Multi-network Mesh Expansion

This example provides instructions to integrate VMs and bare metal hosts into an Istio mesh deployed on Kubernetes with gateways. No VPN connectivity nor direct network access between workloads in VMs, bare metals and clusters is required.


  • One or more Kubernetes clusters with versions: 1.13, 1.14, 1.15.

  • Mesh expansion machines must have IP connectivity to the Ingress gateways in the mesh.

  • Install the Helm client. Helm is needed to enable mesh expansion.

Installation steps

Setup consists of preparing the mesh for expansion and installing and configuring each VM.

Customized installation of Istio on the Cluster

The first step when adding non-Kubernetes services to an Istio mesh is to configure the Istio installation itself, and generate the configuration files that let mesh expansion VMs connect to the mesh. To prepare the cluster for mesh expansion, run the following commands on a machine with cluster admin privileges:

  1. Generate a meshexpansion-gateways Istio configuration file using helm:

    $ helm template install/kubernetes/helm/istio --name istio --namespace istio-system \
        -f \ > $HOME/istio-mesh-expansion-gatways.yaml

    For further details and customization options, refer to the Installation with Helm instructions.

  2. Deploy Istio control plane into the cluster

    $ kubectl create namespace istio-system
    $ helm template  install/kubernetes/helm/istio-init --name istio-init --namespace istio-system  | kubectl apply -f -
    $ kubectl apply -f $HOME/istio-mesh-expansion-gatways.yaml
  3. Verify Istio is installed successfully

    $ istioctl verify-install -f $HOME/istio-mesh-expansion-gatways.yaml
  4. Create vm namespace for the VM services.

    $ kubectl create ns vm
  5. Define the namespace the VM joins. This example uses the SERVICE_NAMESPACE environment variable to store the namespace. The value of this variable must match the namespace you use in the configuration files later on.

    $ export SERVICE_NAMESPACE="vm"
  6. Extract the initial keys the service account needs to use on the VMs.

    $ kubectl -n $SERVICE_NAMESPACE get secret istio.default  \
        -o jsonpath='{.data.root-cert\.pem}' | base64 --decode > root-cert.pem
    $ kubectl -n $SERVICE_NAMESPACE get secret istio.default  \
        -o jsonpath='{.data.key\.pem}' | base64 --decode > key.pem
    $ kubectl -n $SERVICE_NAMESPACE get secret istio.default  \
          -o jsonpath='{.data.cert-chain\.pem}' | base64 --decode > cert-chain.pem
  7. Determine and store the IP address of the Istio ingress gateway since the mesh expansion machines access Citadel and Pilot and workloads on cluster through this IP address.

    $ export GWIP=$(kubectl get -n istio-system service istio-ingressgateway -o jsonpath='{.status.loadBalancer.ingress[0].ip}')
    $ echo $GWIP
  8. Generate a cluster.env configuration to deploy in the VMs. This file contains the Kubernetes cluster IP address ranges to intercept and redirect via Envoy.

  9. Check the contents of the generated cluster.env file. It should be similar to the following example:

    $ cat cluster.env

Setup DNS

Providing DNS resolution to allow services running on VM can access the services running in the cluster. Istio itself does not use the DNS for routing requests between services. Services local to a cluster share a common DNS suffix(e.g., svc.cluster.local). Kubernetes DNS provides DNS resolution for these services.

To provide a similar setup to allow services accessible from VMs, you name services in the clusters in the format <name>.<namespace>.global. Istio also ships with a CoreDNS server that will provide DNS resolution for these services. In order to utilize this DNS, Kubernetes’ DNS must be configured to stub a domain for .global.

Create one of the following ConfigMaps, or update an existing one, in each cluster that will be calling services in remote clusters (every cluster in the general case):

For clusters that use kube-dns:

$ kubectl apply -f - <<EOF
apiVersion: v1
kind: ConfigMap
  name: kube-dns
  namespace: kube-system
  stubDomains: |
    {"global": ["$(kubectl get svc -n istio-system istiocoredns -o jsonpath={.spec.clusterIP})"]}

For clusters that use CoreDNS:

$ kubectl apply -f - <<EOF
apiVersion: v1
kind: ConfigMap
  name: coredns
  namespace: kube-system
  Corefile: |
    .:53 {
        kubernetes cluster.local {
           pods insecure
        prometheus :9153
        proxy . /etc/resolv.conf
        cache 30
    global:53 {
        cache 30
        proxy . $(kubectl get svc -n istio-system istiocoredns -o jsonpath={.spec.clusterIP})

