Set up for GKE clusters

This document shows you how to set up Binary Authorization for GKE clusters. It then shows you how to configure a sample Binary Authorization policy.

Before you begin

  1. You must have GKE clusters registered with Connect. Binary Authorization supports the following environments.

    GKE on Bare Metal

    GKE on Bare Metal 1.14 or later.

    GKE on VMware

    GKE on VMware 1.4 or later.

  2. The Binary Authorization service uses a public IP address, accessible through a regular internet connection. Configure your firewall rules for HTTPS to let the user cluster access the endpoint binaryauthorization.googleapis.com.

    GKE on Bare Metal

    Configure GKE on Bare Metal firewall rules.

    GKE on VMware

    Configure GKE on VMware firewall rules.

  3. If you want to use centralized Cloud Audit Logs to view audit log entries, including those from Binary Authorization for GKE clusters, you must configure Cloud Audit Logs in your cluster configuration.

    GKE on Bare Metal

    Configure Cloud Audit Logs in GKE on Bare Metal.

    GKE on VMware

    Configure Cloud Audit Logs in GKE on VMware.

  4. You must enable the Binary Authorization API as follows:

    1. Go to Google Cloud console.

      Enable APIs

    2. In the project drop-down list, select your Connect project. You can find this Google Cloud project in the gkeConnect section of your user cluster configuration file. This is the Google Cloud project that connects your user cluster to Google Cloud.

Set up Binary Authorization

In this section, you set up Binary Authorization for GKE clusters in your cluster.

Specify installation environment variables

To specify the environment variables, do the following:

Using Workload Identity

  1. Specify your Connect project:

    export PROJECT_ID=PROJECT_ID
    
  2. Specify the Fleet membership ID of your cluster:

    export MEMBERSHIP_ID=MEMBERSHIP_ID
    

Using service account key

  1. Specify your Connect project:

    export PROJECT_ID=PROJECT_ID
    

    Replace PROJECT_ID with the Google Cloud project in the gkeConnect section of your user cluster configuration file.

  2. Specify the path of the user cluster's kubeconfig file:

    export KUBECONFIG=PATH
    

    Replace PATH with the path of your user cluster kubeconfig file.

  3. Choose a name for the Binary Authorization API access service account:

    export SA_NAME=SERVICE_ACCOUNT_NAME
    

    Replace SERVICE_ACCOUNT_NAME with the service account name of your choice. The Binary Authorization Module uses this service account to access the Binary Authorization API.

  4. Specify the path to the service account key file you download later in this guide:

    export SA_JSON_PATH=SA_KEY_FILE_PATH
    

    Replace SA_KEY_FILE_PATH with the path of the JSON key file for the service account.

Install the Binary Authorization module in your user cluster

To install the Binary Authorization module, do the following:

Using Workload Identity

Fleet Workload Identity lets the workloads in your cluster authenticate to Google without requiring you to download, manually rotate, and generally manage Google Cloud service account keys. You can find out more about how fleet Workload Identity works and the advantages of using it in Use fleet Workload Identity.

  1. Grant the binaryauthorization.policyEvaluator role to the Kubernetes service account on your Connect project:

    gcloud projects add-iam-policy-binding ${PROJECT_ID} \
        --member="serviceAccount:${PROJECT_ID}.svc.id.goog[binauthz-system/binauthz-admin]" \
        --role="roles/binaryauthorization.policyEvaluator"
    
  2. Create a working directory:

    1. Make a directory called binauthz.

    2. Change to the directory.

  3. Download the manifest-wi-0.2.6.yaml.tmpl file, which you use to install the Binary Authorization Module in your GKE clusters user cluster:

    GKE on Bare Metal

    gsutil cp gs://anthos-baremetal-release/binauthz/manifest-wi-0.2.6.yaml.tmpl .
    

    GKE on VMware

    gsutil cp gs://gke-on-prem-release/binauthz/manifest-wi-0.2.6.yaml.tmpl .
    
  4. Substitute the environment variables in the template:

    envsubst < manifest-wi-0.2.6.yaml.tmpl > manifest-0.2.6.yaml
    
  5. Install the Binary Authorization Module in your user cluster:

    kubectl apply -f manifest-0.2.6.yaml
    
  6. Verify the Deployment was created:

    kubectl get pod --namespace binauthz-system
    

    You see the Pod binauthz-module-deployment-* listed with Status of Running and 1/1 Pods ready, similar to this output:

    NAME                                          READY   STATUS    RESTARTS   AGE
    binauthz-module-deployment-5fddf9594f-qjprz   1/1     Running   0          11s
    

Using service account key

  1. Set the default project for the Google Cloud CLI:

    gcloud config set project ${PROJECT_ID}
    
  2. Create a Binary Authorization API access service account:

    gcloud iam service-accounts create ${SA_NAME}
    
  3. Grant the binaryauthorization.policyEvaluator role to the Binary Authorization API access service account on your Connect project:

    gcloud projects add-iam-policy-binding ${PROJECT_ID}\
        --member="serviceAccount:${SA_NAME}@${PROJECT_ID}.iam.gserviceaccount.com" \
        --role="roles/binaryauthorization.policyEvaluator"
    
  4. Create a working directory:

    1. Make a directory called binauthz.

    2. Change to the directory.

  5. Download the manifest-0.2.6.yaml file, which you use to install the Binary Authorization Module in your GKE clusters user cluster:

    anthos_clusters_on_bare_metal

    gsutil cp gs://anthos-baremetal-release/binauthz/manifest-0.2.6.yaml .
    

    anthos_clusters_on_vmware

    gsutil cp gs://gke-on-prem-release/binauthz/manifest-0.2.6.yaml .
    
