Draft: properly using metrics backends

[harmw]

Currently, the vamp-node-client supports Elasticsearch and Prometheus as a metrics datasource.
This client is used in various workflows, which both share an implementation that is pretty much hardcoded to only use ES.

• https://github.com/magneticio/vamp/blob/c3fa5fd9d7f4e5f591a2b9547e7ba0aced898fd9/lifter/src/main/resources/io/vamp/lifter/artifact/breeds/health.js
• https://github.com/magneticio/vamp/blob/c3fa5fd9d7f4e5f591a2b9547e7ba0aced898fd9/lifter/src/main/resources/io/vamp/lifter/artifact/breeds/metrics.js

To properly decouple these workflows from the metrics backend used in a Vamp environment/deployment, there is need for a common abstraction.

General configuration:

• Configure a metrics backend to use in a certain workflow using the environment:
  VAMP_METRICS_BACKEND=${implementation}

Interfaces to implement:

• Generic interface to return total throughput (rps)
  Only count and return requests within a certain configurable window
• Generic interface to return health
  Using status_code HTTP/500 and/or some other code
• Generic interface to execute raw queries
• Generic interface to return HTTP requests per second, using a specific HTTP status_code
• Generic interface to return latency
  Latency measured between haproxy and the backend server (container), in ms

All interfaces should support returning data for:

• gateways
• gateway ports