Authorization on Ingress Gateway

This task shows you how to enforce IP-based access control on an Istio ingress gateway using an authorization policy.

Before you begin

Before you begin this task, do the following:

  • Read the Authorization conceptual documentation.

  • Install Istio using the Istio installation guide.

  • Deploy a workload, httpbin in a namespace, for example foo, and expose it through the Istio ingress gateway with this command:

    ZipZip

    1. $ kubectl create ns foo
    2. $ kubectl apply -f <(istioctl kube-inject -f @samples/httpbin/httpbin.yaml@) -n foo
    3. $ kubectl apply -f <(istioctl kube-inject -f @samples/httpbin/httpbin-gateway.yaml@) -n foo

  • Turn on RBAC debugging in Envoy for the ingress gateway:

    1. $ kubectl get pods -n istio-system -o name -l istio=ingressgateway | sed 's|pod/||' | while read -r pod; do istioctl proxy-config log "$pod" -n istio-system --level rbac:debug; done

  • Follow the instructions in Determining the ingress IP and ports to define the INGRESS_HOST and INGRESS_PORT environment variables.

  • Verify that the httpbin workload and ingress gateway are working as expected using this command:

    1. $ curl "$INGRESS_HOST:$INGRESS_PORT"/headers -s -o /dev/null -w "%{http_code}\n"
    2. 200

If you don’t see the expected output, retry after a few seconds. Caching and propagation overhead can cause a delay.

Getting traffic into Kubernetes and Istio

All methods of getting traffic into Kubernetes involve opening a port on all worker nodes. The main features that accomplish this are the NodePort service and the LoadBalancer service. Even the Kubernetes Ingress resource must be backed by an Ingress controller that will create either a NodePort or a LoadBalancer service.

  • A NodePort just opens up a port in the range 30000-32767 on each worker node and uses a label selector to identify which Pods to send the traffic to. You have to manually create some kind of load balancer in front of your worker nodes or use Round-Robin DNS.

  • A LoadBalancer is just like a NodePort, except it also creates an environment specific external load balancer to handle distributing traffic to the worker nodes. For example, in AWS EKS, the LoadBalancer service will create a Classic ELB with your worker nodes as targets. If your Kubernetes environment does not have a LoadBalancer implementation, then it will just behave like a NodePort. An Istio ingress gateway creates a LoadBalancer service.

What if the Pod that is handling traffic from the NodePort or LoadBalancer isn’t running on the worker node that received the traffic? Kubernetes has its own internal proxy called kube-proxy that receives the packets and forwards them to the correct node.

Source IP address of the original client

If a packet goes through an external proxy load balancer and/or kube-proxy, then the original source IP address of the client is lost. Below are some strategies for preserving the original client IP for logging or security purposes.

A critical bug has been identified in Envoy that the proxy protocol downstream address is restored incorrectly for non-HTTP connections.

Please DO NOT USE the remoteIpBlocks field and remote_ip attribute with proxy protocol on non-HTTP connections until a newer version of Istio is released with a proper fix.

Note that Istio doesn’t support the proxy protocol and it can be enabled only with the EnvoyFilter API and should be used at your own risk.

If you are using a TCP/UDP Proxy external load balancer (AWS Classic ELB), it can use the Proxy Protocol to embed the original client IP address in the packet data. Both the external load balancer and the Istio ingress gateway must support the proxy protocol for it to work. In Istio, you can enable it with an EnvoyFilter like below:

  1. apiVersion: networking.istio.io/v1alpha3
  2. kind: EnvoyFilter
  3. metadata:
  4.   name: proxy-protocol
  5.   namespace: istio-system
  6. spec:
  7.   configPatches:
  8.   - applyTo: LISTENER
  9.     patch:
  10.       operation: MERGE
  11.       value:
  12.         listener_filters:
  13.         - name: envoy.listener.proxy_protocol
  14.         - name: envoy.listener.tls_inspector
  15.   workloadSelector:
  16.     labels:
  17.       istio: ingressgateway

Here is a sample of the IstioOperator that shows how to configure the Istio ingress gateway on AWS EKS to support the Proxy Protocol:

  1. apiVersion: install.istio.io/v1alpha1
  2. kind: IstioOperator
  3. spec:
  4.   meshConfig:
  5.     accessLogEncoding: JSON
  6.     accessLogFile: /dev/stdout
  7.   components:
  8.     ingressGateways:
  9.     - enabled: true
  10.       k8s:
  11.         hpaSpec:
  12.           maxReplicas: 10
  13.           minReplicas: 5
  14.         serviceAnnotations:
  15.           service.beta.kubernetes.io/aws-load-balancer-access-log-emit-interval: "5"
  16.           service.beta.kubernetes.io/aws-load-balancer-access-log-enabled: "true"
  17.           service.beta.kubernetes.io/aws-load-balancer-access-log-s3-bucket-name: elb-logs
  18.           service.beta.kubernetes.io/aws-load-balancer-access-log-s3-bucket-prefix: k8sELBIngressGW
  19.           service.beta.kubernetes.io/aws-load-balancer-proxy-protocol: "*"
  20.         affinity:
  21.           podAntiAffinity:
  22.             preferredDuringSchedulingIgnoredDuringExecution:
  23.             - podAffinityTerm:
  24.                 labelSelector:
  25.                   matchLabels:
  26.                     istio: ingressgateway
  27.                 topologyKey: failure-domain.beta.kubernetes.io/zone
  28.               weight: 1
  29.       name: istio-ingressgateway