  6. Create a YAML file for thebinauthz-system namespace.

    Copy the following to a file called namespace.yaml:

    apiVersion: v1
    kind: Namespace
    metadata:
      labels:
        control-plane: binauthz-controller
      name: binauthz-system
    
  7. Create the namespace in your user cluster:

    kubectl apply -f namespace.yaml
    

    You see output similar to the following:

    namespace/binauthz-system created
    
  8. Download a JSON key file for your service account:

    gcloud iam service-accounts keys create ${SA_JSON_PATH} --iam-account ${SA_NAME}@${PROJECT_ID}.iam.gserviceaccount.com
    
  9. Save the service account key as a Kubernetes Secret in your user cluster:

    kubectl --namespace binauthz-system create secret generic binauthz-sa --from-file=key.json=${SA_JSON_PATH}
    
  10. Install the Binary Authorization Module in your user cluster:

    kubectl apply -f manifest-0.2.6.yaml
    
  11. Verify the Deployment was created:

    kubectl get pod --namespace binauthz-system
    

    You see the Pod binauthz-module-deployment-* listed with Status of Running and 1/1 Pods ready, similar to this output:

    NAME                                          READY   STATUS    RESTARTS   AGE
    binauthz-module-deployment-5fddf9594f-qjprz   1/1     Running   0          11s
    

Set up and use Binary Authorization policies

This section shows you how to set up and use Binary Authorization policies for GKE clusters.

In each example, you configure the policy and then test it by attempting to deploy a container image in your GKE cluster.

Allow all

This section demonstrates a success case. You configure the Binary Authorization policy so that a container image satisfies the policy and is deployed.

In Google Cloud, do the following:

Console

  1. In the Google Cloud console, go to the Binary Authorization page.

    Go to Binary Authorization

  2. Make sure to select your Connect project ID.

  3. Click Edit Policy.

  4. Under Project default rule, select Allow all images.

  5. Click Save Policy.

gcloud

  1. Set the PROJECT_ID for your Connect project. You can find this project ID in the gkeConnect field in your user cluster configuration file.

    export PROJECT_ID=PROJECT_ID
    

    Set the default Google Cloud project.

    gcloud config set project ${PROJECT_ID}
    
  2. Export the policy YAML file to your local system:

    gcloud container binauthz policy export  > policy.yaml
    

    Your YAML file looks like the following:

    admissionWhitelistPatterns:
    - namePattern: gcr.io/google_containers/*
    - namePattern: gcr.io/google-containers/*
    - namePattern: k8s.gcr.io/**
    - namePattern: gke.gcr.io/**
    - namePattern: gcr.io/stackdriver-agents/*
    globalPolicyEvaluationMode: ENABLE
    defaultAdmissionRule:
      evaluationMode: ALWAYS_ALLOW
      enforcementMode: ENFORCED_BLOCK_AND_AUDIT_LOG
    name: projects/<var>PROJECT_ID</var>/policy
    
  3. Edit policy.yaml.

  4. Set evaluationMode to ALWAYS_ALLOW.

  5. If you have a requireAttestationsBy block in the file, delete this block.

  6. Save the file.

  7. Import policy.yaml as follows:

    gcloud container binauthz policy import policy.yaml
    

On your GKE clusters admin workstation, do the following:

  1. Create a manifest file for a Pod.

    Save the following to a file called pod.yaml:

    apiVersion: v1
    kind: Pod
    metadata:
      name: test-pod
    spec:
      containers:
      - name: test-container
        image: gcr.io/google-samples/hello-app@sha256:c62ead5b8c15c231f9e786250b07909daf6c266d0fcddd93fea882eb722c3be4
    
  2. Create the Pod:

    kubectl apply -f pod.yaml
    

    You see that the Pod was successfully deployed.

  3. Delete the Pod:

    kubectl delete -f pod.yaml
    

Disallow all

This section demonstrates a failure case. In this section, you configure the default policy to disallow your container image from being deployed.

