[openstack-dev] [nova][scheduler] Instance Group Model and APIs - Updated document with an example request payload

Russell Bryant rbryant at redhat.com
Tue Oct 29 20:48:48 UTC 2013


On 10/29/2013 04:32 PM, Mike Spreitzer wrote:
> I should clarify my comment about invoking Heat to do the orchestration.
>  I think we have a choice between designing a 1-stage API vs a 2-stage
> API.  The 2-stage API goes like this: first the client defines the
> top-level group and everything inside it, then the client makes more
> calls to create the resources (with reference to the groups and
> policies).  In the 2-stage API, there is no need to use Heat for
> orchestration --- the client can orchestrate in any way it wants (BTW,
> we should eventually get around to talking about what happens if the
> client is the heat engine).  In the 1-stage API, the client just makes
> one call to define the top-level group and everything inside it; the
> implementation takes care of all the rest; in this style of API, I think
> it would be natural for the implementation to call Heat to do the
> orchestration.

Nova calling heat to orchestrate Nova seems fundamentally wrong.

Based on your description, the 2-stage bits belong in Nova, and the
1-stage part should just be talking to the Heat API directly, not Nova.
 Once Nova has instance groups and policies for those groups, Heat
should be updated to allow you to include those in your stacks.

-- 
Russell Bryant



More information about the OpenStack-dev mailing list