Open Stack

Quick note - I want to keep this discussion a bit high-level and not to 
get into big implementation details. For everyone, please, let's agree 
in this thread on the direction and approach and we can start follow-up 
threads with bigger details of how to get those things done.

On 2013/13/12 12:04, Tzu-Mainn Chen wrote:
>> *VERSION 0*
>> ===========
>> Enable user to deploy OpenStack with the simpliest TripleO way, no
>> difference between hardware.
>>
>> Target:
>> - end of icehouse-2
>
> My impression was that some of these features required features to be developed in other
> OpenStack services - if so, should we call those out so that we can see if they'll be
> available in the icehouse-2 timeframe?
As for below listed features for v0 - it is the smallest set of what we 
have to have in the UI - if there is some delay in other services, we 
have to put attention there as well. But I don't think there is anything 
blocking us at the moment.

>> Features we need to get in:
>> - Enable manual nodes registration (Ironic)
>> - Get images available for user (Glance)
>
> Are we still providing the Heat template?  If so, are there image requirements that we
> need to take into account?
I am not aware of any special requirements, but I will let experts to 
answer here...

>
>> - Node roles (hardcode): Controller, Compute, Object Storage, Block Storage
>> - Design deployment (number of nodes per role)
>
> We're only allowing a single deployment, right?
Correct. For the whole Icehouse. I don't think we can get multiple 
deployments in time, there are much more important features.

>> - Deploy (Heat + Nova)
>
> What parameters are we passing in for deploy?  Is it limited to the # of nodes/role, or
> are we also passing in the image?
I think it is # nodes/role and image as well. Though images might be 
hardcoded for the very first iteration. Soon we should be able to let 
user assign images to roles.

> Do we also need the following?
>
> * unregister a node in Ironic
> * update a deployment (add or destroy instances)
> * destroy a deployment
> * view information about management node (instance?)
> * list nodes/instances by role
> * view deployment configuration
> * view status of deployment as it's being deployed
Some of that is part of above mentioned, some a bit later down the road 
(not far away though). We need all of that, but let's enable user to 
deploy first and we can add next features after we get something working 
then.

-- Jarda