Attach Cluster using tunnel

Using the CLI to attach a Kubernetes Cluster using a Tunnel

Creating a gateway

Identify the management cluster endpoint

Obtain the hostname and CA certificate for the management cluster:

hostname=$(kubectl get service -n kommander kommander-traefik -o go-template='{{with index .status.loadBalancer.ingress 0}}{{or .hostname .ip}}{{end}}')
b64ca_cert=$(kubectl get secret -n kommander kommander-bootstrap-root-ca -o=go-template='{{index .data "tls.crt"}}')

Specify a workspace namespace

Obtain the desired workspace namespace on the management cluster for the tunnel gateway:

namespace=$(kubectl get workspace default-workspace -o jsonpath="{.status.namespaceRef.name}")

Alternatively, if you wish to create a new workspace instead of using an existing workspace:

workspace=sample
namespace=${workspace}

cat > workspace.yaml <<EOF
apiVersion: workspaces.kommander.mesosphere.io/v1alpha1
kind: Workspace
metadata:
  annotations:
    kommander.mesosphere.io/display-name: ${workspace}
  name: ${workspace}
spec:
  namespaceName: ${namespace}
EOF

kubectl apply -f workspace.yaml

You can verify the workspace exists using:

kubectl get workspace ${workspace}

Create a tunnel gateway

Create a tunnel gateway on the management cluster to listen for tunnel agents on remote clusters:

NOTE: Kommander uses Traefik 2 ingress, which requires explicit definition of strip prefix middleware as a Kubernetes API object, opposed to a simple annotation. Kommander provides default middleware that supports creating tunnels only on the /dkp/tunnel URL prefix. This is indicated by using the extra annotation, traefik.ingress.kubernetes.io/router.middlewares: kommander-stripprefixes-kubetunnel@kubernetescrd as shown in the code sample that follows. If you want to expose a tunnel on a different URL prefix, you must manage your own middleware configuration.

cacert_secret=kubetunnel-ca
gateway=sample-gateway

cat > gateway.yaml <<EOF
apiVersion: v1
kind: Secret
metadata:
  namespace: ${namespace}
  name: ${cacert_secret}
data:
  ca.crt:
    ${b64ca_cert}
---
apiVersion: kubetunnel.d2iq.io/v1alpha1
kind: TunnelGateway
metadata:
  namespace: ${namespace}
  name: ${gateway}
spec:
  ingress:
    caSecretRef:
      namespace: ${namespace}
      name: ${cacert_secret}
    loadBalancer:
      hostname: ${hostname}
    urlPathPrefix: /dkp/tunnel
    extraAnnotations:
      kubernetes.io/ingress.class: kommander-traefik
      traefik.ingress.kubernetes.io/router.tls: "true"
      traefik.ingress.kubernetes.io/router.middlewares: kommander-stripprefixes-kubetunnel@kubernetescrd
EOF

kubectl apply -f gateway.yaml

You can verify the gateway exists using the command:

kubectl get tunnelgateway -n ${namespace} ${gateway}

Connecting a remote cluster

Create a tunnel connector

Create a tunnel connector on the management cluster for the remote cluster:

connector=sample-connector

cat > connector.yaml <<EOF
apiVersion: kubetunnel.d2iq.io/v1alpha1
kind: TunnelConnector
metadata:
  namespace: ${namespace}
  name: ${connector}
spec:
  gatewayRef:
    name: ${gateway}
EOF

kubectl apply -f connector.yaml

You can verify the connector exists using:

kubectl get tunnelconnector -n ${namespace} ${connector}

Wait for the tunnel connector to reach Listening state and then export the agent manifest:

while [ "$(kubectl get tunnelconnector -n ${namespace} ${connector} -o jsonpath="{.status.state}")" != "Listening" ]
do
  sleep 5
done

manifest=$(kubectl get tunnelconnector -n ${namespace} ${connector} -o jsonpath="{.status.tunnelAgent.manifestsRef.name}")
while [ -z ${manifest} ]
do
  sleep 5
  manifest=$(kubectl get tunnelconnector -n ${namespace} ${connector} -o jsonpath="{.status.tunnelAgent.manifestsRef.name}")
done

kubectl get secret -n ${namespace} ${manifest} -o jsonpath='{.data.manifests\.yaml}' | base64 -d > manifest.yaml

Create a network policy for the tunnel server

This step is optional, but improves security by restricting which remote hosts can connect to the tunnel.

