new to me.

[jpunzel]

@tdooner Looks like your PR was denied by the CLA bot. I'd love to see these changes merged.

[vmihailenco]

Ping. Any news on this?

[martinb3]

#95 appears to be good now, passing the CLA check. I'd love to see it merged! 👍

[mattsmitton]

Bump, would love to have this!

[Paladin]

Just a comment, here:

Device Mapping is still invalid and there needs to be an indication that boot-volume-size is now required (or a workable default plugged in).

[martinb3]

@Paladin we should probably open this as a new issue, since this issue is a year old and closed. Could you open an issue and show us the command you're trying to run?