Create a GKE on Bare Metal hybrid cluster on Compute Engine VMs

Version 1.28 (latest)

• Version 1.28 (latest)
• Version 1.16
• Version 1.15
• Earlier versions (unsupported)

This page shows you how to set up a GKE on Bare Metal hybrid cluster in High Availability (HA) mode using Virtual Machines (VMs) running on Compute Engine.

Important: This deployment is for evaluation purposes only. Don't run GKE on Bare Metal on Compute Engine VMs in production. For production, follow the steps in our Quickstart to install GKE on Bare Metal on your hardware.

You can try out GKE on Bare Metal quickly and without having to prepare any hardware. Completing the steps on this page provides you with a working GKE on Bare Metal test environment that runs on Compute Engine.

To try GKE on Bare Metal on Compute Engine VMs, complete the following steps:

1. Create six VMs in Compute Engine
2. Create a vxlan network between all VMs with L2 connectivity
3. Install prerequisites for GKE on Bare Metal

4. Deploy a GKE on Bare Metal hybrid cluster

5. Verify your cluster

Before you begin

1. Sign in to your Google Cloud account. If you're new to Google Cloud, create an account to evaluate how our products perform in real-world scenarios. New customers also get $300 in free credits to run, test, and deploy workloads.

2. In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

   Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

   Go to project selector

3. Make sure that billing is enabled for your Google Cloud project.

4. In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

   Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

   Go to project selector

5. Make sure that billing is enabled for your Google Cloud project.

6. Make a note of the project ID because you need it to set an environment variable that is used in the scripts and commands on this page. If you selected an existing project, make sure that you are either a project owner or editor.
7. On your Linux workstation, make sure you have installed the latest Google Cloud CLI, the command line tool for interacting with Google Cloud. If you already have gcloud CLI installed, update its components by running the following command:

   gcloud components update

   Depending on how the gcloud CLI was installed, you might see the following message: "You cannot perform this action because the Google Cloud CLI component manager is disabled for this installation. You can run the following command to achieve the same result for this installation:" Follow the instructions to copy and paste the command to update the components.

The steps in this guide are taken from the installation script in the anthos-samples repository. The FAQ section has more information on how to customize this script to work with some popular variations.

Create six VMs in Compute Engine

Complete these steps to create the following VMs:

• One VM for the admin workstation. An admin workstation hosts command-line interface (CLI) tools and configuration files to provision clusters during installation, and CLI tools for interacting with provisioned clusters post-installation. The admin workstation will have access to all the other nodes in the cluster via SSH.
• Three VMs for the three control plane nodes needed to run the GKE on Bare Metal control plane.
• Two VMs for the two worker nodes needed to run workloads on the GKE on Bare Metal cluster.

1. Setup environment variables:

   export PROJECT_ID=PROJECT_ID
   export ZONE=ZONE
   export CLUSTER_NAME=CLUSTER_NAME
   export BMCTL_VERSION=1.28.400-gke.77
   

   For the ZONE, you can use us-central1-a or any of the other Compute Engine zones .

2. Run the following commands to log in with your Google account and set your project as the default:

   gcloud auth login
   gcloud config set project $PROJECT_ID
   gcloud config set compute/zone $ZONE
   

3. Create the baremetal-gcr service account:

   gcloud iam service-accounts create baremetal-gcr
   
   gcloud iam service-accounts keys create bm-gcr.json \
       --iam-account=baremetal-gcr@"${PROJECT_ID}".iam.gserviceaccount.com

4. Enable Google Cloud APIs and services:

   gcloud services enable \
       anthos.googleapis.com \
       anthosaudit.googleapis.com \
       anthosgke.googleapis.com \
       cloudresourcemanager.googleapis.com \
       connectgateway.googleapis.com \
       container.googleapis.com \
       gkeconnect.googleapis.com \
       gkehub.googleapis.com \
       serviceusage.googleapis.com \
       stackdriver.googleapis.com \
       monitoring.googleapis.com \
       logging.googleapis.com \
       opsconfigmonitoring.googleapis.com

