Cluster-Aware Service Routing

9 minute read

This example shows how to configure a multicluster mesh with a single control plane topology and using Istio’s Split-horizon EDS (Endpoints Discovery Service) feature (introduced in Istio 1.1) to route service requests to other clusters via their ingress gateway. Split-horizon EDS enables Istio to route requests to different endpoints, depending on the location of the request source.

By following the instructions in this example, you will setup a two cluster mesh as shown in the following diagram:

Single Istio control plane topology spanning multiple Kubernetes clusters with Split-horizon EDS configured

The primary cluster, cluster1, runs the full set of Istio control plane components while cluster2 only runs Istio Citadel, Sidecar Injector, and Ingress gateway. No VPN connectivity nor direct network access between workloads in different clusters is required.

Before you begin

In addition to the prerequisites for installing Istio, the following is required for this example:

Two Kubernetes clusters (referred to as cluster1 and cluster2).
The Kubernetes API server of cluster2 MUST be accessible from cluster1 in order to run this configuration.

The kubectl command is used to access both the cluster1 and cluster2 clusters with the --context flag. Use the following command to list your contexts:

$ kubectl config get-contexts
CURRENT   NAME       CLUSTER    AUTHINFO       NAMESPACE
*         cluster1   cluster1   user@foo.com   default
          cluster2   cluster2   user@foo.com   default

Export the following environment variables with the context names of your configuration:

$ export CTX_CLUSTER1=<KUBECONFIG_CONTEXT_NAME_FOR_CLUSTER_1>
$ export CTX_CLUSTER2=<KUBECONFIG_CONTEXT_NAME_FOR_CLUSTER_2>

Example multicluster setup

In this example you will install Istio with mutual TLS enabled for both the control plane and application pods. For the shared root CA, you create a cacerts secret on both cluster1 and cluster2 clusters using the same Istio certificate from the Istio samples directory.

The instructions, below, also set up cluster2 with a selector-less service and an endpoint for istio-pilot.istio-system that has the address of cluster1 Istio ingress gateway. This will be used to access pilot on cluster1 securely using the ingress gateway without mutual TLS termination.

Setup cluster 1 (primary)

Use Helm to create the Istio deployment YAML for cluster1:

$ helm template --name=istio --namespace=istio-system \
  --set global.mtls.enabled=true \
  --set security.selfSigned=false \
  --set global.controlPlaneSecurityEnabled=true \
  --set global.proxy.accessLogFile="/dev/stdout" \
  --set global.meshExpansion.enabled=true \
  --set 'global.meshNetworks.network2.endpoints[0].fromRegistry'=n2-k8s-config \
  --set 'global.meshNetworks.network2.gateways[0].address'=0.0.0.0 \
  --set 'global.meshNetworks.network2.gateways[0].port'=443 \
  install/kubernetes/helm/istio > istio-auth.yaml

Deploy Istio to cluster1:

$ kubectl create --context=$CTX_CLUSTER1 ns istio-system
$ kubectl create --context=$CTX_CLUSTER1 secret generic cacerts -n istio-system --from-file=samples/certs/ca-cert.pem --from-file=samples/certs/ca-key.pem --from-file=samples/certs/root-cert.pem --from-file=samples/certs/cert-chain.pem
$ for i in install/kubernetes/helm/istio-init/files/crd*yaml; do kubectl apply --context=$CTX_CLUSTER1 -f $i; done
$ kubectl create --context=$CTX_CLUSTER1 -f istio-auth.yaml

Wait for the Istio pods on cluster1 to become ready:

$ kubectl get pods --context=$CTX_CLUSTER1 -n istio-system
NAME                                      READY   STATUS      RESTARTS   AGE
istio-citadel-9bbf9b4c8-nnmbt             1/1     Running     0          2m8s
istio-cleanup-secrets-1.1.0-x9crw         0/1     Completed   0          2m12s
istio-galley-868c5fff5d-9ph6l             1/1     Running     0          2m9s
istio-ingressgateway-6c756547b-dwc78      1/1     Running     0          2m8s
istio-pilot-54fcf8db8-sn9cn               2/2     Running     0          2m8s
istio-policy-5fcbd55d8b-xhbpz             2/2     Running     2          2m8s
istio-security-post-install-1.1.0-ww5zz   0/1     Completed   0          2m12s
istio-sidecar-injector-6dcc9d5c64-7hnnl   1/1     Running     0          2m8s
istio-telemetry-57875ffb6d-n2vmf          2/2     Running     3          2m8s
prometheus-66c9f5694-8pccr                1/1     Running     0          2m8s

