Kubernetes can not see my application at the given IP

Question

I am trying to run my app using Kubernetes I have created a deployment from file called deployment.yaml a service from the file called service.yaml

Here is the content of deployment.yaml:

apiVersion: v1
kind: Pod

    metadata:
      name: kubectl-test
    spec:
      containers:
        - name: kubectl-test
          image: gcr.io/[my-name]/node-app:0.0.1
          imagePullPolicy: Always
          ports:
            - containerPort: 8080
              hostPort: 8080

This is my services.yaml

kind: Service
apiVersion: v1
metadata:
  #Service name
  name: kubectl-test-node-app
spec:
  selector:
    app: kubectl-test-189010
  ports:
    - protocol: TCP
      port: 8000
      targetPort: 8000
  type: LoadBalancer

When I run the command kubectl get deployments, I can see:

NAME                  DESIRED   CURRENT   UP-TO-DATE   AVAILABLE   AGE
kubernetes-bootcamp   1         1         1            1           16m

I can see my services by using kubectl get services and I can see:

NAME                    TYPE           CLUSTER-IP      EXTERNAL-IP   PORT(S)          AGE
kubectl-test-node-app   LoadBalancer   10.59.242.211   35.195.2.76   8000:31592/TCP   8m
kubernetes              ClusterIP      10.59.240.1     <none>        443/TCP          1d

When I try to visit the EXTERNAL-IP:8000 I cant see my app there, I don't get any error, but I don't what is going wrong!

EDIT

I had to create deployment using:

`kubectl run kubernetes-bootcamp --image=[image-name] --port=8080`

but not with this command

`kubectl create -f deployment.yaml`

Is this related to the port: 8000:31592/TCP ?

kubectl get pods
kubectl-test                           1/1       Running   0          4h
kubernetes-bootcamp-1654181842-czwqg   1/1       Running   0          3h

EDIT

Running the command kubectl describe services kubectl-test-node-app returns:

Name:                     kubectl-test-node-app
Namespace:                default
Labels:                   <none>
Annotations:              <none>
Selector:                 app=kubectl-test-189010
Type:                     LoadBalancer
IP:                       10.59.242.211
LoadBalancer Ingress:     35.195.2.76
Port:                     <unset>  8000/TCP
TargetPort:               8000/TCP
NodePort:                 <unset>  31592/TCP
Endpoints:                <none>
Session Affinity:         None
External Traffic Policy:  Cluster
Events:
  Type    Reason                Age   From                Message
  ----    ------                ----  ----                -------
  Normal  CreatingLoadBalancer  58m   service-controller  Creating load balancer
  Normal  CreatedLoadBalancer   57m   service-controller  Created load balancer

There are no endpoints defined. I guess this is the problem? How can I build the end points?

Answer 1

Your service specification target all pods with label "app=kubectl-test-189010" which your pod's specification doesn't have.

You need to update pod's metadata so it will match selector from service spec:

    metadata:
       name: kubectl-test
       app: kubectl-test-189010

Also your service accept connections on port 8000 and then forward it to port 8000 (targetPort) on pod. But pod listen on port 8080 (containerPort). So one of them should be changed to match another - targetPort or containerPort.

You can read more about configuring services in Kubernetes here: https://kubernetes.io/docs/concepts/services-networking/service/

Answer 2

As @Leonid Talalaev mentioned, your selectors in your pod and service aren't in sync. Also, looking at the output for kubectl get services, your app is exposed outside the cluster on NodePort 31592. Did you try accessing the app using this port?

I would also recommend using a Deployment instead of a Pod. Using a Deployment will automatically create the required ReplicaSet and Pod(s) when you create the Deployment with kubectl.

I'm not sure what your desired NodePort is, but if you want to expose your app outside the cluster on port 32000(for example), you could use the configs below. Then you should be able to access your app with http://<CLUSTER_IP>:32000/

# deployment.yml
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: kubectl-test
  labels:
    app: kubectl-test
spec:
  replicas: 3
  strategy: {}
  template:
    metadata:
      labels:
        app: kubectl-test
    spec:
      containers:
      - image: gcr.io/[my-name]/node-app:0.0.1
        name: kubectl-test
        ports:
        - containerPort: 8080
        resources: {}
      restartPolicy: Always
status: {}

# service.yml
kind: Service
metadata:  
  name: kubectl-test-node-app   
  labels:
    app: kubectl-test
spec:
  ports:
    - protocol: TCP
      name: "8080"
      port: 8080
      targetPort: 8080
      nodePort: 32000
  selector:
    app: kubectl-test
  type: LoadBalancer
  sessionAffinity: None
  externalTrafficPolicy: Cluster
status: 
  loadBalancer: {}

kubectl create -f deployment.yml
kubectl create -f service.yml

Answer 3

For starters, make sure you know what you want to accomplish. You say service and deployment, but your files refer to Service and Pod.

Deployment is an object, that will create ReplicaSets and scale them so that they result with a particular group of underlying Pods launched in cluster. Taken from Deployments metadata, they will have some labels. These labels are what Service will match it's selector against (labels&selectors). I do assume you do want to use Deployment, to leverage stuff like rolling updates (it's generaly very hard to update Pod in many cases, and you end up deleting and recreating it manualy) or scaling.

With that assumption, you need Deployment (or Pod if you really want to go the "manual" way), that has something like folowing

metadata:
  labels:
    app: myappname

followed by Service that has a selector defined similarly

selector:
  app: myappname

that should be enough for you to have a working setup, and you should see your service will now have active endpoints (assuming your pods are launched and in Ready state.

Also, if you want to contain version in your labels, so that you can ie. target particular pod version with a service, or maybe use it for another reason like pod affinity, you should split it into two labels like

metadata:
  labels:
    app: myappname
    ver: v12

Answer 4

I was able to fix it, I had to change the dokcerfile and made some adjustments, then my deployment.yaml file had to be also changed a little, mainly adding the correct name and port number and the same in my service.yaml, and finally my webpack configuration had lo be changed!