Machine id option to deploy

[Mark Ramm]

On 04/08/2013 03:09 PM, roger peppe wrote:
> I feel that deploy-to is possibly a first step towards placement, which
> I think can be seen as largely orthogonal to constraints,
> although naturally they will interact, and we will need to decide
> how they should do so.
Yea, and I think the fundamental issue here is that we don't yet have 
placement and co-location fully figured out.

I doubt they are truly orthogonal (there seems like there would have to 
be *some* interaction) but I think your intuition that they are separate 
concerns seems reasonable, and is definitely something to explore.

But we need to do something to support existing open stack deployments, 
as well as the needs of Canonical IS (using Juju+MaaS to deploy 
postgres).  And we need to do it in order to get the great work that 
we've put into the GO port into usable form for them ASAP (ideally by 
the 13.04 release which is just a couple of weeks away!!!).

Whatever we end up doing we need to make sure that we can do the "right 
thing" in the future when we fully understand what that is. And we need 
to do "something" now to support people who are actually trying to use 
Juju to get real work done in production environments.

And that's the real issue here:

  * Will this lead us down a bad path from which it will be difficult to
    recover?
  * Will we be painting --force-machine users into a corner?

My guess based on the discussions I have seen (and I fully admit I 
haven't been around long enough to remember when this came up for the 
first time) is that this is a compromise that will turn out OK.

We get something a bit rough around the edges that serves users now, and 
still give ourselves lots of room to create a "proper" solution when we 
implement containerization and real placement logic.

--Mark Ramm

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/juju/attachments/20130408/ac790e8a/attachment-0001.html>