Mutual TLS Migration

This task shows how to migrate your existing Istio services’ traffic from plain text to mutual TLS without breaking live traffic.

In the scenario where there are many services communicating over the network, it may be desirable to gradually migrate them to Istio. During the migration, some services have Envoy sidecars while some do not. For a service with a sidecar, if you enable mutual TLS on the service, the connections from legacy clients (i.e., clients without Envoy) will lose communication since they do not have Envoy sidecars and client certificates. To solve this issue, Istio authentication policy provides a PERMISSIVE mode to solve this problem. When PERMISSIVE mode is enabled, a service can take both HTTP and mutual TLS traffic.

You can configure Istio services to send mutual TLS traffic to that service while connections from legacy services will not lose communication. Moreover, you can use the Grafana dashboard to check which services are still sending plaintext traffic to the service in PERMISSIVE mode and choose to lock them down once the migration is done.

Before you begin

  • Understand Istio authentication policy and related mutual TLS authentication concepts.

  • Read the authentication policy task to learn how to configure authentication policy.

  • Have a Kubernetes cluster with Istio installed, without global mutual TLS enabled (e.g use the demo configuration profile as described in installation steps, or set the global.mtls.enabled installation option to false).

  • Ensure that your cluster is in PERMISSIVE mode before migrating to mutual TLS. Run the following command to check:

    $ kubectl get meshpolicy default -o yaml
    ...
    spec:
      peers:
      - mtls:
          mode: PERMISSIVE
    

    If the output is the same as above, you can skip the next step.

  • Run the following command to enable PERMISSIVE mode for the cluster. In general, this operation does not cause any interruption to your workloads, but note the warning message below.

    $ kubectl apply -f - <<EOF
    apiVersion: "authentication.istio.io/v1alpha1"
    kind: "MeshPolicy"
    metadata:
      name: "default"
    spec:
      peers:
      - mtls:
          mode: PERMISSIVE
    EOF
    

The rest of this task is divided into two parts.

Option 1: gradually enable mutual TLS for services

In this section, you can try out the migration process by creating sample workloads and modifying the policies to enforce STRICT mutual TLS between the workloads.

Set up the cluster

  • Create the following namespaces and deploy httpbin and sleep with sidecars on both of them.

    • foo
    • bar
  • Create the following namespace and deploy sleep without a sidecar

    • legacy
    ZipZipZipZipZip
    $ kubectl create ns foo
    $ kubectl apply -f <(istioctl kube-inject -f @samples/httpbin/httpbin.yaml@) -n foo
    $ kubectl apply -f <(istioctl kube-inject -f @samples/sleep/sleep.yaml@) -n foo
    $ kubectl create ns bar
    $ kubectl apply -f <(istioctl kube-inject -f @samples/httpbin/httpbin.yaml@) -n bar
    $ kubectl apply -f <(istioctl kube-inject -f @samples/sleep/sleep.yaml@) -n bar
    $ kubectl create ns legacy
    $ kubectl apply -f @samples/sleep/sleep.yaml@ -n legacy
    
  • Verify setup by sending an http request (using curl command) from any sleep pod (among those in namespace foo, bar or legacy) to httpbin.foo. All requests should success with HTTP code 200.

    $ for from in "foo" "bar" "legacy"; do kubectl exec $(kubectl get pod -l app=sleep -n ${from} -o jsonpath={.items..metadata.name}) -c sleep -n ${from} -- curl http://httpbin.foo:8000/ip -s -o /dev/null -w "sleep.${from} to httpbin.foo: %{http_code}\n"; done
    sleep.foo to httpbin.foo: 200
    sleep.bar to httpbin.foo: 200
    sleep.legacy to httpbin.foo: 200
    
  • Also verify that there are no authentication policies or destination rules (except control plane’s) in the system:

    $ kubectl get policies.authentication.istio.io --all-namespaces
    NAMESPACE      NAME                          AGE
    istio-system   grafana-ports-mtls-disabled   3m
    
    $ kubectl get destinationrule --all-namespaces
    NAMESPACE      NAME              AGE
    istio-system   istio-policy      25m
    istio-system   istio-telemetry   25m
    

Configure clients to send mutual TLS traffic

Configure Istio services to send mutual TLS traffic by setting DestinationRule.

