Install marketplace apps on a cluster

A cluster’s Marketplace tab installs common add-on tools on the cluster for you, without Helm charts or manifests of your own.

Before you begin

• Your cluster reads Running.
• For the NGINX Ingress Controller, the cluster’s LB IP Pool has a free IP; otherwise its external IP stays on <pending>. See Expose a service with a load balancer.
• (Optional) To remove a tool later, install Helm on your computer and point it at the cluster’s kubeconfig.

Choose a tool

The tab, headed Market Place, offers:

• Monitoring Stack, under Monitoring Tools: Prometheus and Grafana, for metrics and dashboards of your pods, containers, and nodes.
• Cert-Manager, under Network Tools: gets TLS certificates for your host names and renews them before they expire.
• NGINX Ingress Controller, under Network Tools: routes HTTP and HTTPS traffic to several applications by host name, through one LoadBalancer service.

Install a tool

1. In MyAccount, under Compute, click Kubernetes, click your cluster’s name, and open the Marketplace tab.
2. Click Install on the tool you want.

The tool reads Installing…, then Installed. If the install fails, the tool shows Retry with the error; read the error, and click Retry to try again.

To check the tool, list the pods in its namespace, given in the tool’s section:

kubectl get pods --namespace <namespace>

Every pod reads Running.

Each tool installs as a Helm release, and you remove it with helm uninstall; each tool’s section gives the command.

Use the Monitoring Stack

Monitoring Stack installs the kube-prometheus-stack chart as the Helm release prometheus, in the namespace kube-prometheus-stack: Prometheus, Grafana, Alertmanager, kube-state-metrics, and a node exporter. They are reached from inside the cluster.

1. Read Grafana’s sign-in details from the secret prometheus-grafana:

   kubectl get secret prometheus-grafana --namespace kube-prometheus-stack -o jsonpath="{.data.admin-user}" | base64 --decode
   kubectl get secret prometheus-grafana --namespace kube-prometheus-stack -o jsonpath="{.data.admin-password}" | base64 --decode

   The first command prints the user name, and the second the password.

2. Forward a local port to Grafana, and leave the command running:

   kubectl port-forward --namespace kube-prometheus-stack svc/prometheus-grafana 3000:80

   kubectl prints lines such as:

   Forwarding from 127.0.0.1:3000 -> 3000
   Forwarding from [::1]:3000 -> 3000

3. In a browser, open http://localhost:3000. Grafana’s sign-in page opens; sign in with the user name and password.

To remove it, run:

helm uninstall prometheus --namespace kube-prometheus-stack

Use Cert-Manager

Cert-Manager installs cert-manager version 1.12 as the Helm release cert-manager, in the namespace cert-manager. To get certificates, create an issuer and certificates as the cert-manager documentation describes.

To remove it, run:

helm uninstall cert-manager --namespace cert-manager

Use the NGINX Ingress Controller

NGINX Ingress Controller installs as the Helm release ingress-nginx, in the namespace ingress-nginx, with the IngressClass nginx, which your Ingress resources name. Its controller is a LoadBalancer service, so it takes a free IP from the cluster’s LB IP Pool, private LB IPs included; to make sure it gets a public address, free a public service IP, or name one in its MetalLB annotation as Route several applications through one IP shows. Check that it has one:

kubectl get service --namespace ingress-nginx

The controller’s row shows the IP it took from the pool:

NAME                       TYPE           CLUSTER-IP     EXTERNAL-IP     PORT(S)                      AGE
ingress-nginx-controller   LoadBalancer   <cluster-ip>   <external-ip>   80:<port>/TCP,443:<port>/TCP 1m

To route requests to your applications through it, see Route several applications through one IP.

To remove it, run:

helm uninstall ingress-nginx --namespace ingress-nginx

The tools you installed run in your cluster, and each can be removed with one Helm command. If a tool’s pods do not reach Running, see A pod stays Pending and Troubleshooting Kubernetes.