Traffic splitting based on weight : Nginx Ingress controller AKS

Question

I have a simple use case; I have 4 Microservices lets say service-a, service-b, service-c, service-d.

For the purpose of testing, I want to split traffic based on weight like

• 40% to service-a
• 20% to service-b
• 10% to service-c
• 30% to service-d

Considering they will be accessed over same path: example.com/

I am planinng to go with NGINX Ingress Controller but I saw a limitation on documentation: https://kubernetes.github.io/ingress-nginx/user-guide/nginx-configuration/annotations/#canary

Currently a maximum of one canary ingress can be applied per Ingress rule.

There is a GitHub issue too related to the same: https://github.com/kubernetes/ingress-nginx/issues/5848

I am not able to understand what does this actually mean and will this limitation not allow me implement with 4 services. Does that mean I have to create 4 canary ingress with single ingress rule for all 4 services? All the example for traffic splitting using ingress controller is of 2 services. should I consider Istio as it does not have this limitation

Can someone please explain me with simple yaml code example about this limitation?

Answer 1

If you need to implement traffic splitting across multiple microservices in AKS using a load balancer like NGINX Ingress Controller, and are facing limitations with the canary configuration, a better approach could be using Istio, a service mesh that provides advanced traffic management capabilities.

install and setup istio first

Create Deployment and Service for Each Microservice (service-a, service-b, service-c, and service-d) using your docker image ( I'll use nginxdemos/hello image here)

# deployment-service-a.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: service-a
spec:
  replicas: 2
  selector:
    matchLabels:
      app: service-a
  template:
    metadata:
      labels:
        app: service-a
    spec:
      containers:
      - name: service-a
        image: nginxdemos/hello
        ports:
        - containerPort: 80

---
apiVersion: v1
kind: Service
metadata:
  name: service-a
spec:
  ports:
  - port: 80
    targetPort: 80
  selector:
    app: service-a

repeat for b, c and d same way as above and apply kubectl apply -f <filename>.yaml

next steps involve setting up the Istio Gateway and Virtual Service to handle the traffic distribution according to your specified weights. This will enable you to route incoming traffic to your services service-a, service-b, and service-c at the specified ratios

Create Istio Gateway

apiVersion: networking.istio.io/v1beta1
kind: Gateway
metadata:
  name: example-gateway
spec:
  selector:
    istio: ingressgateway
  servers:
  - port:
      number: 80
      name: http
      protocol: HTTP
    hosts:
    - "*"

Create Istio VirtualService

apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: my-virtualservice
  namespace: default  # Ensure this is the correct namespace for your services
spec:
  hosts:
  - "*"  # This can be specific to your domain if needed
  gateways:
  - my-gateway
  http:
  - route:
    - destination:
        host: service-a.default.svc.cluster.local  # Adjust the FQDN as necessary
      weight: 40
    - destination:
        host: service-b.default.svc.cluster.your-cluster.com
      weight: 20
    - destination:
        host: service-c.default.svc.cluster.your-cluster.com
      weight: 10
    - destination:
        host: service-d.default.svc.cluster.your-cluster.com
      weight: 30

Adjust the namespace, host, and port parameters accordingly

kubectl apply -f istio-gateway.yaml
kubectl apply -f istio-virtualservice.yaml

kubectl get svc istio-ingressgateway -n istio-system

You can now verify the VirtualService Configuration if it is matching your split -

kubectl describe virtualservice my-virtualservice -n default