Upcoming changes for the ec2 provider

Andrew Wilkins andrew.wilkins at canonical.com
Thu Oct 6 23:33:56 UTC 2016


On Fri, Oct 7, 2016 at 7:26 AM Mark Shuttleworth <mark at ubuntu.com> wrote:

> On 06/10/16 16:14, Andrew Wilkins wrote:
>
> On Thu, Oct 6, 2016 at 8:30 PM Rick Harding <rick.harding at canonical.com>
> wrote:
>
> Thanks Andrew, this is great to hear. Can I bug you about details as to
> how it works? Does this introduce their pricing API as a blocker to
> deploying with Juju? If they introduce a change to the API we miss or their
> API goes down is there any sort of cache of the info that users can
> continue with until service is restored?
>
>
> So I said API because that's what it's called by AWS, but IMO it's a bit
> of an overstatement. It's just a set of (very large) JSON files in a well
> defined location. Very large, as in 48M for the one file that we need.
>
> Because it's so big, pulling that down each time you go to bootstrap isn't
> really reasonable. So for now, we're doing that at build time, and the
> information is still hard coded into the client. The process can be
> automated, so there shouldn't be a great lag time between updates and
> bringing them into a Juju release.
>
> We intend to update Juju later to periodically pull down the JSON file
> server-side, asynchronously, so a running controller will get the updates
> without us having to release a new Juju or you having to upgrade.
>
>
> Could we not process the file on our side regularly and publish a
> (smaller) just-what-juju-needs bit of json at a well-known location
> ourselves?
>

Yes, we could. There was some discussion about doing that, and someone
brought up concerns that I can't recall (or find) now. We will explore this
option further.

Cheers,
Andrew
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/juju/attachments/20161006/d7d1e984/attachment.html>


More information about the Juju mailing list