We need to integrate txstatsd into lp:uci-engine so we can start to build up metrics about how long operations are taking. One big thing to consider is the server itself. Should we use a pre-existing server and make this a configuration option when deploying. This would be nice because the stats could span the use of how we teardown and re-deploy things. The other option would be to deploy a txstatsd server inside. This might be a pain because you'd have to add charm hooks for everything to know about this server. I'd probably vote for:
- make it configurable (and optionally just disabled if no server is provided via unit_config) and not deploy the server.
We need to then collect metrics like:
* how long does it take to "clean" a ppa
* how long does it take to [build|upload] a package
* how long does it take to [build|test] an image