state: unit assignment respects machine series

state: unit assignment respects machine series

That is: it is no longer possible to assign a unit to a machine with the
wrong series, and machines created for the use of a given unit now take that
unit's series. A "Series" field was added to the unit document to make it
possible to check this without way-too-many roundtrips.

https://code.launchpad.net/~fwereade/juju-core/state-assign-respects-series/+merge/153332

(do not edit description out of merge proposal)

[fwereade, 2013-03-14 11:39:30]

Please take a look.

[dimitern, 2013-03-14 11:47:18]

LGTM. As Mr. Burns would say "Eeeexcellent" :)

[dave_cheney.net, 2013-03-15 00:21:59]

On 2013/03/14 11:47:18, dimitern wrote:
> LGTM. As Mr. Burns would say "Eeeexcellent" :)

To clarify. Prior to this CL, if I had used default-series set to quantal, then
deployed cs:precise/mysql, it will deploy the precise mysql charm on quantal. 

^ note - this is the behavior today, I just confirmed.

With this proposal, are you saying that the machine created by the PA will
always be of the series of the charm that was deployed ?

[rog, 2013-03-15 10:10:56]

On 2013/03/14 11:39:30, fwereade wrote:
> Please take a look.

LGTM, nice.

[rog, 2013-03-15 10:10:56]

On 2013/03/14 11:39:30, fwereade wrote:
> Please take a look.

LGTM, nice.

[fwereade, 2013-03-15 10:29:10]

On 2013/03/15 00:21:59, dfc wrote:
> On 2013/03/14 11:47:18, dimitern wrote:
> > LGTM. As Mr. Burns would say "Eeeexcellent" :)
> 
> To clarify. Prior to this CL, if I had used default-series set to quantal,
then
> deployed cs:precise/mysql, it will deploy the precise mysql charm on quantal. 
> 
> ^ note - this is the behavior today, I just confirmed.
> 
> With this proposal, are you saying that the machine created by the PA will
> always be of the series of the charm that was deployed ?

As of this change, machines in state will know what series they ought to run, as
determined by the charm; but Environ.StartInstance doesn't yet accept a series
parameter, so we remain precisely as broken as before. The StartInstance
signature fix will come soon, from either thumper or myself, and after that
we'll start getting the individual providers to handle it appropriately. That is
to say: a fix is in sight, but actual provisioning behaviour has not yet
changed.

[fwereade, 2013-03-15 11:51:46]

*** Submitted:

state: unit assignment respects machine series

That is: it is no longer possible to assign a unit to a machine with the
wrong series, and machines created for the use of a given unit now take that
unit's series. A "Series" field was added to the unit document to make it
possible to check this without way-too-many roundtrips.

R=dimitern, dfc, rog
CC=
https://codereview.appspot.com/7715046