Setting up the VM

Next, run the following commands on each machine that you want to add to the mesh:

  1. Copy the previously created cluster.env and *.pem files to the VM.

  2. Install the Debian package with the Envoy sidecar.

    $ curl -L > istio-sidecar.deb
    $ sudo dpkg -i istio-sidecar.deb
  3. Add the IP address of the Istio gateway to /etc/hosts. Revisit the Customized installation of Istio on the Cluster section to learn how to obtain the IP address. The following example updates the /etc/hosts file with the Istio gateway address:

    $ echo " istio-citadel istio-pilot istio-pilot.istio-system" | sudo tee -a /etc/hosts
  4. Install root-cert.pem, key.pem and cert-chain.pem under /etc/certs/.

    $ sudo mkdir -p /etc/certs
    $ sudo cp {root-cert.pem,cert-chain.pem,key.pem} /etc/certs
  5. Install cluster.env under /var/lib/istio/envoy/.

    $ sudo cp cluster.env /var/lib/istio/envoy
  6. Transfer ownership of the files in /etc/certs/ and /var/lib/istio/envoy/ to the Istio proxy.

    $ sudo chown -R istio-proxy /etc/certs /var/lib/istio/envoy
  7. Verify the node agent works:

    $ sudo node_agent
    CSR is approved successfully. Will renew cert in 1079h59m59.84568493s
  8. Start Istio using systemctl.

    $ sudo systemctl start istio-auth-node-agent
    $ sudo systemctl start istio

Added Istio resources

Below Istio resources are added to support Mesh Expansion with gateways. This released the flat network requirement between the VM and cluster.

Resource KindResource NameFunction
configmapcorednsSend *.global request to istiocordns service
serviceistiocorednsResolve *.global to Istio Ingress gateway
gateway.networking.istio.iomeshexpansion-gatewayOpen port for Pilot, Citadel and Mixer
gateway.networking.istio.ioistio-multicluster-egressgatewayOpen port 15443 for outbound *.global traffic
gateway.networking.istio.ioistio-multicluster-ingressgatewayOpen port 15443 for inbound *.global traffic
envoyfilter.networking.istio.ioistio-multicluster-ingressgatewayTransform *.global to *. svc.cluster.local
destinationrule.networking.istio.ioistio-multicluster-destinationruleSet traffic policy for 15443 traffic
destinationrule.networking.istio.iomeshexpansion-dr-pilotSet traffic policy for istio-pilot
destinationrule.networking.istio.ioistio-policySet traffic policy for istio-policy
destinationrule.networking.istio.ioistio-telemetrySet traffic policy for istio-telemetry
virtualservice.networking.istio.iomeshexpansion-vs-pilotSet route info for istio-pilot
virtualservice.networking.istio.iomeshexpansion-vs-citadelSet route info for istio-citadel

Expose service running on cluster to VMs

Every service in the cluster that needs to be accessed from the VM requires a service entry configuration in the cluster. The host used in the service entry should be of the form where name and namespace correspond to the service’s name and namespace respectively.

To demonstrate access from VM to cluster services, configure the the httpbin service in the cluster.

  1. Deploy the httpbin service in the cluster

    $ kubectl create namespace bar
    $ kubectl label namespace bar istio-injection=enabled
    $ kubectl apply -n bar -f @samples/httpbin/httpbin.yaml@
  2. Create a service entry for the httpbin service in the cluster.

    To allow services in VM to access httpbin in the cluster, we need to create a service entry for it. The host name of the service entry should be of the form <name>.<namespace>.global where name and namespace correspond to the remote service’s name and namespace respectively.

    For DNS resolution for services under the *.global domain, you need to assign these services an IP address.

    If the global services have actual VIPs, you can use those, but otherwise we suggest using IPs from the loopback range that are not already allocated. These IPs are non-routable outside of a pod. In this example we’ll use IPs in which avoids conflicting with well known IPs such as (localhost). Application traffic for these IPs will be captured by the sidecar and routed to the appropriate remote service.

