On Wed, 2011-06-01 at 12:27 -0400, Scott Seago wrote:
On 06/01/2011 11:18 AM, Mark McLoughlin wrote:
> On Fri, 2011-05-27 at 12:50 -0400, Hugh Brock wrote:
>> On Fri, May 27, 2011 at 05:20:51PM +0100, Mark McLoughlin wrote:
>>> On Thu, 2011-05-26 at 23:49 -0400, Scott Seago wrote:
>>>> On 05/26/2011 10:57 AM, Mark McLoughlin wrote:
>>>> We should clarify at what levels we store quota. We've already
>>>> implemented instance quota in conductor -- i.e. how many instances can
>>>> be run at once by various measures:
>>>> 1) user quota: how many instances an end user can run at once,
>>>> regardless of pool or provider account
>>>> 2) pool quota: how many instances can run in a pool, regardless of
>>>> running user or provider account
>>>> 3) environment/pool family quota: how many instances can run in a
>>>> particular environment, regardless of pool, running user, or provider
>>>> 4) provider account quota: how many instances can run in a provider
>>>> account, regardless of user, pool, or environment.
>>> Well, I don't know what the rationale for those quotas was, so I
>>> really comment.
>>> To me, quota is only really important as a replacement for billing.
>>> Anything an admin might want to set up billing for, they will need
>>> quotas for in the absence of billing.
>> I think that's correct, with the exception of the pool quota and the
>> pool family/environment quota. The only reason that quota exists is to
>> keep one environment from DOS-ing the provider accounts that back it,
>> in the event another environment is also allowed to access those
> Apologies for getting off-topic a little, but ...
> Say you have two provider accounts - one for RHEV-M and one for EC2.
> Your RHEV-M install is fairly small and you can't handle more than 20
> instances. Your EC2 account can handle an unlimited number of instances.
Ahh -- that's what the provider account quota is for -- at least at the
aggregate level. So in addition to whatever other quotas you want to
enforce below, in this case you'd also want to set a RHEV account quota
to 20 so condor won't try to schedule more than 20 there.
Hmm, no. See this bit:
> You want to allow some dev workload on RHEV-M, but not at the
> forcing all production workload onto EC2.
A provider quota account doesn't prevent the dev workload from being
DOSed from RHEV-M onto EC2.
You would need to be able to set a quota on how many instances a pool
family can start on a provider account.
> That's the kind of case you mean, right?
> With your description of a pool family quota, you'd need to limit the
> dev pool family to e.g. 10 instances in order to guarantee that there is
> always at least 10 instances available for the production pool family.
> That seems like a unnatural constraint, since you have unlimited
> resources on the EC2 account.
I don't think you'd need to limit pool family in this way -- unless
there's some requirement that each pool family schedule equally on both
providers (which would be yet another policy question), you could simply
set the pool family quotas as high as you'd like. Now if it violates
some requirement for 'dev' to use up rhev and have all of prod in ec2,
The requirement here is for ec2 to just be used for spillover. And you
want to avoid production workload spilling over.
Yet you do want to allow some dev workload on RHEV.
then some of the "combined" quota types you mention below
created -- or even a non-quota policy -- i.e. "prefer the provider with
the most available quota before the one that's almost full"
That's actually the opposite of the case I'm describing - you want to
prefer RHEV-M with its smaller quota over EC2
> I would have imagined something like "the dev pool family
can only have
> 5 instances on RHEV-M, but can have an unlimited number of EC2
Sure, that would make sense too if we want to define various "combined"
or "composite" quota types. In other words, right now we have 4 types of
3) Pool Family/Environment
4) Provider account
The above would be a fifth one:
But we could add 10 more by combining the basic types in more ways, so
we really need a clear set of use cases to help define where we'll need
to restrict instance usage.
And clear use cases for the "uncombined quotas" - e.g. if we said that
the UI cannot cope with more than 4 quota types without frying the
user's brain, and we decide the need an per-env-and-account quota ...
which of the other quotas would we drop?