Apply a network policy that restricts access to the tunnel to specific namespaces and IP blocks. The following example permits connections from pods running in the kommander-federation, kommander, and kubeaddons namespaces, from pods running in namespaces with a label kubetunnel.d2iq.io/networkpolicy matching the tunnel name and namespace, and to remote clusters with IP addresses in the ranges 192.0.2.0 to 192.0.2.255 and 203.0.113.0 to 203.0.113.255:

cat > net.yaml <<EOF
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  namespace: ${namespace}
  name: ${connector}-deny
  labels:
    kubetunnel.d2iq.io/tunnel-connector: ${connector}
    kubetunnel.d2iq.io/networkpolicy-type: "tunnel-server"
spec:
  podSelector:
    matchLabels:
      kubetunnel.d2iq.io/tunnel-connector: ${connector}
  policyTypes:
  - Ingress
---
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  namespace: ${namespace}
  name: ${connector}-allow
  labels:
    kubetunnel.d2iq.io/tunnel-connector: ${connector}
    kubetunnel.d2iq.io/networkpolicy-type: "tunnel-server"
spec:
  podSelector:
    matchLabels:
      kubetunnel.d2iq.io/tunnel-connector: ${connector}
  policyTypes:
  - Ingress
  ingress:
  - from:
    - namespaceSelector:
        matchLabels:
          app: "kommander-federation"
    - namespaceSelector:
        matchLabels:
          app: "kommander"
    - namespaceSelector:
        matchLabels:
          app: "kubeaddons"
    - namespaceSelector:
        matchLabels:
          kubetunnel.d2iq.io/networkpolicy: ${connector}-${namespace}
    - ipBlock:
        cidr: 192.0.2.0/24
    - ipBlock:
        cidr: 203.0.113.0/24
EOF

kubectl apply -f net.yaml

Setup the managed cluster

Copy the manifest.yaml file to the managed cluster and deploy the tunnel agent:

kubectl apply --context managed -f manifest.yaml

You can check the status of the created pods using:

kubectl get pods --context managed -n kubetunnel

After a short time, expect to see a post-kubeconfig pod that reaches Completed state and a tunnel-agent pod that stays in Running state.

NAME                           READY   STATUS      RESTARTS   AGE
post-kubeconfig-j2ghk          0/1     Completed   0          14m
tunnel-agent-f8d9f4cb4-thx8h   0/1     Running     0          14m

Add the managed cluster into Kommander

On the management cluster, wait for the tunnel to be connected by the tunnel agent:

while [ "$(kubectl get tunnelconnector -n ${namespace} ${connector} -o jsonpath="{.status.state}")" != "Connected" ]
do
  sleep 5
done

Add the cluster into Kommander:

managed=private-cluster
display_name=${managed}

cat > kommander.yaml <<EOF
apiVersion: kommander.mesosphere.io/v1beta1
kind: KommanderCluster
metadata:
  namespace: ${namespace}
  name: ${managed}
  annotations:
    kommander.mesosphere.io/display-name: ${display_name}
spec:
  clusterTunnelConnectorRef:
    name: ${connector}
EOF

kubectl apply -f kommander.yaml

Wait for the managed cluster to join the Kommander cluster:

while [ "$(kubectl get kommandercluster -n ${namespace} ${managed} -o jsonpath='{.status.phase}')" != "Joined" ]
do
  sleep 5
done

kubefed=$(kubectl get kommandercluster -n ${namespace} ${managed} -o jsonpath="{.status.kubefedclusterRef.name}")
while [ -z "${kubefed}" ]
do
  sleep 5
  kubefed=$(kubectl get kommandercluster -n ${namespace} ${managed} -o jsonpath="{.status.kubefedclusterRef.name}")
done

kubectl wait --for=condition=ready --timeout=60s kubefedcluster -n kommander ${kubefed}

kubectl get kubefedcluster -n kommander ${kubefed}

Using a remote cluster

To access services running on the remote cluster from the management cluster, connect to the tunnel proxy.

There are three methods:

  1. If the client program supports use of a kubeconfig file, use the managed cluster’s kubeconfig.
  2. If the client program supports SOCKS5 proxying, use the proxy directly.
  3. Otherwise, deploy a proxy server on the management cluster.

Managed cluster service

These sections require a service to run on the managed cluster.

