Configure ServiceRoute for (multi-port, multi-protocol) services

This how-to document will show you how to configure routes to services exposing multiple ports through a single ServiceRoute config.

Scenario

Consider a backend service named tcp-echo which exposes two ports, 9000 and 9001 over TCP. The service has two versions v1 and v2 and traffic splitting needs to be achieved between these two versions for both ports. In order to achieve this, a ServiceRoute with port level settings needs to be configured.

Deploy the tcp-echo Service

Deploy the tcp-echo application from the Istio's samples directory into the echo namespace by installing these manifests.

TSB configuration

Deploy a Workspace and Traffic Group

Apply the following configuration to create a Workspace and a Traffic Group.

note

The examples assume that you have already created an organization named tetrateio and a tenant named tetrate.

apiVersion: api.tsb.tetrate.io/v2
kind: Workspace
metadata:
 name: tcp-multiport-ws
 organization: tetrateio
 tenant: tetrate
spec:
 namespaceSelector:
   names:
     - "*/echo"
---
apiVersion: traffic.tsb.tetrate.io/v2
kind: Group
metadata:
 name: tcp-multiport-tg
 organization: tetrateio
 tenant: tetrate
 workspace: tcp-multiport-ws
spec:
 configMode: BRIDGED
 namespaceSelector:
   names:
   - "*/echo"

Deploy the ServiceRoute

Apply the following configuration to create the ServiceRoute that configures both ports.

apiVersion: traffic.tsb.tetrate.io/v2
kind: ServiceRoute
metadata:
  name: tcp-multiport-service-route
  organization: tetrateio
  tenant: tetrate
  workspace: tcp-multiport-ws
  group: tcp-multiport-tg
spec:
  service: "echo/tcp-echo.svc.cluster.local"
  portLevelSettings:
    - port: 9000
      trafficType: TCP
    - port: 9001
      trafficType: TCP
  subsets:
    - name: v1
      labels:
        version: v1
      weight: 80
    - name: v2
      labels:
        version: v2
      weight: 20

Testing

To verify that the routes have been set successfully, try curling several times to the echo pod. The request will be forwarded to the v1 pod most of the times because of the 80:20 weight ratio set between v1:v2.

For testing TCP traffic, use nc instead.

kubectl -n echo exec -it <pod-name> -c <container-name> -- curl -H "X-B3-Sampled: 1" -sv tcp-echo.svc.cluster.local:9000
kubectl -n echo exec -it <pod-name> -c <container-name> -- sh -c "echo hello | nc -v tcp-echo.svc.cluster.local:9001"