Merge lp://staging/~psivaa/uci-engine/image-watcher into lp://staging/uci-engine

Psivaa,

Thanks for working on this script.

I have nothing against monitoring build-info.txt instead of using simplestreams (python-simplestreams), as long as it works, as it seems to be. We can invest time later to rewrite the whole service to benefit of simplestreams goodies (supports the whole procedure from stream to glance).

There are some questions and remarks inline, specially regarding the topology of the new CLIs and how the charm is supposed to deploy them and its implication on configuration.

Thanks cprov for the comments.

I have removed using a special config file and relying on the amqp_utils.get_config() for the config information. For this as you pointed out, amqp-relation hook needs to be implemented in the cron-cmd charm.

Thanks again.

Psivaa,

Thanks for the cleanup, I have just a minor suggestion about allowing image_watch.main() to be properly tested, but it's not a blocker.

Don't forget to coordinate with Joe for supporting amqp-relation in cron-cmd charm.

> Psivaa,
>
> Thanks for working on this script.
>
> I have nothing against monitoring build-info.txt instead of using
> simplestreams (python-simplestreams), as long as it works, as it seems to be.
> We can invest time later to rewrite the whole service to benefit of
> simplestreams goodies (supports the whole procedure from stream to glance).
This is actually possible now, we discussed yesterday and I think we all agreed *not* to do that though right?

Celso mentioned (if I understood correctly) in a MP for me that we should take all of these services for ubuntucore/snappy and organize them under a single directory rather than having an individual directory for each one off the root of our tree. I think this probably makes sense because they are all related to ubuntucore/snappy, rather than, for instance, a generic image watcher that may have to monitor some other type of source and handle things differently.

Some additional comments below

Psivaa, Paul,

I retracted my concerns about having one directory per service (CLIs + worker python module), it the right direction.

With that tree organisation we run CLIs locally or via with `run-python.py` and properly configure rabbit-worker. To me, it looks perfectly fine and acceptable and if it needs rework we can deal with it later.

[]

Thanks,
@plars: I have addressed your comments inline. and fixed where appropriate. About checking the latest against the glance image-list, we need to have a uniquely corresponding image in glance to the image in cloud-image. Pending that, I believe we only could check against a local cache. Please let me know what you think. Thanks

I'm fine if we sort out that detail later. I believe we can just call the glance image whatever we want though. The image we download (file) will be something like: devel-core-amd64-disk1.img so maybe we could call it something like devel-core-amd64-${imageid} (ex. devel-core-amd64-20150115.1)

In light of the shift of direction though, I think I'd rather see this get merged and we can sort out those details later. It's easy enough to change things once they are in place, and you've done a great job of getting the ball rolling with this. Thanks for working on this!