5. Give the baremetal-gcr service account additional permissions to avoid needing multiple service accounts for different APIs and services:

   gcloud projects add-iam-policy-binding "$PROJECT_ID" \
     --member="serviceAccount:baremetal-gcr@$PROJECT_ID.iam.gserviceaccount.com" \
     --role="roles/gkehub.connect" \
     --no-user-output-enabled
   
   gcloud projects add-iam-policy-binding "$PROJECT_ID" \
     --member="serviceAccount:baremetal-gcr@$PROJECT_ID.iam.gserviceaccount.com" \
     --role="roles/gkehub.admin" \
     --no-user-output-enabled
   
   gcloud projects add-iam-policy-binding "$PROJECT_ID" \
     --member="serviceAccount:baremetal-gcr@$PROJECT_ID.iam.gserviceaccount.com" \
     --role="roles/logging.logWriter" \
     --no-user-output-enabled
   
   gcloud projects add-iam-policy-binding "$PROJECT_ID" \
     --member="serviceAccount:baremetal-gcr@$PROJECT_ID.iam.gserviceaccount.com" \
     --role="roles/monitoring.metricWriter" \
     --no-user-output-enabled
   
   gcloud projects add-iam-policy-binding "$PROJECT_ID" \
     --member="serviceAccount:baremetal-gcr@$PROJECT_ID.iam.gserviceaccount.com" \
     --role="roles/monitoring.dashboardEditor" \
     --no-user-output-enabled
   
   gcloud projects add-iam-policy-binding "$PROJECT_ID" \
     --member="serviceAccount:baremetal-gcr@$PROJECT_ID.iam.gserviceaccount.com" \
     --role="roles/stackdriver.resourceMetadata.writer" \
     --no-user-output-enabled
   
   gcloud projects add-iam-policy-binding "$PROJECT_ID" \
     --member="serviceAccount:baremetal-gcr@$PROJECT_ID.iam.gserviceaccount.com" \
     --role="roles/opsconfigmonitoring.resourceMetadata.writer" \
     --no-user-output-enabled

6. Create the variables and arrays needed for all the commands on this page:

   MACHINE_TYPE=n1-standard-8
   VM_PREFIX=abm
   VM_WS=$VM_PREFIX-ws
   VM_CP1=$VM_PREFIX-cp1
   VM_CP2=$VM_PREFIX-cp2
   VM_CP3=$VM_PREFIX-cp3
   VM_W1=$VM_PREFIX-w1
   VM_W2=$VM_PREFIX-w2
   declare -a VMs=("$VM_WS" "$VM_CP1" "$VM_CP2" "$VM_CP3" "$VM_W1" "$VM_W2")
   declare -a IPs=()

7. Use the following loop to create six VMs:

   Note: If you intend to deploy VMs using the Anthos VM Runtime, for better performance enable hardware virtualization by adding --enable-nested-virtualization to the following gcloud command.

   for vm in "${VMs[@]}"
   do
       gcloud compute instances create "$vm" \
         --image-family=ubuntu-2004-lts --image-project=ubuntu-os-cloud \
         --zone="${ZONE}" \
         --boot-disk-size 200G \
         --boot-disk-type pd-ssd \
         --can-ip-forward \
         --network default \
         --tags http-server,https-server \
         --min-cpu-platform "Intel Haswell" \
         --enable-nested-virtualization \
         --scopes cloud-platform \
         --machine-type "$MACHINE_TYPE" \
         --metadata "cluster_id=${CLUSTER_NAME},bmctl_version=${BMCTL_VERSION}"
       IP=$(gcloud compute instances describe "$vm" --zone "${ZONE}" \
            --format='get(networkInterfaces[0].networkIP)')
       IPs+=("$IP")
   done

   This command creates VM instances with the following names:

   • abm-ws: The VM for the admin workstation.
   • abm-cp1, abm-cp2, abm-cp3: The VMs for the control plane nodes.
   • abm-w1, abm-w2: The VMs for the nodes that run workloads.

