I’m using AWS Network Load Balancer as a k8s service for my deployment. The reason I use NLB instead of ALB is that I need a static host to expose my service inside my Elastic Kubernetes Service(EKS). The NLB I delpoied uses such yaml:

apiVersion: v1
kind:Service
metadata:
  annotations:
    service.beta.kubernetes.io/aws-load-balancer-nlb-target-type: ip
    service.beta.kubernetes.io/aws-load-balancer-scheme: internet-facing
    service.beta.kubernetes.io/aws-load-balancer-type: external
  name: my-nlb
  namespace: my-ns
spec:
  ports:
    - name: webport
      port: 80
      protocol: TCP
      targetPort: 80
    selector:
      app: my-app
    type: LoadBalancer

The problem is that, when I made a deployment with 2 replicas, and checked they are selected by NLB. I do some test like:

ab -c 100 -n 3000 http://AWS-NLB-Hostname

It seems that all the requests went to the same pod.

I know NLB is using hash for load balancing, does it mean it will load all requests from the same IP to the same target? If so (or not), what should I do to fix that?

Because my service is stafulless, I hope that all the requests are balanced across all the pods.