Setting up Multi Cluster Ingress


This page shows you how to route traffic across multiple Google Kubernetes Engine (GKE) clusters in different regions using Multi Cluster Ingress, with an example using two clusters.

For a detailed comparison between Multi Cluster Ingress (MCI), Multi-cluster Gateway (MCG), and load balancer with Standalone Network Endpoint Groups (LB and Standalone NEGs), see Choose your multi-cluster load balancing API for GKE.

To learn more about deploying Multi Cluster Ingress, see Deploying Ingress across clusters.

These steps require elevated permissions and should be performed by a GKE administrator.

Before you begin

Before you start, make sure you have performed the following tasks:

  • Enable the Google Kubernetes Engine API.
  • Enable Google Kubernetes Engine API
  • If you want to use the Google Cloud CLI for this task, install and then initialize the gcloud CLI. If you previously installed the gcloud CLI, get the latest version by running gcloud components update.

Requirements and limitations

Multi Cluster Ingress has the following requirements:

  • Google Cloud CLI version 290.0.0 and later.

If you use Standard mode clusters, ensure that you meet the following requirements. Autopilot clusters already meet these requirements.

Multi Cluster Ingress has the following limitations:

  • Only supported with an external Application Load Balancer.
  • Don't create Compute Engine load balancers in the same project with the prefix mci- that are not managed by Multi Cluster Ingress or they will be deleted. Google Cloud uses the prefix mci-[6 char hash] to manage the Compute Engine resources that Multi Cluster Ingress deploys.
  • Configuration of HTTPS requires a pre-allocated static IP address. HTTPS is not supported with ephemeral IP addresses.

Overview

In this exercise, you perform the following steps:

  1. Select the pricing you want to use.
  2. Deploy clusters.
  3. Configure cluster credentials.
  4. Register the clusters to a fleet.
  5. Specify a config cluster. This cluster can be a dedicated control plane, or it can run other workloads.

The following diagram shows what your environment will look like after you complete the exercise:

Cluster topology showing the relationships between regions, fleet, and project.

In the diagram, there are two GKE clusters named gke-us and gke-eu in the regions europe-west1 and us-central1. The clusters are registered to a fleet so that the Multi Cluster Ingress controller can recognize them. A fleet lets you logically group and normalize your GKE clusters, making administration of infrastructure easier and enabling the use of multi-cluster features such as Multi Cluster Ingress. You can learn more about the benefits of fleets and how to create them in the fleet management documentation.

Select pricing

If Multi Cluster Ingress is the only GKE Enterprise capability that you are using, then we recommend that you use standalone pricing. If your project is using other GKE Enterprise on Google Cloud components or capabilities, you should enable the entire GKE Enterprise platform. This lets you use all GKE Enterprise features for a single per-vCPU charge.

The APIs that you must enable depend on the Multi Cluster Ingress pricing that you use.

  • If the GKE Enterprise API (anthos.googleapis.com) is enabled, then your project is billed according to the number of cluster vCPUs and GKE Enterprise pricing.
  • If the GKE Enterprise API is disabled, then your project is billed according to the number of backend Multi Cluster Ingress pods in your project.

You can change the Multi Cluster Ingress billing model from standalone to GKE Enterprise, or from GKE Enterprise to standalone at any time without impacting Multi Cluster Ingress resources or traffic.

Standalone pricing

To enable standalone pricing, perform the following steps:

  1. Confirm that the GKE Enterprise API is disabled in your project:

    gcloud services list --project=PROJECT_ID | grep anthos.googleapis.com
    

    Replace PROJECT_ID with the project ID where your GKE clusters are running.

    If the output is an empty response, the GKE Enterprise API is disabled in your project and any Multi Cluster Ingress resources are billed using standalone pricing.