8. Use the following loop to verify that SSH is ready on all VMs:

   for vm in "${VMs[@]}"
   do
       while ! gcloud compute ssh root@"$vm" --zone "${ZONE}" --command "printf 'SSH to $vm succeeded\n'"
       do
           printf "Trying to SSH into %s failed. Sleeping for 5 seconds. zzzZZzzZZ" "$vm"
           sleep  5
       done
   done

Create a vxlan network with L2 connectivity between VMs

Use the standard vxlan functionality of Linux to create a network that connects all the VMs with L2 connectivity.

The following command contains two loops that perform the following actions:

1. SSH into each VM.
2. Update and install needed packages.

3. Execute the required commands to configure the network with vxlan.

   i=2 # We start from 10.200.0.2/24
   for vm in "${VMs[@]}"
   do
       gcloud compute ssh root@"$vm" --zone "${ZONE}" << EOF
           apt-get -qq update > /dev/null
           apt-get -qq install -y jq > /dev/null
           set -x
           ip link add vxlan0 type vxlan id 42 dev ens4 dstport 0
           current_ip=\$(ip --json a show dev ens4 | jq '.[0].addr_info[0].local' -r)
           printf "VM IP address is: \$current_ip"
           for ip in ${IPs[@]}; do
               if [ "\$ip" != "\$current_ip" ]; then
                   bridge fdb append to 00:00:00:00:00:00 dst \$ip dev vxlan0
               fi
           done
           ip addr add 10.200.0.$i/24 dev vxlan0
           ip link set up dev vxlan0
   
   EOF
       i=$((i+1))
   done

You now have L2 connectivity within the 10.200.0.0/24 network. The VMs have the following IP addresses:

• Admin workstation VM: 10.200.0.2
• VMs running the control plane nodes:
  • 10.200.0.3
  • 10.200.0.4
  • 10.200.0.5
• VMs running the worker nodes:
  • 10.200.0.6
  • 10.200.0.7

Install prerequisites for GKE on Bare Metal

You need to install the following tools on the admin workstation before installing GKE on Bare Metal:

• bmctl
• kubectl
• Docker

To install the tools and prepare for GKE on Bare Metal installation:

1. Run the following commands to download the service account key to the admin workstation and install the required tools:

   gcloud compute ssh root@$VM_WS --zone "${ZONE}" << EOF
   set -x
   
   export PROJECT_ID=\$(gcloud config get-value project)
   BMCTL_VERSION=\$(curl http://metadata.google.internal/computeMetadata/v1/instance/attributes/bmctl_version -H "Metadata-Flavor: Google")
   export BMCTL_VERSION
   
   gcloud iam service-accounts keys create bm-gcr.json \
     --iam-account=baremetal-gcr@\${PROJECT_ID}.iam.gserviceaccount.com
   
   curl -LO "https://storage.googleapis.com/kubernetes-release/release/$(curl -s https://storage.googleapis.com/kubernetes-release/release/stable.txt)/bin/linux/amd64/kubectl"
   
   chmod +x kubectl
   mv kubectl /usr/local/sbin/
   mkdir baremetal && cd baremetal
   gsutil cp gs://anthos-baremetal-release/bmctl/$BMCTL_VERSION/linux-amd64/bmctl .
   chmod a+x bmctl
   mv bmctl /usr/local/sbin/
   
   cd ~
   printf "Installing docker"
   curl -fsSL https://get.docker.com -o get-docker.sh
   sh get-docker.sh
   EOF

