Open Stack

On 04/23/2015 12:13 AM, Mike Smith wrote:
>    No shared services between them.   Some of the separation is due to the kind of instances they contain (“Dev/Test" vs “Prod” for example), but it is largely due to the location diversity and the desire to not require everything to be upgraded at once.
This is an interesting point;  if a mix-and-match approach were takien, 
it would require a lot of internal maturing of the OpenStack 
interfaces.  We would end up with case of a Kilo Glance being called by 
a Liberty Nova, for example.  That might not be a bad thing, as it would 
require a degree of discovery and flexibility we don't currently emphasize.

The bigger issue, I suspect, is the message queue.  Getting that to API 
stability would be a huge effort, I suspect.

I am less interested in what people are actually doing than in the 
things that you would want to be doing but cannot due to limitations in 
the authorization and sharing assumptions that make up OpenStack.  The 
Cinder client that runs on the compute node is "owned" buy the cinder 
server, so to speak.  This means that two Cinders with different access 
control levels cannot be shared by the same cloud.