Open Stack

On 10 September 2013 08:04, Mike Spreitzer <mspreitz at us.ibm.com> wrote:

>
>> My vision for TripleO/undercloud and scale in the long term is:
>> - A fully redundant self-healing undercloud
>>   - (implies self hosting)
> ...
>
> Robert, what do you mean by "self hosting"?  If a cloud can self-host, why
> do we need two clouds (under and over)?

Running the baremetal scheduler and a regular scheduler is still not
pretty; the baremetal scheduler exact-matches on memory etc, vs a
subdividing scheduler which looks for 'fits in'. You can work around
this by running cells.

However, there is a very nice separation of concerns in having your
production cloud and your deployment cloud completely disconnected.
For instance, a configuration error is much less able to permit a
tenant of your production cloud deploying onto baremetal when they
shouldn't. Relatedly, being able to spin up multiple overclouds gives
you a great dev/CI story: test your overcloud by having a parallel
test overcloud sitting on the same undercloud substrate.

That said, I think it's feasible to support a single cloud for folk
that want it - and initially when we started this I wanted it, I just
no longer think that it's actually the ideal configuration :).

-Rob

-- 
Robert Collins <rbtcollins at hp.com>
Distinguished Technologist
HP Converged Cloud