In Google Cloud do the following:

Console

  1. In the Google Cloud console, go to the Binary Authorization page.

    Go to Binary Authorization

  2. Make sure your Connect project is selected.

  3. Click Edit Policy.

  4. Under Project default rule, select Disallow all images.

  5. Click Save policy.

gcloud

  1. Set the PROJECT_ID to your Connect project ID.

    export PROJECT_ID=PROJECT_ID
    
  2. Set the default Google Cloud project.

    gcloud config set project ${PROJECT_ID}
    
  3. Export the policy YAML file:

    gcloud container binauthz policy export  > policy.yaml
    
  4. Edit policy.yaml.

  5. Set evaluationMode to ALWAYS_DENY.

  6. If you have a requireAttestationsBy block in the file, delete this block.

  7. Save the file.

  8. Import policy.yaml as follows:

    gcloud container binauthz policy import policy.yaml
    

On your GKE clusters admin workstation, do the following:

  1. Create a manifest file for a Pod.

    Save the following to a file called pod.yaml:

    apiVersion: v1
    kind: Pod
    metadata:
      name: test-pod
    spec:
      containers:
      - name: test-container
        image: gcr.io/google-samples/hello-app@sha256:c62ead5b8c15c231f9e786250b07909daf6c266d0fcddd93fea882eb722c3be4
    
  2. Create the Pod:

    kubectl apply -f pod.yaml
    

    You see that the Pod was was blocked from being deployed. The output looks like the following:

    Error from server (VIOLATES_POLICY): error when creating "pod.yaml": admission webhook "binaryauthorization.googleapis.com" denied the request: Denied by default admission rule. Overridden by evaluation mode
    

Get the user cluster resource ID

This section shows you how to compose the cluster resource ID for your user cluster. In your Binary Authorization policy, you can create cluster-specific rules. You associate these rules with a cluster-specific resource ID, which is based on your cluster ID.

You get the resource ID as follows:

Console

  1. In the Google Cloud console, go to the GKE Enterprise Clusters page.

    Go to Clusters

  2. Select the Connect project ID for GKE clusters. You can find this project ID in the gkeConnect section of your user cluster configuration file.

  3. Under Anthos Managed Clusters, find your cluster ID under the Name column.

  4. To create the resource ID, add the prefix global. to the cluster ID so that the resource ID has the following format: global.CLUSTER_ID.

gcloud

  1. Use SSH to connect to your GKE clusters admin workstation.

  2. On the admin workstation, run the following command:

    kubectl get membership -o yaml
    
  3. Get the cluster ID from spec.owner.id field of the output. Example output follows:

    apiVersion: v1
    items:
    - apiVersion: hub.gke.io/v1
      kind: Membership
      ...
      spec:
        owner:
          id: //gkehub.googleapis.com/projects/PROJECT_NUMBER/locations/global/memberships/my-cluster-id
    

    In the example output, the cluster ID is my-cluster-id.

  4. To create the resource ID, add the prefix global. to the cluster ID. In the example, the resource ID is global.my-cluster-id.

You use this resource ID when you define cluster-specific rules. Learn how to set cluster-specific rules by using the Google Cloud console or the gcloud CLI.

Update the failure policy

The Binary Authorization Module webhook can be configured to fail open or fail close.

Fail close

To update the failure policy to fail close, do the following:

  1. Edit the manifest-0.2.6.yaml file and set failurePolicy to Fail

  2. Re-enable the webhook:

    kubectl apply -f manifest-0.2.6.yaml
    

    You see output similar to the following:

    serviceaccount/binauthz-admin unchanged
    role.rbac.authorization.k8s.io/binauthz-role configured
    clusterrole.rbac.authorization.k8s.io/binauthz-role configured
    rolebinding.rbac.authorization.k8s.io/binauthz-rolebinding unchanged
    clusterrolebinding.rbac.authorization.k8s.io/binauthz-rolebinding unchanged
    secret/binauthz-tls unchanged
    service/binauthz unchanged
    deployment.apps/binauthz-module-deployment unchanged
    validatingwebhookconfiguration.admissionregistration.k8s.io/binauthz-validating-webhook-configuration configured
    

Fail open

To update the failure policy to fail open, do the following:

  1. Edit the manifest-0.2.6.yaml file and set failurePolicy to Ignore

  2. Re-enable the webhook:

    kubectl apply -f manifest-0.2.6.yaml
    

    You see output similar to the following:

    serviceaccount/binauthz-admin unchanged
    role.rbac.authorization.k8s.io/binauthz-role configured
    clusterrole.rbac.authorization.k8s.io/binauthz-role configured
    rolebinding.rbac.authorization.k8s.io/binauthz-rolebinding unchanged
    clusterrolebinding.rbac.authorization.k8s.io/binauthz-rolebinding unchanged
    secret/binauthz-tls unchanged
    service/binauthz unchanged
    deployment.apps/binauthz-module-deployment unchanged
    validatingwebhookconfiguration.admissionregistration.k8s.io/binauthz-validating-webhook-configuration configured
    

To learn more, see webhook failure policy.

Clean up

  1. The following code sample shows how to disable the webhook:

    kubectl delete ValidatingWebhookConfiguration/binauthz-validating-webhook-configuration
    
  2. The following code sample shows how to re-enable the webhook:

    kubectl apply -f manifest-0.2.6.yaml
    
  3. The following code sample shows how to delete all of the resources related to Binary Authorization:

    kubectl delete -f manifest-0.2.6.yaml
    kubectl delete namespace binauthz-system
    

What's next