Provide a Service Mesh Interface (SMI) scaler

[tomkerkhove]

Provide a Service Mesh Interface (SMI) scaler.

Use-Case

People who are using service meshes would love to autoscale their workloads as well.

Proposal

Service Mesh Interface (SMI) is a specification which introduces a universal way to interact with service meshes.

As part of the specification, it covers how to retrieve traffic metrics which allows us to very easily integrate with it.

Good to know is that Service Mesh Interface (SMI) is being donated to CNCF (deck), provides a Go SDK and is supported by a variety of services meshes already such as Linkerd (CNCF), SuperGloo (Solo.io), Istio*, Consul Connect* and more.

(*) Via adapter

Relates to #538
Relates to servicemeshinterface/smi-spec#96

[tomkerkhove]

Here is a proposal for the trigger configuration:

  triggers:
  - type: service-mesh-interface
    metadata:
      metricName: "p90_response_latency" # Required - Name of the metric in the SMI's TrafficMetrics resource.
      value: "30m"                       # Required - Value defining when we have to start scaling.

Example is based on https://github.com/servicemeshinterface/smi-spec/blob/master/traffic-metrics.md#specification

It would then pull the full TrafficMetrics object for the scaleTargetRef and filter out the configured metricName and make a decision on the value to scale on.

We can use the official Go SDK for accessing the information: https://github.com/deislabs/smi-sdk-go

[tomkerkhove]

This PR is adding all the supported Service Meshes - servicemeshinterface/smi-spec.io#7

[adnxn]

taking a look at this and going off docs here

[tomkerkhove]

Awesome, thanks @adnxn!

[adnxn]

@tomkerkhove:

Here is a proposal for the trigger configuration:

should the trigger configuration include a unit specification as well? it is part of the smi spec here https://github.com/servicemeshinterface/smi-spec/blob/master/traffic-metrics.md

for example,

 triggers:
  - type: service-mesh-interface
    metadata:
      metricName: "p90_response_latency"
      unit: "seconds"
      value: "30m"    

[tomkerkhove]

Sounds good to me!

[tomkerkhove]

Any update on this @adnxn?

[tomkerkhove]

Feel free to let us know if you need help @adnxn

[adnxn]

hey i havent had time to follow up on this. if someone else wants to take it over, feel free. fwiw #783 builds off v1.

[arschles]

I'm not sure why it didn't show up, but we discussed this in #538 (comment) (cc @tomkerkhove)

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.

[tomkerkhove]

This will be part of our add-on as per kedacore/http-add-on#6

[arschles]

Specifically: kedacore/http-add-on#6