... "docker exec configure file" is a sad but common pattern ...
Interesting discussions happen when you hang out with straight-talking Paul Czarkowski. There’s a long chain of circumstance that lead us from an Interop panel together at Barcelona (video) to bemoaning Ansible and Docker integration early one Sunday morning outside a gate in IAD.
What started as a rant about czray ways people find of injecting configuration into containers (we seemed to think file mounting configs was “least horrific”) turned into an discussion about how to retro-fit application registry features (like consul or etcd) into legacy applications.
Ansible Inventory is basically a static registry service.
While we both acknowledge that Ansible inventory is distinctly not a registry service, the idea is a useful way to help explain the interaction between registry and configuration. The most basic goal of a registry (there are others!) is to have system components be able to find and integrate with other system components. In that sense, the inventory creates allows operators to pre-wire this information in advance in a functional way.
The utility quickly falls apart because it’s difficult to create re-runable Ansible (people can barely pronounce idempotent as it is) that could handle incremental updates. Also, a registry provides many other important functions like service health and basic cross node storage that are import.
It may not be perfect, but I thought it was @pczarkowski insight worth passing on. What do you think?