Juju providers should be able to get image id's automagically
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
juju-core |
Fix Released
|
High
|
Ian Booth | ||
pyjuju |
Triaged
|
Low
|
Unassigned |
Bug Description
When configuring Juju for the HP public cloud I needed to enter an image-id, like this:
environments:
hpcloud:
default-
default-series: precise
We value a "just works" approach, which in part means minimising the amount of information that a user needs to glean, when the system can infer that from things the user will definitely know.
In this case, while it's useful to be able to specify the image-id, it would be better if that were not required, but instead if juju, by policy and online services, could work that out from the "default-series" preference. Providers for different environments will need to map that series preference to the appropriate image automagically, so that users don't need to update that when, for example, a newer base image rolling up fixes is published.
Changed in juju-core: | |
importance: | Undecided → High |
status: | New → Confirmed |
Changed in juju-core: | |
status: | Confirmed → Triaged |
Changed in juju: | |
status: | Confirmed → Triaged |
importance: | High → Low |
tags: | added: images |
Changed in juju-core: | |
status: | Triaged → Fix Released |
assignee: | nobody → Ian Booth (wallyworld) |
This was discussed at length at UDS-Q, as part of this blueprint:
https:/ /blueprints. launchpad. net/ubuntu/ +spec/serverclo ud-r-juju- resource- map
It wasn't worked on because of the decision to focus on the Go port, but it is still needed. I've re-targetted it at R and proposed it for UDS-R.
We can use https:/ /cloud- images. ubuntu. com to lookup image IDs on public clouds, but we will also need to offer users the ability to have a local resource map for private clouds. There is also a need to map instance types to their constraints in EC2, since Amazon does not offer this via an API.