I have the following deployment and a load balancer service to connect to it:

apiVersion: apps/v1beta1
kind: Deployment
metadata:
  name: slave
spec:
  replicas: 2
  template:
    metadata:
      labels:
        app: slave
    spec:
      containers:
      - name: slave
        envFrom:
        - configMapRef:
            name: my-env-var
        image: registry.azurecr.io/slave:latest
        ports:
        - containerPort: 8080
      imagePullSecrets:
      - name: acr-auth
---
apiVersion: v1
kind: Service
metadata:
  name: slave
spec:
  type: LoadBalancer
  ports:
  - port: 8080
  selector:
    app: slave

I connect to this deployment from another deployment, let's called it master, through "slave:8080" which using DNS it gets resolved to an IP. So the master sends a request to the slave for a job. The slave does a job and reply with a response containing a link to the result file. The link is something like slave:8080/results.txt. So the master can now download the results file. This works fine when I have one slave. Now, I want to have more than one by using replicas. The problem is that how could the master get the results file from a certain slave since they are all sitting behind a load balancer.

This is the description of my slave service:

Name:                     slave
Namespace:                default
Labels:                   <none>
Annotations:              kubectl.kubernetes.io/last-applied-configuration={"apiVersion":"v1","kind":"Service","metadata":{"annotations":{},"name":"slave","namespace":"default"},"spec":{"ports":[{"port":8080}],"selector...
Selector:                 app=slave
Type:                     LoadBalancer
IP:                       10.0.233.16
LoadBalancer Ingress:     41.137.169.141
Port:                     <unset>  8080/TCP
TargetPort:               8080/TCP
NodePort:                 <unset>  32546/TCP
Endpoints:                10.244.0.213:8080,10.244.0.214:8080
Session Affinity:         None
External Traffic Policy:  Cluster
Events:                   <none>

I see two endpoints for the replicas but these are internal hence unreachable.

So how can the master get the results file from the replica slave that performed the job?

I want to know if an easy solution can be done before considering having a mounted volume or a different approach.

Thanks!