As an example, start the following service:

service_namespace=test
service_name=webserver
service_port=8888
service_endpoint=${service_name}.${service_namespace}.svc.cluster.local:${service_port}

cat > nginx.yaml <<EOF
apiVersion: v1
kind: Namespace
metadata:
  name: ${service_namespace}
---
apiVersion: apps/v1
kind: Deployment
metadata:
  namespace: ${service_namespace}
  name: nginx-deployment
  labels:
    app: nginx-deployment
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx-app
  template:
    metadata:
      labels:
        app: nginx-app
    spec:
      containers:
      - name: nginx
        image: nginx:1.14.2
        ports:
        - containerPort: 80
---
apiVersion: v1
kind: Service
metadata:
  namespace: ${service_namespace}
  name: ${service_name}
spec:
  selector:
    app: nginx-app
  type: ClusterIP
  ports:
  - targetPort: 80
    port: ${service_port}
EOF

kubectl apply -f nginx.yaml

kubectl rollout status deploy -n ${service_namespace} nginx-deployment

On the managed cluster, a client Job can access this service using:

cat > curl.yaml <<EOF
apiVersion: batch/v1
kind: Job
metadata:
  name: curl
spec:
  template:
    spec:
      containers:
      - name: curl
        image: curlimages/curl:7.76.0
        command: ["curl", "--silent", "--show-error", "http://${service_endpoint}"]
      restartPolicy: Never
  backoffLimit: 4
EOF

kubectl apply -f curl.yaml

kubectl wait --for=condition=complete job curl

podname=$(kubectl get pods --selector=job-name=curl --field-selector=status.phase=Succeeded -o jsonpath='{.items[0].metadata.name}')

kubectl logs ${podname}

The final command returns the default Nginx web page:

<!DOCTYPE html>
<html>
<head>
<title>Welcome to nginx!</title>
<style>
    body {
        width: 35em;
        margin: 0 auto;
        font-family: Tahoma, Verdana, Arial, sans-serif;
    }
</style>
</head>
<body>
<h1>Welcome to nginx!</h1>
<p>If you see this page, the nginx web server is successfully installed and
working. Further configuration is required.</p>

<p>For online documentation and support please refer to
<a href="http://nginx.org/">nginx.org</a>.<br/>
Commercial support is available at
<a href="http://nginx.com/">nginx.com</a>.</p>

<p><em>Thank you for using nginx.</em></p>
</body>
</html>

Use of kubeconfig file

This is primarily useful for running kubectl commands on the management cluster to monitor the managed cluster.

On the management cluster, a kubeconfig file for the managed cluster configured to use the tunnel proxy is available as a Secret. The Secret’s name can be identified using:

kubeconfig_secret=$(kubectl get tunnelconnector -n ${namespace} ${connector} -o jsonpath='{.status.kubeconfigRef.name}')

After setting service_namespace and service_name to the managed service resource, on the management cluster run:

cat > get-service.yaml <<EOF
apiVersion: batch/v1
kind: Job
metadata:
  name: get-service
spec:
  template:
    spec:
      containers:
      - name: kubectl
        image: bitnami/kubectl:1.19
        command: ["kubectl", "get", "service", "-n", "${service_namespace}", "${service_name}"]
        env:
        - name: KUBECONFIG
          value: /tmp/kubeconfig/kubeconfig
        volumeMounts:
        - name: kubeconfig
          mountPath: /tmp/kubeconfig
      volumes:
      - name: kubeconfig
        secret:
          secretName: "${kubeconfig_secret}"
      restartPolicy: Never
  backoffLimit: 4
EOF

kubectl apply -n ${namespace} -f get-service.yaml

kubectl wait --for=condition=complete --timeout=5m job -n ${namespace} get-service

podname=$(kubectl get pods -n ${namespace} --selector=job-name=get-service --field-selector=status.phase=Succeeded -o jsonpath='{.items[0].metadata.name}')

kubectl logs -n ${namespace} ${podname}

Direct use of SOCKS5 proxy

To use the SOCKS5 proxy directly, obtain the SOCKS5 proxy endpoint using:

proxy_service=$(kubectl get tunnelconnector -n ${namespace} ${connector} -o jsonpath='{.status.tunnelServer.serviceRef.name}')

socks_proxy=$(kubectl get service -n ${namespace} "${proxy_service}" -o jsonpath='{.spec.clusterIP}{":"}{.spec.ports[?(@.name=="proxy")].port}')

Provide the value of ${socks_proxy} as the SOCKS5 proxy to your client.

