[openstack-dev] [nova][ceilometer] model for ceilo/nova interaction going forward

Mark McLoughlin markmc at redhat.com
Mon Nov 26 23:01:28 UTC 2012


On Mon, 2012-11-26 at 10:39 -0500, Doug Hellmann wrote:
> 
> 
> On Fri, Nov 23, 2012 at 1:26 AM, Mark McLoughlin <markmc at redhat.com>
> wrote:
>         On Thu, 2012-11-22 at 16:32 -0500, Eoghan Glynn wrote:

>         > > If we were just designing a solution for metering, would
>         we go for
>         > > the notifications option?
>         >
>         > Probably, but the scope of ceilo has widened a bit from pure
>         > metering.
>         
>         
>         Right. From a far remove, that looks like a mistake at this
>         early stage.
>         
>         As Sandy rightly says, if we just focused on metering, then
>         the
>         conclusion is simple - improve the notifications
>         infrastructure. If
>         Ceilometer focused on its original mission, forward progress
>         would be
>         easier.
> 
> 
> We've had conflicting messages on that point. When we applied for
> incubation I understood the TC mandate for the project to be "measure
> everything," and that has led to consideration of features like
> monitoring and collecting data more frequently and in different ways
> than would be necessary for just billing-related purposes. However,
> when changes related to those features are proposed we've had some
> push-back because of scope creep. Are you suggesting we should drop
> the monitoring work?

Uggh, I don't want to be talking about "TC mandates" here :)

A project gets accepted into Incubation partly based on the fact that
the TC has a reasonable level of trust in the judgement of the project
leaders. I'd hate to see the TC get into mandating the detailed
technical direction of an incubating project.

The way I'd personally like to see Ceilometer approach incubation is to
focus completely on metering, "knock the ball out of the park" and be a
kick-ass metering solution when Grizzly is released.

I worry that if the scope expands, you might get to Grizzly release time
and still have a bunch of loose ends. I mention it in this Nova
integration thread because it looks like a good example. IMHO, if
Ceilometer was focused (for now) purely on metering then this thread
would have been shorter and the solution might already be in place.

So, it's not about TC mandates. It's about TC folks giving their best
advice on the trade-offs you guys are facing. It's for you to make the
decision on those trade-offs.

(In fact, I only expressed an opinion here because the idea came up of
moving Nova's virt drivers out as an Oslo library)

That's not to say monitoring isn't important, or that there isn't a
whole lot of benefit to monitoring an metering folks working together,
or that no-one should be working on monitoring right now, or that Nova
doesn't need to support monitoring requirements, or ... just let's not
allow monitoring slow down progress on metering.

Cheers,
Mark.




More information about the OpenStack-dev mailing list