Deploying units with a specific AMI
clint at ubuntu.com
Fri Apr 27 07:47:36 UTC 2012
Excerpts from Graham Binns's message of Thu Apr 26 12:47:41 -0700 2012:
> Hi all,
> Long story short: I'd like to be able to launch Juju nodes on EC2
> using a specific AMI. I know I used to be able to do this by setting
> default-image-id in my environments.yaml, but this now causes Juju to
> shout at me and then exit. Clint mentioned that he thought there was a
> way to use default-image-id still, but he couldn't recall it. Does
> anyone know what it might be?
> To save bikeshedding, I know that this is not the best way to do
> things. I'm looking to solve a particular problem and whilst there are
> other, more elegant solutions I'd quite like to know if my original,
> ugly solution could actually work.
I've had all day to think about this, and I actually think we should
really think about taking more advantage of snapshots and images in Juju.
I have to wonder if we should consider having a flag for deploy that
will take a snapshot right after the 'install' hook has run.
In theory, install should always produce the same result. There should
be nothing machine specific happening there. So if we were to take
a snapshot, then every add-unit after that could boot from that snapshot.
Also we should be able to keep track of these snapshots for each
charm+version and boot future instances from those snapshots.
Seems like a lofty, but worthy goal, for juju optimization.
More information about the Juju