Code review comment for lp:~eday/nova/compute-abstraction

The stuff in irc was simply noticing that exception handling on image_get wasn't handled consistently. Giving the workers an api seems reasonable to me. I think it is easier conceptually than having the same class work both locally and remote.

How do you see this working for the volume? Code from volume_manager runs on the api host (create the db record), the compute host (discover the volume), and the volume host (create and export the volume). Do we have three separate classes? Or would VolumeAPI encompass all of the functions that are called by other workers (the api host and compute host code)?

Regardless, we need move the other workers code into this format ASAP.