$ cat <<EOF | kubectl apply -n foo -f -
apiVersion: "networking.istio.io/v1alpha3"
kind: "DestinationRule"
metadata:
  name: "example-httpbin-istio-client-mtls"
spec:
  host: httpbin.foo.svc.cluster.local
  trafficPolicy:
    tls:
      mode: ISTIO_MUTUAL
EOF

sleep.foo and sleep.bar should start sending mutual TLS traffic to httpbin.foo. And sleep.legacy still sends plaintext traffic to httpbin.foo since it does not have sidecar thus DestinationRule does not apply.

Now we confirm all requests to httpbin.foo still succeed.

$ for from in "foo" "bar" "legacy"; do kubectl exec $(kubectl get pod -l app=sleep -n ${from} -o jsonpath={.items..metadata.name}) -c sleep -n ${from} -- curl http://httpbin.foo:8000/ip -s -o /dev/null -w "sleep.${from} to httpbin.foo: %{http_code}\n"; done
200
200
200

You can also specify a subset of the clients’ request to use ISTIO_MUTUAL mutual TLS in DestinationRule. After verifying it works by checking Grafana to monitor, then increase the rollout scope and finally apply to all Istio client services.

Lock down to mutual TLS

After migrating all clients to Istio services and injecting the Envoy sidecar, we can lock down the httpbin.foo to only accept mutual TLS traffic.

$ cat <<EOF | kubectl apply -n foo -f -
apiVersion: "authentication.istio.io/v1alpha1"
kind: "Policy"
metadata:
  name: "example-httpbin-strict"
  namespace: foo
spec:
  targets:
  - name: httpbin
  peers:
  - mtls:
      mode: STRICT
EOF

Now, you should see the request from sleep.legacy fail.

$ for from in "foo" "bar" "legacy"; do kubectl exec $(kubectl get pod -l app=sleep -n ${from} -o jsonpath={.items..metadata.name}) -c sleep -n ${from} -- curl http://httpbin.foo:8000/ip -s -o /dev/null -w "sleep.${from} to httpbin.foo: %{http_code}\n"; done
200
200
503

If you can’t migrate all your services to Istio (injecting Envoy sidecar), you have to stay at PERMISSIVE mode. However, when configured with PERMISSIVE mode, no authentication or authorization checks will be performed for plaintext traffic by default. We recommend you use Istio Authorization to configure different paths with different authorization policies.

Clean up the example

To remove all resources created in this section:

$ kubectl delete ns foo bar legacy
Namespaces foo bar legacy deleted.

Option 2: globally enable mutual TLS for the cluster

This section describes how to apply the configuration to enforce mutual TLS for a cluster. For more details, see the Authentication policy task.

Configure all clients to send mutual TLS traffic

Run the following command to enable all Envoy sidecars to send mutual TLS traffic to the servers.

$ kubectl apply -f - <<EOF
apiVersion: "networking.istio.io/v1alpha3"
kind: "DestinationRule"
metadata:
  name: "default"
  namespace: "istio-system"
spec:
  host: "*.local"
  trafficPolicy:
    tls:
      mode: ISTIO_MUTUAL
EOF

The connections between services should not be interrupted.

Lock down to mutual TLS for the entire cluster

Run the following command to enforce all Envoy sidecars to only receive mutual TLS traffic.

$ kubectl apply -f - <<EOF
apiVersion: "authentication.istio.io/v1alpha1"
kind: "MeshPolicy"
metadata:
  name: "default"
spec:
  peers:
  - mtls: {}
EOF

The connections between services should not be interrupted.

Clean up global mutual TLS configuration

Choose one of the following, depending on the mode you want to switch to:

  • To switch to PERMISSIVE mode for the cluster:

    $ kubectl apply -f - <<EOF
    apiVersion: "authentication.istio.io/v1alpha1"
    kind: "MeshPolicy"
    metadata:
      name: "default"
    spec:
      peers:
      - mtls:
          mode: PERMISSIVE
    EOF
    
  • To disable the global mutual TLS and switch to plaintext only:

    $ kubectl delete meshpolicy default
    $ kubectl delete destinationrule default -n istio-system
    
Was this information useful?
Do you have any suggestions for improvement?

Thanks for your feedback!