Create an ingress gateway to access service(s) in cluster2:
```
$ kubectl create --context=$CTX_CLUSTER1 -f - <<EOF
apiVersion: networking.istio.io/v1alpha3
kind: Gateway
metadata:
  name: cluster-aware-gateway
  namespace: istio-system
spec:
  selector:
    istio: ingressgateway
  servers:
  - port:
      number: 443
      name: tls
      protocol: TLS
    tls:
      mode: AUTO_PASSTHROUGH
    hosts:
    - "*.local"
EOF
```
This Gateway configures 443 port to pass incoming traffic through to the target service specified in a request’s SNI header, for SNI values of the local top-level domain (i.e., the Kubernetes DNS domain). Mutual TLS connections will be used all the way from the source to the destination sidecar.
Although applied to cluster1, this Gateway instance will also affect cluster2 because both clusters communicate with the same Pilot.

Setup cluster 2

Export the cluster1 gateway address:
```
$ export LOCAL_GW_ADDR=$(kubectl get --context=$CTX_CLUSTER1 svc --selector=app=istio-ingressgateway \
    -n istio-system -o jsonpath='{.items[0].status.loadBalancer.ingress[0].ip}') && echo ${LOCAL_GW_ADDR}
```
This command sets the value to the gateway’s public IP and displays it.
The command fails if the load balancer configuration doesn’t include an IP address. The implementation of DNS name support is pending.

Use Helm to create the Istio deployment YAML for cluster2:

$ helm template --name istio-remote --namespace=istio-system \
  --values @install/kubernetes/helm/istio/values-istio-remote.yaml@ \
  --set global.mtls.enabled=true \
  --set gateways.enabled=true \
  --set security.selfSigned=false \
  --set global.controlPlaneSecurityEnabled=true \
  --set global.createRemoteSvcEndpoints=true \
  --set global.remotePilotCreateSvcEndpoint=true \
  --set global.remotePilotAddress=${LOCAL_GW_ADDR} \
  --set global.remotePolicyAddress=${LOCAL_GW_ADDR} \
  --set global.remoteTelemetryAddress=${LOCAL_GW_ADDR} \
  --set gateways.istio-ingressgateway.env.ISTIO_META_NETWORK="network2" \
  --set global.network="network2" \
  install/kubernetes/helm/istio > istio-remote-auth.yaml

Deploy Istio to cluster2:

$ kubectl create --context=$CTX_CLUSTER2 ns istio-system
$ kubectl create --context=$CTX_CLUSTER2 secret generic cacerts -n istio-system --from-file=samples/certs/ca-cert.pem --from-file=samples/certs/ca-key.pem --from-file=samples/certs/root-cert.pem --from-file=samples/certs/cert-chain.pem
$ kubectl create --context=$CTX_CLUSTER2 -f istio-remote-auth.yaml

Wait for the Istio pods on cluster2, except for istio-ingressgateway, to become ready:

$ kubectl get pods --context=$CTX_CLUSTER2 -n istio-system -l istio!=ingressgateway
NAME                                     READY   STATUS      RESTARTS   AGE
istio-citadel-75c8fcbfcf-9njn6           1/1     Running     0          12s
istio-cleanup-secrets-1.1.0-vtp62        0/1     Completed   0          14s
istio-sidecar-injector-cdb5d4dd5-rhks9   1/1     Running     0          12s

Determine the ingress IP and port for cluster2.
1. Set the current context of kubectl to CTX_CLUSTER2
```
$ export ORIGINAL_CONTEXT=$(kubectl config current-context)
$ kubectl config use-context $CTX_CLUSTER2
```
2. Follow the instructions in Determining the ingress IP and ports, to set the INGRESS_HOST and SECURE_INGRESS_PORT environment variables.
3. Restore the previous kubectl context:
```
$ kubectl config use-context $ORIGINAL_CONTEXT
$ unset ORIGINAL_CONTEXT
```
4. Print the values of INGRESS_HOST and SECURE_INGRESS_PORT:
```
$ echo The ingress gateway of cluster2: address=$INGRESS_HOST, port=$SECURE_INGRESS_PORT
```
Update the gateway address in the mesh network configuration. Edit the istio ConfigMap:
```
$ kubectl edit cm -n istio-system --context=$CTX_CLUSTER1 istio
```
Update the gateway’s address and port of network2 to reflect the cluster2 ingress host and port, respectively, then save and quit.
Once saved, Pilot will automatically read the updated network configuration.

