FlavorDiskTooSmall exception should be communicated to client

[Lars Kellogg-Stedman]

If you attempt to boot an image that is too large for a given flavor:

  $ nova boot --image fedora-21-cloud --flavor m1.tiny test0

The boot will of course fail, but the fault communicated to the client is:

  | fault | {"message": "Build of instance 05f6c907-61ed-4109-bc42-1c6cd1cfa663 aborted: Could not clean up failed build, not rescheduling", "code": 500, "created": "2015-03-28T18:49:55Z"} |

While in nova-compute.log is a more useful error:

  TRACE nova.compute.manager [instance: 05f6c907-61ed-4109-bc42-1c6cd1cfa663] FlavorDiskTooSmall: Flavor's disk is too

It seems as if nova should provide this information in the "fault" key, because it clearly communicates to someone attempting to boot an instance why that instance failed to boot.

[Davanum Srinivas (DIMS) (dims-v)]

Already fixed. Here's what i see in latest devstack.

| fault | {"message": "Build of instance 8eaee112-8fc8-4479-b989-18f127c3927e aborted: Flavor's disk is too small for requested image.", "code": 500, "details": " File \"/opt/stack/nova/nova/compute/manager.py\", line 2175, in _do_build_and_run_instance |