  2. Enable the required APIs in your project:

    gcloud services enable \
        multiclusteringress.googleapis.com \
        gkehub.googleapis.com \
        container.googleapis.com \
        multiclusterservicediscovery.googleapis.com \
        --project=PROJECT_ID
    

GKE Enterprise pricing

To enable GKE Enterprise pricing, enable the required APIs in your project:

gcloud services enable \
    anthos.googleapis.com \
    multiclusteringress.googleapis.com \
    gkehub.googleapis.com \
    container.googleapis.com \
    multiclusterservicediscovery.googleapis.com \
    --project=PROJECT_ID

After anthos.googleapis.com is enabled in your project, any clusters registered to Connect are billed according to GKE Enterprise pricing.

Deploy clusters

Create two GKE clusters named gke-us and gke-eu in the europe-west1and us-central1 regions.

Autopilot

  1. Create the gke-us cluster in the us-central1 region:

    gcloud container clusters create-auto gke-us \
        --region=us-central1 \
        --release-channel=stable \
        --project=PROJECT_ID
    

    Replace PROJECT_ID with your Google Cloud project ID.

  2. Create the gke-eu cluster in the europe-west1 region:

    gcloud container clusters create-auto gke-eu \
        --region=europe-west1 \
        --release-channel=stable \
        --project=PROJECT_ID
    

Standard

Create the two clusters with Workload Identity Federation for GKE enabled.

  1. Create the gke-us cluster in the us-central1 region:

    gcloud container clusters create gke-us \
        --region=us-central1 \
        --enable-ip-alias \
        --workload-pool=PROJECT_ID.svc.id.goog \
        --release-channel=stable \
        --project=PROJECT_ID
    

    Replace PROJECT_ID with your Google Cloud project ID.

  2. Create the gke-eu cluster in the europe-west1 region:

    gcloud container clusters create gke-eu \
        --region=europe-west1 \
        --enable-ip-alias \
        --workload-pool=PROJECT_ID.svc.id.goog \
        --release-channel=stable \
        --project=PROJECT_ID
    

Configure cluster credentials

Configure credentials for your clusters and rename the cluster contexts to make it easier to switch between clusters when deploying resources.

  1. Retrieve the credentials for your clusters:

    gcloud container clusters get-credentials gke-us \
        --region=us-central1 \
        --project=PROJECT_ID
    
    gcloud container clusters get-credentials gke-eu \
        --region=europe-west1 \
        --project=PROJECT_ID
    

    The credentials are stored locally so that you can use your kubectl client to access the cluster API servers. By default, an auto-generated name is created for the credentials.

  2. Rename the cluster contexts:

    kubectl config rename-context gke_PROJECT_ID_us-central1_gke-us gke-us
    kubectl config rename-context gke_PROJECT_ID_europe-west1_gke-eu gke-eu
    

Register clusters to a fleet

Register your clusters to your project's fleet as follows.

  1. Register your clusters:

    gcloud container fleet memberships register gke-us \
        --gke-cluster us-central1/gke-us \
        --enable-workload-identity \
        --project=PROJECT_ID
    
    gcloud container fleet memberships register gke-eu \
        --gke-cluster europe-west1/gke-eu \
        --enable-workload-identity \
        --project=PROJECT_ID
    
  2. Confirm that your clusters have successfully been registered to the fleet:

    gcloud container fleet memberships list --project=PROJECT_ID
    

    The output is similar to the following:

    NAME                                  EXTERNAL_ID
    gke-us                                0375c958-38af-11ea-abe9-42010a800191
    gke-eu                                d3278b78-38ad-11ea-a846-42010a840114
    

After you register your clusters, GKE deploys the gke-mcs-importer Pod to your cluster.

You can learn more about registering clusters in Register a GKE cluster to your fleet.

Specify a config cluster

The config cluster is a GKE cluster you choose to be the central point of control for Ingress across the member clusters. This cluster must already be registered to the fleet. For more information, see Config cluster design.