If you are using a TCP/UDP network load balancer that preserves the client IP address (AWS Network Load Balancer, GCP External Network Load Balancer, Azure Load Balancer) or you are using Round-Robin DNS, then you can also preserve the client IP inside Kubernetes by bypassing kube-proxy and preventing it from sending traffic to other nodes. However, you must run an ingress gateway pod on every node. If you don’t, then any node that receives traffic and doesn’t have an ingress gateway will drop the traffic. See Source IP for Services with Type=NodePort for more information. Update the ingress gateway to set externalTrafficPolicy: Local to preserve the original client source IP on the ingress gateway using the following command:

  1. $ kubectl patch svc istio-ingressgateway -n istio-system -p '{"spec":{"externalTrafficPolicy":"Local"}}'

If you are using an HTTP/HTTPS external load balancer (AWS ALB, GCP ), it can put the original client IP address in the X-Forwarded-For header. Istio can extract the client IP address from this header with some configuration. See Configuring Gateway Network Topology. Quick example if using a single load balancer in front of Kubernetes:

  1. apiVersion: install.istio.io/v1alpha1
  2. kind: IstioOperator
  3. spec:
  4.   meshConfig:
  5.     accessLogEncoding: JSON
  6.     accessLogFile: /dev/stdout
  7.     defaultConfig:
  8.       gatewayTopology:
  9.         numTrustedProxies: 1

For reference, here are the types of load balancers created by Istio with a LoadBalancer service on popular managed Kubernetes environments:

Cloud ProviderLoad Balancer NameLoad Balancer Type
AWS EKSClassic Elastic Load BalancerTCP Proxy
GCP GKETCP/UDP Network Load BalancerNetwork
Azure AKSAzure Load BalancerNetwork
DO DOKSLoad BalancerNetwork

You can instruct AWS EKS to create a Network Load Balancer when you install Istio by using a serviceAnnotation like below:

  1. apiVersion: install.istio.io/v1alpha1
  2. kind: IstioOperator
  3. spec:
  4.   meshConfig:
  5.     accessLogEncoding: JSON
  6.     accessLogFile: /dev/stdout
  7.   components:
  8.     ingressGateways:
  9.     - enabled: true
  10.       k8s:
  11.         hpaSpec:
  12.           maxReplicas: 10
  13.           minReplicas: 5
  14.         serviceAnnotations:
  15.           service.beta.kubernetes.io/aws-load-balancer-type: "nlb"

IP-based allow list and deny list

When to use ipBlocks vs. remoteIpBlocks: If you are using the X-Forwarded-For HTTP header or the Proxy Protocol to determine the original client IP address, then you should use remoteIpBlocks in your AuthorizationPolicy. If you are using externalTrafficPolicy: Local, then you should use ipBlocks in your AuthorizationPolicy.

Load Balancer TypeSource of Client IPipBlocks vs. remoteIpBlocks
TCP ProxyProxy ProtocolremoteIpBlocks
Networkpacket source addressipBlocks
HTTP/HTTPSX-Forwarded-ForremoteIpBlocks
  • The following command creates the authorization policy, ingress-policy, for the Istio ingress gateway. The following policy sets the action field to ALLOW to allow the IP addresses specified in the ipBlocks to access the ingress gateway. IP addresses not in the list will be denied. The ipBlocks supports both single IP address and CIDR notation.

Create the AuthorizationPolicy:

  1. $ kubectl apply -f - <<EOF
  2. apiVersion: security.istio.io/v1beta1
  3. kind: AuthorizationPolicy
  4. metadata:
  5.   name: ingress-policy
  6.   namespace: istio-system
  7. spec:
  8.   selector:
  9.     matchLabels:
  10.       app: istio-ingressgateway
  11.   action: ALLOW
  12.   rules:
  13.   - from:
  14.     - source:
  15.         ipBlocks: ["1.2.3.4", "5.6.7.0/24"]
  16. EOF
  1. $ kubectl apply -f - <<EOF
  2. apiVersion: security.istio.io/v1beta1
  3. kind: AuthorizationPolicy
  4. metadata:
  5.   name: ingress-policy
  6.   namespace: istio-system
  7. spec:
  8.   selector:
  9.     matchLabels:
  10.       app: istio-ingressgateway
  11.   action: ALLOW
  12.   rules:
  13.   - from:
  14.     - source:
  15.         remoteIpBlocks: ["1.2.3.4", "5.6.7.0/24"]
  16. EOF

