introduce service graph metrics

[frzifus]

Area(s)

area:new

Is your change request related to a problem? Please describe.

Generating metrics to describe the relationship between services is generally very useful for inspecting systems. The servicegraphconnector provides a way to generate service graph metrics based on trace data. While details about the metrics are listed in the documentation those are tied to its implementation. Having a aligned spec would allow serve as a source of truth for other producers to be compatible (e.g. envoy) and consumers (e.g. grafana).

Describe the solution you'd like

Having a common schema for service graph metrics.

Describe alternatives you've considered

No response

Additional context

cc @Juraci @andreasgerstmayr

[frzifus]

@carsonip

[joaopgrassi]

This sounds very cool and useful! But I do worry that when bringing the metrics to semconv, we will have to adjust the naming. Given there are products already offering service graph support for them, I wonder how breaking changes will be seen. Was this discussed in any way in the collector SIG @frzifus ?

[frzifus]

Not yet official. The topic came up during a brainstorming debate at kubecon.

As of today, the metric list is not yet very large. Thats why I wouldn't be too worried about breaking changes in naming. More about the question of whether there is time to deal with the topic appropriately.

[joaopgrassi]

Alright thanks for the additional info! I think this is a good idea and semconv is the right place for it. I will remove the triage label and keep it as is.