Enable Multi Cluster Ingress and select gke-us as the config cluster:

gcloud container fleet ingress enable \
    --config-membership=gke-us \
    --location=us-central1 \
    --project=PROJECT_ID

The config cluster takes up to 15 minutes to register. Successful output is similar to the following:

Waiting for Feature to be created...done.
Waiting for controller to start...done.

The unsuccessful output is similar to the following:

Waiting for controller to start...failed.
ERROR: (gcloud.container.fleet.ingress.enable) Controller did not start in 2 minutes. Please use the `describe` command to check Feature state for debugging information.

If a failure occurred in the previous step, then check the feature state:

gcloud container fleet ingress describe \
    --project=PROJECT_ID

The successful output is similar to the following:

createTime: '2021-02-04T14:10:25.102919191Z'
membershipStates:
 projects/PROJECT_ID/locations/global/memberships/CLUSTER_NAME:
 state:
   code: ERROR
   description: '...is not a VPC-native GKE Cluster.'
   updateTime: '2021-08-10T13:58:50.298191306Z'
 projects/PROJECT_ID/locations/global/memberships/CLUSTER_NAME:
 state:
   code: OK
   updateTime: '2021-08-10T13:58:08.499505813Z'

To learn more about troubleshooting errors with Multi Cluster Ingress, see Troubleshooting and operations.

Impact on live clusters

You can safely enable Multi Cluster Ingress using gcloud container fleet ingress enable on a live cluster, as it does not result in any downtime or impact to traffic on the cluster.

Shared VPC

You can deploy a MultiClusterIngress resource for clusters in a Shared VPC network, but all the participating backend GKE clusters must be in the same project. Having GKE clusters in different projects using the same Cloud Load Balancing VIP is not supported.

In non-Shared VPC networks, the Multi Cluster Ingress controller manages firewall rules to allow health checks to pass from the load balancer to container workloads.

In a Shared VPC network, a host project administrator must manually create the firewall rules for load balancer traffic on behalf of the Multi Cluster Ingress controller.

The following command shows the firewall rule that you must create if your clusters are on a Shared VPC network. The source ranges are the ranges that load balancer uses to send traffic to backends. This rule must exist for the operational lifetime of a MultiClusterIngress resource.

If your clusters are on a Shared VPC network, create the firewall rule:

gcloud compute firewall-rules create FIREWALL_RULE_NAME \
    --project=HOST_PROJECT \
    --network=SHARED_VPC \
    --direction=INGRESS \
    --allow=tcp:0-65535 \
    --source-ranges=130.211.0.0/22,35.191.0.0/16

Replace the following:

  • FIREWALL_RULE_NAME: the name of the new firewall rule that you choose.
  • HOST_PROJECT: the ID of the Shared VPC host project.
  • SHARED_VPC: the name of the Shared VPC network.

Known issues

This section describes known issues for the Multi Cluster Ingress

InvalidValueError for field config_membership

A known issue prevents the Google Cloud CLI from interacting with Multi Cluster Ingress. This issue was introduced in version 346.0.0 and was fixed in version 348.0.0. We don't recommend using the gcloud CLI versions 346.0.0 and 347.0.0 with Multi Cluster Ingress.

Invalid value for field 'resource'

Google Cloud Armor cannot communicate with Multi Cluster Ingress config clusters running on the following GKE versions:

  • 1.18.19-gke.1400 and later
  • 1.19.10-gke.700 and later
  • 1.20.6-gke.700 and later

When you configure a Google Cloud Armor security policy, the following message appears:

Invalid value for field 'resource': '{"securityPolicy": "global/securityPolicies/"}': The given policy does not exist

To avoid this issue, upgrade your config cluster to version 1.21 or later, or use the following command to update the BackendConfig CustomResourceDefinition:

kubectl patch crd backendconfigs.cloud.google.com --type='json' -p='[{"op": "replace", "path": "/spec/versions/1/schema/openAPIV3Schema/properties/spec/properties/securityPolicy", "value":{"properties": {"name": {"type": "string"}}, "required": ["name" ],"type": "object"}}]'

What's next