I found building services with the layers in mind, starting from the highest level:
- Transport Layer (Http/grpc)
- Service Metrics Layer
- Service Safety (Balancing & Limiting)
- Biz Analytics
- Biz Metrics
- Biz Logic
The separation of layers promotes robust, reliable, maintainable micro services (modeled after go-kit). I tried to implement as many as possible to give an example of what I look for in the genesis of a rest/service.
For my repo / storage layer, I chose to wrap sync.Map instead of implementing my own map w/ locks after learning that
it takes care of thundering heard automagically (mentioning because I am aware of a bare map w/ locks).
- Would be a nice middleware or wrapper ideally around a client
- Timeouts are set at the server level; however it would be nice to incorporate this at the service method level as well.
- I can write a mean dockerfile
- I would like to have an error type, and bubble it up to a mapper[errorType] which would return the proper error/response code.
- Example of a PR I made
- The repo/store would need to be centeralized at the least. At the moment, each instance has its own store.