2. Run the following commands to ensure that root@10.200.0.x works. The commands perform these tasks:

   1. Generate a new SSH key on the admin workstation.
   2. Add the public key to all the other VMs in the deployment.

   gcloud compute ssh root@$VM_WS --zone "${ZONE}" << EOF
   set -x
   ssh-keygen -t rsa -N "" -f /root/.ssh/id_rsa
   sed 's/ssh-rsa/root:ssh-rsa/' ~/.ssh/id_rsa.pub > ssh-metadata
   for vm in ${VMs[@]}
   do
       gcloud compute instances add-metadata \$vm --zone ${ZONE} --metadata-from-file ssh-keys=ssh-metadata
   done
   EOF

Deploy a GKE on Bare Metal hybrid cluster

The following code block contains all commands and configurations needed to complete the following tasks:

1. Create the configuration file for the needed hybrid cluster.
2. Run the preflight checks.
3. Deploy the cluster.

gcloud compute ssh root@$VM_WS --zone "${ZONE}" <<EOF
set -x
export PROJECT_ID=$(gcloud config get-value project)
CLUSTER_NAME=\$(curl http://metadata.google.internal/computeMetadata/v1/instance/attributes/cluster_id -H "Metadata-Flavor: Google")
BMCTL_VERSION=\$(curl http://metadata.google.internal/computeMetadata/v1/instance/attributes/bmctl_version -H "Metadata-Flavor: Google")
export CLUSTER_NAME
export BMCTL_VERSION
bmctl create config -c \$CLUSTER_NAME
cat > bmctl-workspace/\$CLUSTER_NAME/\$CLUSTER_NAME.yaml << EOB
---
gcrKeyPath: /root/bm-gcr.json
sshPrivateKeyPath: /root/.ssh/id_rsa
gkeConnectAgentServiceAccountKeyPath: /root/bm-gcr.json
gkeConnectRegisterServiceAccountKeyPath: /root/bm-gcr.json
cloudOperationsServiceAccountKeyPath: /root/bm-gcr.json
---
apiVersion: v1
kind: Namespace
metadata:
  name: cluster-\$CLUSTER_NAME
---
apiVersion: baremetal.cluster.gke.io/v1
kind: Cluster
metadata:
  name: \$CLUSTER_NAME
  namespace: cluster-\$CLUSTER_NAME
spec:
  type: hybrid
  anthosBareMetalVersion: \$BMCTL_VERSION
  gkeConnect:
    projectID: \$PROJECT_ID
  controlPlane:
    nodePoolSpec:
      clusterName: \$CLUSTER_NAME
      nodes:
      - address: 10.200.0.3
      - address: 10.200.0.4
      - address: 10.200.0.5
  clusterNetwork:
    pods:
      cidrBlocks:
      - 192.168.0.0/16
    services:
      cidrBlocks:
      - 172.26.232.0/24
  loadBalancer:
    mode: bundled
    ports:
      controlPlaneLBPort: 443
    vips:
      controlPlaneVIP: 10.200.0.49
      ingressVIP: 10.200.0.50
    addressPools:
    - name: pool1
      addresses:
      - 10.200.0.50-10.200.0.70
  clusterOperations:
    # might need to be this location
    location: us-central1
    projectID: \$PROJECT_ID
  storage:
    lvpNodeMounts:
      path: /mnt/localpv-disk
      storageClassName: node-disk
    lvpShare:
      numPVUnderSharedPath: 5
      path: /mnt/localpv-share
      storageClassName: local-shared
  nodeConfig:
    podDensity:
      maxPodsPerNode: 250
---
apiVersion: baremetal.cluster.gke.io/v1
kind: NodePool
metadata:
  name: node-pool-1
  namespace: cluster-\$CLUSTER_NAME
spec:
  clusterName: \$CLUSTER_NAME
  nodes:
  - address: 10.200.0.6
  - address: 10.200.0.7
EOB

bmctl create cluster -c \$CLUSTER_NAME
EOF

Verify your cluster

You can find your cluster's kubeconfig file on the admin workstation in the bmctl-workspace directory of the root account. To verify your deployment, complete the following steps.