Prepare environment variables for building the n2-k8s-config file for the service account istio-multi:

$ CLUSTER_NAME=$(kubectl --context=$CTX_CLUSTER2 config view --minify=true -o jsonpath='{.clusters[].name}')
$ SERVER=$(kubectl --context=$CTX_CLUSTER2 config view --minify=true -o jsonpath='{.clusters[].cluster.server}')
$ SECRET_NAME=$(kubectl --context=$CTX_CLUSTER2 get sa istio-multi -n istio-system -o jsonpath='{.secrets[].name}')
$ CA_DATA=$(kubectl get --context=$CTX_CLUSTER2 secret ${SECRET_NAME} -n istio-system -o jsonpath="{.data['ca\.crt']}")
$ TOKEN=$(kubectl get --context=$CTX_CLUSTER2 secret ${SECRET_NAME} -n istio-system -o jsonpath="{.data['token']}" | base64 --decode)

Create the n2-k8s-config file in the working directory:

$ cat <<EOF > n2-k8s-config
apiVersion: v1
kind: Config
clusters:
  - cluster:
      certificate-authority-data: ${CA_DATA}
      server: ${SERVER}
    name: ${CLUSTER_NAME}
contexts:
  - context:
      cluster: ${CLUSTER_NAME}
      user: ${CLUSTER_NAME}
    name: ${CLUSTER_NAME}
current-context: ${CLUSTER_NAME}
users:
  - name: ${CLUSTER_NAME}
    user:
      token: ${TOKEN}
EOF

Start watching cluster 2

Execute the following commands to add and label the secret of the cluster2 Kubernetes. After executing these commands Istio Pilot on cluster1 will begin watching cluster2 for services and instances, just as it does for cluster1.
```
$ kubectl create --context=$CTX_CLUSTER1 secret generic n2-k8s-secret --from-file n2-k8s-config -n istio-system
$ kubectl label --context=$CTX_CLUSTER1 secret n2-k8s-secret istio/multiCluster=true -n istio-system
```

Wait for istio-ingressgateway to become ready:

$ kubectl get pods --context=$CTX_CLUSTER2 -n istio-system -l istio=ingressgateway
NAME                                    READY     STATUS    RESTARTS   AGE
istio-ingressgateway-5c667f4f84-bscff   1/1       Running   0          16m

Now that you have your cluster1 and cluster2 clusters set up, you can deploy an example service.

Example service

In this demo you will see how traffic to a service can be distributed across the two clusters. As shown in the diagram, above, you will deploy two instances of the helloworld service, one on cluster1 and one on cluster2. The difference between the two instances is the version of their helloworld image.

Deploy helloworld v2 in cluster 2

Create a sample namespace with a sidecar auto-injection label:

$ kubectl create --context=$CTX_CLUSTER2 ns sample
$ kubectl label --context=$CTX_CLUSTER2 namespace sample istio-injection=enabled

Deploy helloworld v2:

$ kubectl create --context=$CTX_CLUSTER2 -f @samples/helloworld/helloworld.yaml@ -l app=helloworld -n sample
$ kubectl create --context=$CTX_CLUSTER2 -f @samples/helloworld/helloworld.yaml@ -l version=v2 -n sample

Confirm helloworld v2 is running:

$ kubectl get po --context=$CTX_CLUSTER2 -n sample
NAME                             READY     STATUS    RESTARTS   AGE
helloworld-v2-7dd57c44c4-f56gq   2/2       Running   0          35s

Deploy helloworld v1 in cluster 1

Create a sample namespace with a sidecar auto-injection label:

