On 15/02/16 12:58, Tom Barber wrote: > In an alternative world, it would be great to be able to extend these > locally so when Amazon or whoever launch new instance types we could have a > local instance-types.yaml file or something that lets us declare new node > types without waiting for a new juju release, or perhaps you're stuck on a > specific rev it would alleviate that problem... rainy day project perhaps.
On a similar note, I would really like to enable bundles to be usable sanely across multiple different clouds, with detailed expressions of the best machine instance types for particular roles in the topology. So what's needed is an effective way to say "on the Azure cloud use this instance type, on AWS use that, and on Google do the following". I think we'd want to be able to say that a bundle has "small, medium and large" footprint variations, or perhaps "test, dense production and scale-out production" variations, and then have a mapping inside the bundle to specific machine types. If we have benchmarks for the bundles, then we could even automatically validate and tweak those machine types, optimising for value. Mark -- Juju mailing list [email protected] Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/juju
