Implement SMI Metrics

[snehachhabria]

this will help extending the k8s API for users to make metrics requests (for resource defined in SMI spec)

• Extend Envoy to get custom metrics around source of traffic
  • Investigate how other meshes extend Envoy to add detailed src/dst data for metrics #998
  • Compile and run Envoy with example extensions #1013
• Implement SMI metrics api server

[michelleN]

blocked by #563
We need to figure out how to compose the service level metrics before we can build out the smi metrics implementation fully.

[michelleN]

Once we've completed the research on how to get envoy to emit the metrics/metadata we need, we'll also need to create an issue for doing the work to implement those things.

[michelleN]

[Task list moved to description]

[jshaughn]

Hi @michelleN . does this open issue mean that OSM does not yet make request metrics available for consumption?

[michelleN]

@jshaughn - OSM does make available metrics around success/error rate and latency available and you can view them already via the osm dashboard command. The SMI metrics spec requires us to compute some more granular metrics between a source and destination (for example between a two specific Pods or Deployments) so that is what is being done with this work.

[jshaughn]

@michelleN Thanks for the response. Pod-specific metrics are less interesting given the more dynamic nature of pods but the traffic for more logical aggregations of pods, like Deployments, is what I was looking for.

[snehachhabria]

closing this issue as we have an implementation of smi metrics already.