    $ kubectl apply  -n bar -f - <<EOF
    kind: ServiceEntry
      # must be of form
      location: MESH_INTERNAL
      - name: http1
        number: 8000
        protocol: http
      resolution: DNS
      # the IP address to which will resolve to
      # must be unique for each service, within a given cluster.
      # This address need not be routable. Traffic for this IP will be captured
      # by the sidecar and routed appropriately.
      # This address will also be added into VM's /etc/hosts
      # This is the routable address of the ingress gateway in the cluster.
      # Traffic from the VMs will be
      # routed to this address.
      - address: ${CLUSTER_GW_ADDR}
          http1: 15443 # Do not change this port value

    The configurations above will result in all traffic from VMs for on any port to be routed to the endpoint <IPofClusterIngressGateway>:15443 over a mutual TLS connection.

    The gateway for port 15443 is a special SNI-aware Envoy preconfigured and installed as part of the meshexpansion with gateway Istio installation step in the Customized installation of Istio on the Cluster section. Traffic entering port 15443 will be load balanced among pods of the appropriate internal service of the target cluster (in this case, in the cluster).

Send requests from VM to Kubernetes services

After setup, the machine can access services running in the Kubernetes cluster.

The following example shows accessing a service running in the Kubernetes cluster from a mesh expansion VM using /etc/hosts/, in this case using a service from the httpbin service.

  1. On the mesh expansion machine, add the service name and address to its /etc/hosts file. You can then connect to the cluster service from the VM, as in the example below:

    $ echo "" | sudo tee -a /etc/hosts
    $ curl -v
    < HTTP/1.1 200 OK
    < server: envoy
    < content-type: text/html; charset=utf-8
    < content-length: 9593
    ... html content ...

The server: envoy header indicates that the sidecar intercepted the traffic.

Running services on a mesh expansion machine

  1. Setup an HTTP server on the VM instance to serve HTTP traffic on port 8888:

    $ python -m SimpleHTTPServer 8888
  2. Determine the VM instance’s IP address.

  3. Configure a service entry to enable service discovery for the VM. You can add VM services to the mesh using a service entry. Service entries let you manually add additional services to Pilot’s abstract model of the mesh. Once VM services are part of the mesh’s abstract model, other services can find and direct traffic to them. Each service entry configuration contains the IP addresses, ports, and appropriate labels of all VMs exposing a particular service, for example:

    $ kubectl -n ${SERVICE_NAMESPACE} apply -f - <<EOF
    kind: ServiceEntry
      name: vmhttp
      - vmhttp.${SERVICE_NAMESPACE}.svc.cluster.local
      - number: 8888
        name: http
        protocol: HTTP
      resolution: STATIC
      - address: ${VM_IP}
          http: 8888
          app: vmhttp
          version: "v1"
  4. The workloads in a Kubernetes cluster need a DNS mapping to resolve the domain names of VM services. To integrate the mapping with your own DNS system, use istioctl register and creates a Kubernetes selector-less service, for example:

    $ istioctl  register -n ${SERVICE_NAMESPACE} vmhttp ${VM_IP} 8888
  5. Deploy a pod running the sleep service in the Kubernetes cluster, and wait until it is ready:

    $ kubectl apply -f @samples/sleep/sleep.yaml@
    $ kubectl get pod
    NAME                             READY     STATUS    RESTARTS   AGE
    productpage-v1-8fcdcb496-xgkwg   2/2       Running   0          1d
    sleep-88ddbcfdd-rm42k            2/2       Running   0          1s
  6. Send a request from the sleep service on the pod to the VM’s HTTP service:

    $ kubectl exec -it sleep-88ddbcfdd-rm42k -c sleep -- curl vmhttp.${SERVICE_NAMESPACE}.svc.cluster.local:8888

    If configured properly, you will see something similar to the output below.

    <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 3.2 Final//EN"><html>
    <title>Directory listing for /</title>
    <h2>Directory listing for /</h2>
    <li><a href=".bashrc">.bashrc</a></li>
    <li><a href=".ssh/">.ssh/</a></li>

Congratulations! You successfully configured a service running in a pod within the cluster to send traffic to a service running on a VM outside of the cluster and tested that the configuration worked.


Run the following commands to remove the expansion VM from the mesh’s abstract model.

$ istioctl deregister -n ${SERVICE_NAMESPACE} vmhttp ${VM_IP}
2019-02-21T22:12:22.023775Z     info    Deregistered service successfull
$ kubectl delete ServiceEntry vmhttp -n ${SERVICE_NAMESPACE} "vmhttp" deleted