For example, since curl supports SOCKS5 proxies, the managed service started above can be accessed from the management cluster by adding the SOCKS5 proxy to the curl command. After setting service_endpoint to the service endpoint, on the management cluster run:

cat > curl.yaml <<EOF
apiVersion: batch/v1
kind: Job
metadata:
  name: curl
spec:
  template:
    spec:
      containers:
      - name: curl
        image: curlimages/curl:7.76.0
        command: ["curl", "--silent", "--show-error", "--socks5-hostname", "${socks_proxy}", "http://${service_endpoint}"]
      restartPolicy: Never
  backoffLimit: 4
EOF

kubectl apply -f curl.yaml

kubectl wait --for=condition=complete --timeout=5m job curl

podname=$(kubectl get pods --selector=job-name=curl --field-selector=status.phase=Succeeded -o jsonpath='{.items[0].metadata.name}')

kubectl logs ${podname}

The final command returns the same output as for the job on the managed cluster, demonstrating that the job on the management cluster accessed the service running on the managed cluster.

Use of deployed proxy on management cluster

To deploy a proxy on the management cluster, obtain the SOCKS5 proxy endpoint using:

proxy_service=$(kubectl get tunnelconnector -n ${namespace} ${connector} -o jsonpath='{.status.tunnelServer.serviceRef.name}')

socks_proxy=$(kubectl get service -n ${namespace} "${proxy_service}" -o jsonpath='{.spec.clusterIP}{":"}{.spec.ports[?(@.name=="proxy")].port}')

Provide the value of ${socks_proxy} as the SOCKS5 proxy to a proxy deployed on the management cluster. After setting service_endpoint to the service endpoint, on the management cluster run:

cat > nginx-proxy.yaml <<EOF
apiVersion: cert-manager.io/v1
kind: Certificate
metadata:
  name: nginx-proxy-crt
spec:
  secretName: nginx-proxy-crt-secret
  dnsNames:
  - nginx-proxy-service.${namespace}.svc.cluster.local
  issuerRef:
    group: cert-manager.io
    kind: ClusterIssuer
    name: kubernetes-ca
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-proxy
  labels:
    app: nginx-proxy-deployment
spec:
  replicas: 1
  selector:
    matchLabels:
      app: nginx-proxy-app
  template:
    metadata:
      labels:
        app: nginx-proxy-app
    spec:
      containers:
      - name: nginx-proxy
        image: mesosphere/ghostunnel:v1.5.3-server-backend-proxy
        args:
        - "server"
        - "--listen=:443"
        - "--target=${service_endpoint}"
        - "--cert=/etc/certs/tls.crt"
        - "--key=/etc/certs/tls.key"
        - "--cacert=/etc/certs/ca.crt"
        - "--unsafe-target"
        - "--disable-authentication"
        env:
        - name: ALL_PROXY
          value: socks5://${socks_proxy}
        ports:
        - containerPort: 443
        volumeMounts:
        - name: certs
          mountPath: /etc/certs
      volumes:
      - name: certs
        secret:
          secretName: nginx-proxy-crt-secret
---
apiVersion: v1
kind: Service
metadata:
  name: nginx-proxy-service
spec:
  selector:
    app: nginx-proxy-app
  type: ClusterIP
  ports:
  - targetPort: 443
    port: 8765
EOF

kubectl apply -n ${namespace} -f nginx-proxy.yaml

kubectl rollout status deploy -n ${namespace} nginx-proxy

proxy_port=$(kubectl get service -n ${namespace} nginx-proxy-service -o jsonpath='{.spec.ports[0].port}')

Any client running on the management cluster can now access the service running on the managed cluster using the proxy service endpoint. Note in the following that the curl job runs in the same namespace as the proxy, to provide access to the CA certificate secret.

cat > curl.yaml <<EOF
apiVersion: batch/v1
kind: Job
metadata:
  name: curl
spec:
  template:
    spec:
      containers:
      - name: curl
        image: curlimages/curl:7.76.0
        command:
        - curl
        - --silent
        - --show-error
        - --cacert
        - /etc/certs/ca.crt
        - https://nginx-proxy-service.${namespace}.svc.cluster.local:${proxy_port}
        volumeMounts:
        - name: certs
          mountPath: /etc/certs
      volumes:
      - name: certs
        secret:
          secretName: nginx-proxy-crt-secret
      restartPolicy: Never
  backoffLimit: 4
EOF

kubectl apply -n ${namespace} -f curl.yaml

kubectl wait --for=condition=complete --timeout=5m job -n ${namespace} curl

podname=$(kubectl get pods -n ${namespace} --selector=job-name=curl --field-selector=status.phase=Succeeded -o jsonpath='{.items[0].metadata.name}')

kubectl logs -n ${namespace} ${podname}

The final command returns the same output as the job on the managed cluster, demonstrating that the job on the management cluster accessed the service running on the managed cluster.