  • Verify that a request to the ingress gateway is denied:

    1. $ curl "$INGRESS_HOST:$INGRESS_PORT"/headers -s -o /dev/null -w "%{http_code}\n"
    2. 403

  • Update the ingress-policy to include your client IP address:

Find your original client IP address if you don’t know it and assign it to a variable:

  1. $ CLIENT_IP=$(kubectl get pods -n istio-system -o name -l istio=ingressgateway | sed 's|pod/||' | while read -r pod; do kubectl logs "$pod" -n istio-system | grep remoteIP; done | tail -1 | awk -F, '{print $3}' | awk -F: '{print $2}' | sed 's/ //') && echo "$CLIENT_IP"
  2. 192.168.10.15
  1. $ kubectl apply -f - <<EOF
  2. apiVersion: security.istio.io/v1beta1
  3. kind: AuthorizationPolicy
  4. metadata:
  5.   name: ingress-policy
  6.   namespace: istio-system
  7. spec:
  8.   selector:
  9.     matchLabels:
  10.       app: istio-ingressgateway
  11.   action: ALLOW
  12.   rules:
  13.   - from:
  14.     - source:
  15.         ipBlocks: ["1.2.3.4", "5.6.7.0/24", "$CLIENT_IP"]
  16. EOF

Find your original client IP address if you don’t know it and assign it to a variable:

  1. $ CLIENT_IP=$(kubectl get pods -n istio-system -o name -l istio=ingressgateway | sed 's|pod/||' | while read -r pod; do kubectl logs "$pod" -n istio-system | grep remoteIP; done | tail -1 | awk -F, '{print $4}' | awk -F: '{print $2}' | sed 's/ //') && echo "$CLIENT_IP"
  2. 192.168.10.15

Create the AuthorizationPolicy:

  1. $ kubectl apply -f - <<EOF
  2. apiVersion: security.istio.io/v1beta1
  3. kind: AuthorizationPolicy
  4. metadata:
  5.   name: ingress-policy
  6.   namespace: istio-system
  7. spec:
  8.   selector:
  9.     matchLabels:
  10.       app: istio-ingressgateway
  11.   action: ALLOW
  12.   rules:
  13.   - from:
  14.     - source:
  15.         remoteIpBlocks: ["1.2.3.4", "5.6.7.0/24", "$CLIENT_IP"]
  16. EOF

  • Verify that a request to the ingress gateway is allowed:

    1. $ curl "$INGRESS_HOST:$INGRESS_PORT"/headers -s -o /dev/null -w "%{http_code}\n"
    2. 200

  • Update the ingress-policy authorization policy to set the action key to DENY so that the IP addresses specified in the ipBlocks are not allowed to access the ingress gateway:

  1. $ kubectl apply -f - <<EOF
  2. apiVersion: security.istio.io/v1beta1
  3. kind: AuthorizationPolicy
  4. metadata:
  5.   name: ingress-policy
  6.   namespace: istio-system
  7. spec:
  8.   selector:
  9.     matchLabels:
  10.       app: istio-ingressgateway
  11.   action: DENY
  12.   rules:
  13.   - from:
  14.     - source:
  15.         ipBlocks: ["$CLIENT_IP"]
  16. EOF
  1. $ kubectl apply -f - <<EOF
  2. apiVersion: security.istio.io/v1beta1
  3. kind: AuthorizationPolicy
  4. metadata:
  5.   name: ingress-policy
  6.   namespace: istio-system
  7. spec:
  8.   selector:
  9.     matchLabels:
  10.       app: istio-ingressgateway
  11.   action: DENY
  12.   rules:
  13.   - from:
  14.     - source:
  15.         remoteIpBlocks: ["$CLIENT_IP"]
  16. EOF

  • Verify that a request to the ingress gateway is denied:

    1. $ curl "$INGRESS_HOST:$INGRESS_PORT"/headers -s -o /dev/null -w "%{http_code}\n"
    2. 403

  • You could use an online proxy service to access the ingress gateway using a different client IP to verify the request is allowed.

  • If you are not getting the responses you expect, view the ingress gateway logs which should show RBAC debugging information:

    1. $ kubectl get pods -n istio-system -o name -l istio=ingressgateway | sed 's|pod/||' | while read -r pod; do kubectl logs "$pod" -n istio-system; done

Clean up

  • Remove the namespace foo:

    1. $ kubectl delete namespace foo

  • Remove the authorization policy:

    1. $ kubectl delete authorizationpolicy ingress-policy -n istio-system

See also

Authorization Policy Trust Domain Migration

Shows how to migrate from one trust domain to another without changing authorization policy.

Authorization for HTTP traffic

Shows how to set up access control for HTTP traffic.

Authorization for TCP traffic

How to set up access control for TCP traffic.

Authorization policies with a deny action

Shows how to set up access control to deny traffic explicitly.

External authorization with custom action

Shows how to integrate and delegate access control to an external authorization system.

Security

Describes Istio’s authorization and authentication functionality.