internal load balancer loses backend when pods scaled down to 0

[somersru]

Hi all,

I have an internal load balancer setup for access to our pods in GKE autopilot, with this yaml:

apiVersion: v1

kind: Service

metadata:

  name: internal-load-balancer-svc

  namespace: default

  annotations:

    networking.gke.io/load-balancer-type: "Internal"

spec:

  type: LoadBalancer

  loadBalancerIP: 10.128.0.62

  externalTrafficPolicy: Cluster

  selector:

    app: pgbouncer-main

  ports:

  - name: tcp-port

    protocol: TCP

    port: 5432

    targetPort: 5432

This setup works fine to send traffic to our pods. The problem arises when I'm done for the day, I scale our pods down to 0 and then when I return and scale up the pod to 1 to resume testing, there is no backends configured for the load balancer:

So I have to delete the load balancer and reload it from our cloud build/terraform setup. This works fine after I do it.  This is what it looks like:

But I have been scaling down the pod for months, but in the last week or so when I scale back up the pod there is no backend