Quickstart

    • Kubernetes 1.11+
    • CoreDNS installed as (versions 1.3+ supported)
    • Helm v3

    Depending on the tool you used to deploy your cluster you might need to tweak RBAC permissions.

    kubeadm

    If you used kubeadm to deploy your cluster, a fast way to allow the helm installation to perform all steps it needs is to edit the cluster-admin ClusterRoleBinding, adding the following to the subjects section:

    Command

    1. helm repo add traefik-mesh https://helm.traefik.io/mesh
    2. helm repo update
    3. helm install traefik-mesh traefik-mesh/traefik-mesh

    Expected output

    1. [...]
    3. NOTES:
    4. Thank you for installing traefik-mesh.
    6. Your release is named traefik-mesh.
    8. To learn more about the release, try:
    10.   $ helm status traefik-mesh
    11.   $ helm get traefik-mesh

    As an example, let’s deploy a server application and a client application under the test namespace.

    server.yaml

    1. ---
    2. apiVersion: apps/v1
    3. kind: Deployment
    4. metadata:
    5.   name: server
    6.   namespace: test
    7.     app: server
    8. spec:
    9.   replicas: 2
    11.     matchLabels:
    12.       app: server
    13.   template:
    14.     metadata:
    15.       labels:
    16.         app: server
    17.     spec:
    18.       containers:
    19.         - name: server
    20.           image: traefik/whoami:v1.6.0
    21.           ports:
    22.             - containerPort: 80
    23. ---
    24. kind: Service
    25. apiVersion: v1
    26. metadata:
    27.   name: server
    28.   namespace: test
    29. spec:
    30.   selector:
    31.     app: server
    32.   ports:
    33.     - name: web
    34.       protocol: TCP
    35.       port: 80

    Create the namespace then deploy those two applications:

    1. kubectl create namespace test
    2. kubectl apply -f server.yaml
    3. kubectl apply -f client.yaml

    You should now see the following output:

    Command

    1. kubectl get all -n test

    Expected output

    2. pod/client-7446fdf848-x96fq   1/1       Running   0          79s
    3. pod/server-7c8fd58db5-rchg8   1/1       Running   0          77s
    4. pod/server-7c8fd58db5-sd4f9   1/1       Running   0          77s
    6. NAME             TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)   AGE
    7. service/server   ClusterIP   10.43.17.247   <none>        80/TCP    77s
    9. NAME                     READY     UP-TO-DATE   AVAILABLE   AGE
    10. deployment.apps/client   1/1       1            1           79s
    11. deployment.apps/server   2/2       2            2           77s
    13. NAME                                DESIRED   CURRENT   READY     AGE
    14. replicaset.apps/client-7446fdf848   1         1         1         79s
    15. replicaset.apps/server-7c8fd58db5   2         2         2         77s

    Take note of the client app pod name (here it’s client-7446fdf848-x96fq) and open a new terminal session inside this pod using kubectl exec.

    From inside the client container, make sure your server is reachable using the Kubernetes DNS service discovery.

    1. curl server.test.svc.cluster.local

    Expected Output

    1. Hostname: server-7c8fd58db5-sd4f9
    2. IP: 127.0.0.1
    3. IP: ::1
    4. IP: 10.42.2.10
    5. IP: fe80::a4ec:77ff:fe37:1cdd
    6. RemoteAddr: 10.42.2.9:46078
    7. GET / HTTP/1.1
    8. Host: server.test.svc.cluster.local
    9. User-Agent: curl/7.64.0

    You can note that all this server application is doing is to respond with the content of the request it receives.

    Now replace the svc.cluster.local suffix by traefik.mesh, and tada: you are now using Traefik Mesh to reach your server!

    Command

    1. curl server.test.traefik.mesh

    Expected Output

    Note the presence of headers as well as other instrumentation headers like Uber-Trace-Id, indicating than your request has been processed and instrumented by Traefik Mesh.