1. SSH into the admin workstation as root:

   gcloud compute ssh root@abm-ws --zone ${ZONE}
   

   You can ignore any messages about updating the VM and complete this tutorial. If you plan to keep the VMs as a test environment, you might want to update the OS or upgrade to the next release as described in the Ubuntu documentation.

2. Set the KUBECONFIG environment variable with the path to the cluster's configuration file to run kubectl commands on the cluster.

   export clusterid=CLUSTER_NAME
   export KUBECONFIG=$HOME/bmctl-workspace/$clusterid/$clusterid-kubeconfig
   kubectl get nodes
   

3. Set the current context in an environment variable:

   export CONTEXT="$(kubectl config current-context)"
   

4. Run the following gcloud command. This command:

   • Grants your user account the Kubernetes clusterrole/cluster-admin role on the cluster.
   • Configures the cluster so that you can run kubectl commands on your local computer without having to SSH to the admin workstation.

   Replace GOOGLE_ACCOUNT_EMAIL with the email address that is associated with your Google Cloud account. For example: --users=[e-mail].

   gcloud container fleet memberships generate-gateway-rbac  \
       --membership=CLUSTER_NAME \
       --role=clusterrole/cluster-admin \
       --users=GOOGLE_ACCOUNT_EMAIL \
       --project=PROJECT_ID \
       --kubeconfig=$KUBECONFIG \
       --context=$CONTEXT\
       --apply
   

   The output of this command is similar to the following, which is truncated for readability:

   Validating input arguments.
   Specified Cluster Role is: clusterrole/cluster-admin
   Generated RBAC policy is:
   --------------------------------------------
   ...
   
   Applying the generate RBAC policy to cluster with kubeconfig: /root/bmctl-workspace/CLUSTER_NAME/CLUSTER_NAME-kubeconfig, context: CLUSTER_NAME-admin@CLUSTER_NAME
   Writing RBAC policy for user: GOOGLE_ACCOUNT_EMAIL to cluster.
   Successfully applied the RBAC policy to cluster.
   

5. When you are finished exploring, enter exit to log out of the admin workstation.

6. Get the kubeconfig entry that can access the cluster through the Connect gateway.

   gcloud container fleet memberships get-credentials CLUSTER_NAME
   

   The output is similar to the following:

   Starting to build Gateway kubeconfig...
   Current project_id: PROJECT_ID
   A new kubeconfig entry "connectgateway_PROJECT_ID_global_CLUSTER_NAME" has been generated and set as the current context.
   

7. You can now run kubectl commands through the Connect gateway:

   kubectl get nodes
   kubectl get namespaces
   

Log in to your cluster from Google Cloud console

To observe your workloads on GKE on Bare Metal in the Google Cloud console, you need to log in to the cluster. Before you log in to the console for the first time, you need to configure an authentication method. The easiest authentication method to configure is Google identity. This authentication method lets you log in using the email address associated with your Google Cloud account.

The gcloud container fleet memberships generate-gateway-rbac command that you ran in the previous section configures the cluster so that you can log in with your Google identity.

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

   Go to GKE clusters

2. Click Actions next to the registered cluster, then click Login.

3. Select Use your Google identity to log in.

4. Click Login.

Clean up

1. Connect to the admin workstation to reset the cluster VMs to their state prior to installation and unregister the cluster from your Google Cloud project:

   gcloud compute ssh root@abm-ws --zone ${ZONE} << EOF
   set -x
   export clusterid=CLUSTER_NAME
   bmctl reset -c \$clusterid
   EOF
   

2. List all VMs that have abm in their name:

   gcloud compute instances list | grep 'abm'
   

3. Verify that you're fine with deleting all VMs that contain abm in the name.

   After you've verified, you can delete abm VMs by running the following command:

   gcloud compute instances list --format="value(name)" | grep 'abm'  | xargs gcloud \
       --quiet compute instances delete