How to bootstrap canonistack/openstack with trunk now?

[Ian Booth]

On 29/05/13 01:43, Scott Moser wrote:
> On Tue, 28 May 2013, Ian Booth wrote:
> 
>> Hi
>>
>> I am guessing you are running from source. For Canonistack, everything has been
>> set up to "just work", but sadly as of right now only for region lcy01. ie the
>> simplestreams metadata is automatically published to the shared Canonistack
>> public bucket and so long as the public bucket url is correctly set, the
>> metadata should be found. NB there's a ticket raised to (for Canonistack) get
>> the shared public bucket url published in keystone so real soon now you won't
>> need to manually specify that anymore.
> 
> That should be there now.
> 
> $ keystone catalog | sed -n -e '/^Service: product-streams/,/^Service:/p'
> Service: product-streams
> +-------------+------------------------------------------------------+
> |   Property  |      Value                                           |
> +-------------+------------------------------------------------------+
> |   adminURL  | https://swift.canonistack.<trim>/simplestreams/data/ |
> |      id     | db3cb46f5d704ff4a4f834b969ccbe76                     |
> | internalURL | https://swift.canonistack.<trim>/simplestreams/data/ |
> |  publicURL  | https://swift.canonistack.<trim>/simplestreams/data/ |
> |    region   | lcy01                                                |
> +-------------+------------------------------------------------------+
> 
> What did you think was wrong there? I'm mis-remembering something on this
> I think.
>

Yes, the product-streams endpoint is there.
What the new ticket is asking for is a juju-tools endpoint to advertise the URL
of the so-called public bucket. This is where the tools tarballs are stored and
having this endpoint advertised removes the need for each user to have to set up
their local env configuration with the public-bucket-url attribute.

>> With region lcy02, I believe there are issues gathering the image metadata, but
>> could be mis-remembering that. I can't recall the reason but it's being worked
>> on. Scott (who I've cc'ed) can fill in the gaps hopefully.
> 
> The issue with lcy02 is that canonistack is using glance-replicator [1] to
> sync images from lcy01 to lcy02.  Basically, what that means is that the
> publisher creates an image in lcy01, publishes data for that image, and
> then some point in the future that image is synced to lcy02.
> 
> The issue with this is that means the publisher that we have is not
> completely in control of when things appear in lcy02, so we have no easy
> way of knowing when what image ids to publish.
> 

Maybe we could use a separate products json file for lcy02 which would point to
the existing, replicated images and delay the updating of the json file for a
day or so to enable the replication to happen? So at worst case, the images the
user would be getting are a day old.