Open Stack

On 11/17/2016 04:58 AM, Sławek Kapłoński wrote:
> Hello,
>
> Ok. So it's different case. I was thinking only about issues with
> quotas.
> For me this which You described is more related to scheduler/placement
> API(?).
> Do You think that it should be taken in same specs?

Given that we need a spec to cover any changes to publicly-visible behaviour, I 
think it would make sense to lump together any changes around what happens when 
you try to boot multiple instances and only some of them can be booted.

I think it's basically irrelevant what the reasons are why some couldn't be 
booted--as long as the requested "min-count" instances can be booted 
successfully then they should proceed to boot and the request should count as 
successful.

And as I said before, if at least "min-count" instances can be booted 
successfully then the remaining instances should not go into error state but 
rather should just be transparently deleted as though they never existed.

Chris