Open Stack

Sylvain, as I understand in BP description, Christian is about not exactly
reserving tenants itself like we actually do with VMs/hosts - it's just
naming for that. I think he is about two moments:

1) mark some tenants as "needed to be reserved" - speaking about resources
assigned to it
2) reserve these resources via Climate (VMs for first approximation)

I suppose Christian is speaking now about hacking tenants creation process
to mark them as "needed to be reserved" (1st step).

Christian, correct me if I'm wrong, please
Waiting for your comments


On Thu, Feb 20, 2014 at 10:06 PM, Sylvain Bauza <sylvain.bauza at gmail.com>wrote:

> Hi Christian,
>
> 2014-02-20 18:10 GMT+01:00 Martinez, Christian <
> christian.martinez at intel.com>:
>
>   Hello all,
>>
>> I'm working in the following BP:
>> https://blueprints.launchpad.net/climate/+spec/tenant-reservation-concept,
>> in which the idea is to have the possibility to create "special" tenants
>> that have a lease for all of its associated resources.
>>
>>
>>
>> The BP is in discussing phase and we were having conversations on IRC
>> about what approach should we follow.
>>
>>
>>
>
> Before speaking about implementation,  I would definitely know the
> usecases you want to design.
> What kind of resources do you want to provision using Climate ? The basic
> thing is, what is the rationale thinking about hooking tenant creation ?
> Could you please be more explicit ?
>
> At the tenant creation, Climate wouldn't have no information in terms of
> calculating the resources asked, because the resources wouldn't have been
> allocated before. So, generating a lease on top of this would be like a
> non-formal contract in between Climate and the user, accounting nothing.
>
> The main reason behind Climate is to provide SLAs for either user requests
> or projects requests, meaning that's duty of Climate to guarantee that the
> desired associated resource with the lease will be created in the future.
> Speaking of Keystone, the Keystone objects are tenants, users or domains.
> In that case, if Climate would be hooking Keystone, that would say that
> Climate ensures that the cloud will have enough capacity for creating these
> resources in the future.
>
> IMHO, that's not worth implementing it.
>
>
>  First of all, we need to add some "parameters or flags" during the
>> tenant creation so we can know that the associated resources need to have a
>> lease. Does anyone know if Keystone has similar functionality to Nova in
>> relation with Hooks/API extensions (something like the stuff mentioned on
>> http://docs.openstack.org/developer/nova/devref/hooks.html ) ? My first
>> idea is to intercept the tenant creation call (as it's being done with
>> climate-nova) and use that information to associate a lease quota to the
>> resources assigned to that tenant.
>>
>>
>
> Keystone has no way to know which resources are associated within a
> tenant, see how the middleware authentication is done here [1]
> Regarding the BP, the motivation is to possibly 'leasify' all the VMs from
> one single tenant. IIRC, that should still be duty of Nova to handle that
> workflow and send the requests to Climate.
>
> -Sylvain
>
> [1] :
> http://docs.openstack.org/developer/keystone/middlewarearchitecture.html
>
>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>


-- 

Best regards,

Dina Belova

Software Engineer

Mirantis Inc.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140220/727e2215/attachment.html>