[ANN] juju-core 1.9.1 has been released

[Clint Byrum]

Excerpts from David Cheney's message of 2012-11-01 06:33:32 -0700:
> juju-core 1.9.1
> ===============
> 
> A new experimental release of Juju, juju-core 1.9.1 is now available for 
> testing. This is first release of the Go port of Juju and should be 
> considered _experimental_.
> 

WOOOOO HOOOOOO!

> Getting Juju
> ------------
> 
> juju-core 1.9.1 is available from the Gophers PPA
> 
> https://code.launchpad.net/~gophers/+archive/go
> 
> New and Notable
> ---------------
> 
> In addition to the port from Python to Go, this release includes several 
> important improvements
> 
> * The Juju state is now implemented in MongoDB. This replaces ZooKeeper 
> and provides the underlying foundation for making the state highly 
> available.
> * Experimental support for upgrading a Juju environment. This will allow 
> Juju itself to be upgraded inside a running environment. See the juju 
> upgrade-juju command.
> * Internally all Juju agents incorporate the concept of a lifecycle. The 
> lifecycle is a reliable way to manage the death and removal of resources 
> in a predictable manner. For instance, the stop hook now works as expected.
> 
> Known bugs and limitations
> --------------------------
> 
> As this is the first release of Juju in Go there are a large number of 
> features which are still to be implemented. A non exhaustive list of 
> known issues is as follows.
> 
> * juju-core 1.9.1 only supports EC2 as a provider. You may need to 
> remove custom configuration items from your environments.yaml to produce 
> a configuration file that juju-core 1.9.1 accepts.

This seems like a bug. Warning and ignoring on an unimplemented provider
would seem a better option than just refusing it. I don't see a downside
other than a lot of warnings about unimplemented providers until they
are implemented. :)

> * us-east-1 is the most reliable region to deploy into. Due to subtle 
> differences in the EC2 regions, juju bootstrap may fail when creating 
> the control bucket. This is issue #1042107, the current workaround is to 
> use us-east-1.
> * The juju-core package currently cannot be installed in parallel with 
> juju 0.5/0.6 due to conflicting ownership of /usr/bin/juju (and others). 
> This may be mitigated in the future with alternatives.
> * destroy-service is not yet implemented. As a workaround you can remove 
> all the units assigned to a service. This is issue #1027878.
> * remove-unit currently does not cause the underlying instance to shut 
> down. This somewhat reduces the usefulness of the previous workaround. 
> #1067127

Isn't that intentional so we can decide to back up the data before
terminating the machine?