$ kubectl create --context=$CTX_CLUSTER1 ns sample
$ kubectl label --context=$CTX_CLUSTER1 namespace sample istio-injection=enabled

Deploy helloworld v1:

$ kubectl create --context=$CTX_CLUSTER1 -f @samples/helloworld/helloworld.yaml@ -l app=helloworld -n sample
$ kubectl create --context=$CTX_CLUSTER1 -f @samples/helloworld/helloworld.yaml@ -l version=v1 -n sample

Confirm helloworld v1 is running:

$ kubectl get po --context=$CTX_CLUSTER1 -n sample
NAME                            READY     STATUS    RESTARTS   AGE
helloworld-v1-d4557d97b-pv2hr   2/2       Running   0          40s

Split-horizon EDS in action

We will call the helloworld.sample service from another in-mesh sleep service.

Deploy the sleep service:

$ kubectl create --context=$CTX_CLUSTER1 -f @samples/sleep/sleep.yaml@ -n sample

Wait for the sleep service to start:

$ kubectl get po --context=$CTX_CLUSTER1 -n sample -l app=sleep
sleep-754684654f-n6bzf           2/2     Running   0          5s

Call the helloworld.sample service several times:

$ kubectl exec --context=$CTX_CLUSTER1 -it -n sample -c sleep $(kubectl get pod --context=$CTX_CLUSTER1 -n sample -l app=sleep -o jsonpath='{.items[0].metadata.name}') -- curl helloworld.sample:5000/hello

If set up correctly, the traffic to the helloworld.sample service will be distributed between instances on cluster1 and cluster2 resulting in responses with either v1 or v2 in the body:

Hello version: v2, instance: helloworld-v2-758dd55874-6x4t8

Hello version: v1, instance: helloworld-v1-86f77cd7bd-cpxhv

You can also verify the IP addresses used to access the endpoints by printing the log of the sleep’s istio-proxy container.

$ kubectl logs --context=$CTX_CLUSTER1 -n sample $(kubectl get pod --context=$CTX_CLUSTER1 -n sample -l app=sleep -o jsonpath='{.items[0].metadata.name}') istio-proxy
[2018-11-25T12:37:52.077Z] "GET /hello HTTP/1.1" 200 - 0 60 190 189 "-" "curl/7.60.0" "6e096efe-f550-4dfa-8c8c-ba164baf4679" "helloworld.sample:5000" "192.23.120.32:15443" outbound|5000||helloworld.sample.svc.cluster.local - 10.20.194.146:5000 10.10.0.89:59496 -
[2018-11-25T12:38:06.745Z] "GET /hello HTTP/1.1" 200 - 0 60 171 170 "-" "curl/7.60.0" "6f93c9cc-d32a-4878-b56a-086a740045d2" "helloworld.sample:5000" "10.10.0.90:5000" outbound|5000||helloworld.sample.svc.cluster.local - 10.20.194.146:5000 10.10.0.89:59646 -

The gateway IP, 192.23.120.32:15443, of cluster2 is logged when v2 was called and the instance IP, 10.10.0.90:5000, of cluster1 is logged when v1 was called.

Cleanup

Execute the following commands to clean up the demo services and the Istio components.

Cleanup the cluster2 cluster:

$ kubectl delete --context=$CTX_CLUSTER2 -f istio-remote-auth.yaml
$ kubectl delete --context=$CTX_CLUSTER2 ns istio-system
$ kubectl delete --context=$CTX_CLUSTER2 ns sample
$ unset CTX_CLUSTER2 CLUSTER_NAME SERVER SECRET_NAME CA_DATA TOKEN INGRESS_HOST SECURE_INGRESS_PORT INGRESS_PORT
$ rm istio-remote-auth.yaml

Cleanup the cluster1 cluster:

$ kubectl delete --context=$CTX_CLUSTER1 -f istio-auth.yaml
$ kubectl delete --context=$CTX_CLUSTER1 ns istio-system
$ for i in install/kubernetes/helm/istio-init/files/crd*yaml; do kubectl delete --context=$CTX_CLUSTER1 -f $i; done
$ kubectl delete --context=$CTX_CLUSTER1 ns sample
$ unset CTX_CLUSTER1
$ rm istio-auth.yaml n2-k8s-config