Baremetal Kubernetes Cluster - Deploying An Application

This is part of a series on creating a Kubernetes cluster. In the previous post, we dove deeper into Kubernetes networking and created an Ingress Controller to make our Services accessible. In this post, we'll be deploying an application to the Kubernetes (k8s) cluster.

If you are setting up your own baremetal k8s cluster, I assume there's a good chance that you also already run a Plex Media Server. Therefore, I thought it would fun to deploy Tautulli, a monitoring application for Plex, as the first application in the cluster. Specifically, we'll be deploying the docker image for Tautilli that is maintained by the excellent people at

First, define a Deployment for the application. Deployments tell k8s how an application should be managed, such as what image to use, how many replicas should run, what volumes should be mounted, etc. A similar concept, called ReplicaSets, is sometimes used in guides on the internet. I recommend that you follow the guidance of the Kubernetes team and just stick with Deployments.

Save the following yaml into a file named tautulli-deployment.yaml:

apiVersion: apps/v1
kind: Deployment
  name: tautulli
    app: tautulli
  replicas: 1
      app: tautulli
        app: tautulli
        - name: tautulli
          image: linuxserver/tautulli
            - name: TZ
              value: "America/Chicago"
              memory: "2Gi"
              memory: "1Gi"
            - containerPort: 8181
              name: tautulli-web
            - mountPath: /config
              name: default-pvc
              subPath: tautulli
        - name: default-pvc
            claimName: nfs

I'll explain what this defines, starting at the top. First, this is a Deployment named tautulli. It has a label called app with a value of tautulli. This label will help match up the Service to the Deployment later.

The spec indicates it should only have 1 replica. It also has matchLabels defined to match a label named app with a value of tautulli. This needs to match the template label. I know this seems redundant, but it's required per the official docs.

The template spec indicates it will run a container with the docker image of linuxserver/tautulli. It will be created with an environment variable (env) named TZ with a value of America/Chicago. You may update this with your appropriate time zone, which can be found using this online tool.

The container will also request resources of at least 1 GB of memory, limited to 2 GB. The container will also expose port 8181.

The container will mount a volume named default-pvc at a subPath of tautulli to the local path /config. This means your NFS share will have a new directory created called tautulli that will contain the config for the application. This allows the Pod to be recreated without losing your config.

The template, therefore, needs to define a volume named default-pvc. It will provide this volume by binding to the persistentVolumeClaim named nfs (see earlier post).

We can apply this to the cluster using kubectl apply -f tautulli-deployment.yaml.

Now that the application is deployed, we need to define a Service that will expose the Pods outside the cluster network.

Save the following yaml into a file named tautulli-service.yaml:

kind: Service
apiVersion: v1
  name: tautulli
    app: tautulli
    - protocol: TCP
      port: 8181
      targetPort: 8181
      name: tautulli-web

This is much easier to understand. The Service named tautulli will expose TCP on port 8181 to targetPort 8181 on Pods matching the label app with value tautulli. This exposed port will be named tautulli-web for later reference.

Apply this to the cluster using kubectl apply -f tautulli-service.yaml.

The only piece missing now is the Ingress that will allow our host request to the cluster to be router to the appropriate Service.

Save the following yaml into a file named tautulli-ingress.yaml:

apiVersion: extensions/v1beta1
kind: Ingress
  name: tautulli-web
  - host:
      - path: /
          serviceName: tautulli
          servicePort: tautulli-web

This is perhaps the easiest to understand. An Ingress named tautulli-web contains a rule that requests for host should be router to the Service named tautulli on the servicePort named tautulli-web.

Apply this to the cluster using kubectl apply -f tautulli-ingress.yaml.

This will cause the Ingress Controller nginx-ingress to create a new site rule that is a reverse proxy to the Service defined in the Ingress.

Assuming you have the domain pointed to your public IP address and the appropriate firewall port-forwarding rule to forward port 80/443 to your k8s master node IP address, your service should be accessible from the internet.

If you do not have the domain name and port forwarding setup yet, you can still test out the Ingress with curl. Curl allows you to specify headers for an HTTP request and view the response. We can use this to tell the Ingress Controller that we're requesting and make that request directly to the external IP of the Ingress Controller. You can find the external IP using the command at the bottom of the Ingress Controller post.

Here's a curl example specifying the host header.

curl -H "Host:" ""

If everything is working, the response should include a lot of HTML from Tautulli.

Assuming your domain name is working correctly and port 80 is forwarded to your k8s load balancer IP for the ingress controller service ( in my example), you should be able to navigate to and be prompted to setup Tautulli with the connection to your Plex server.

So, there you have it. All of our hard work has finally paid off, and we are hosting Tautulli in our Kubernetes cluster, taking advantage of all the benefits that come with it. It's time to wrap up this series with a few parting thoughts.