Skip to content

Latest commit

 

History

History
 
 

anthos-service-mesh-multicluster

README.md

Contents

Multi-Cluster ASM on Private Clusters

Documentation

Here are several reference documents if you encounter an issue when following the instructions below:

Description

In Adding clusters to an Athos Service Mesh, it shows how to federate service meshes of two Anthos public clusters. However, it misses a key instruction to open the firewall for the service port to the remote cluster. So, your final test of HelloWorld might not work.

This sample builds on the topic of Google's Anthos Service Mesh official installation documents, and adds instructions on how to federate two private clusters, which is more likely in real world environments.

As illustrated in the diagram below, we will create a VPC with three subnets. Two subnets are for private clusters, and one for GCE servers. So, we illustrate using a bastion server to access private clusters as in a real environment.

NetworkImage

The clusters are not accessible from an external network. Users can only log into the bastion server via an IAP tunnel to gain access to this VPC. A firewall rule is built to allow IAP tunneling into the GCE subnet (Subnet C) only. For the bastion server in Subnet C to access Kubernetes APIs of both private clusters, Subnet C's CIDR range is added to the "GKE Control Plane Authorized Network" of both clusters. This is illustrated as blue lines and yellow underscore lines in the diagram above.

Also, in order for both clusters to access the service mesh (Istiod) and service deployed on the other cluster, we need to do the following:

  • The pod CIDR range of one cluster must be added to the "GKE Control Plane Authorized Network" of the other cluster. This enables one cluster to ping istiod on the other cluster.
  • The firewall needs to be open for one cluster's pod CIDR to access the service port on the other cluster. In this sample, it is port 5000 used by the HelloWord testing application. Because the invocation of service is bidirectional in HelloWorld testing application, we will add firewall rules for each direction.

The infrastruction used in this sample is coded in Terraform scripts. The ASM installation steps are coded in a Shell script.

Prerequisites

As mentioned in Add GKE clusters to Anthos Service Mesh, there are several prerequisites.

This guide assumes that you have:

Also, the multi-cluster configuration has these requirements for the clusters in it:

  • All clusters must be on the same network.

    NOTE: ASM 1.7 does not support multiple networks, even peered ones.

  • If you join clusters that are not in the same project, they must be installed using the asm-gcp-multiproject profile and the clusters must be in a shared VPC configuration together on the same network. In addition, we recommend that you have one project to host the shared VPC, and two service projects for creating clusters. For more information, see Setting up clusters with Shared VPC.

In this sample, we create two private clusters in different subnets of the same VPC in the same project, and enable clusters to communicate to each other's API server.

How to set up and run this sample

Build Infrastructure

  1. Create a GCP project.

  2. Create a VPC in GCP project.

  3. Create a subnet in the VPC.

  4. Create a VM in the subnet. This will be the bastion server to simulate an intranet access to GKE clusters.

    • This step is now done by Terraform, in file infrastructure/bastion.tf
    • The Bastion host is used for interraction with the GKE clusters
    • For this demo, we ran Terraform from a local machine, not from the Bastion host
    • Note: you will have to manually create a Google Cloud firewall rule, to allow connection to the bastion server via SSH (port 22). We did not automate this for security reasons.

  5. Set up Git on your local machine, then clone this Github sample. Also clone this Github sample onto the bastion server

  6. Set up Terraform on your local machine, so you will be able to build infrastructure.

  7. On your local machine, update the corresponding parameters for your project.

    • In vars.sh, check to see whether you need to update CLUSTER1_LOCATION,CLUSTER1_CLUSTER_NAME, CLUSTER1_CLUSTER_CTX, CLUSTER2_LOCATION, CLUSTER2_CLUSTER_NAME, CLUSTER2_CLUSTER_CTX.

    • In infrastructure/terraform.example.tfvars, rename this file to terraform.tfvars and update "project_id" and "billing_account".

    • In infrastructure/shared.tf, check whether you need to update "project_prefix" and "region".

    • [OPTIONAL] In the locals section of infrastructure/shared.tf, update CIDR ranges for bastion_cidr and existing_vpc if you need to.

    • Source vars.sh to set up basic environment variables.

      source vars.sh

  8. If you want to run Terraform in your own workspace, create a backend.tf file from infrastructure/backend.tf_tmpl, and update your Terraform workspace information in this file.

  9. Under "infrastructure" directory, run

    • terraform init

      terraform init

    • terraform plan

      terraform plan -out output.tftxt

      NOTE: You may get an error that the Compute Engine API has not been used before in the project. In this case please manually enable the Compute Engine API

      Error: Error when reading or editing GCE default service account: googleapi: Error 403: Compute Engine API has not been used in project XXXXXXXXXXX before or it is disabled. Enable it by visiting https://console.developers.google.com/apis/api/compute.googleapis.com/overview then retry. If you enabled this API recently, wait a few minutes for the action to propagate to our systems and retry., accessNotConfigured

    • terraform apply

      terraform apply output.tftxt

    • If Terraform completes without error, you should have VPC, NAT, a bastion server, two private clusters and firewall rules. Please check all artifacts in GCP Console.

  10. SSH onto the bastion server.

  11. Make sure you have the following tools installed:

    • The Cloud SDK (the gcloud command-line tool)
    • The standard command-line tools: awk, curl, grep, sed, sha256sum, and tr
    • git
    • kpt
    • kubectl
    • jq

Install ASM

  1. On bastion server, go to this source code directory, then source vars.sh

    NOTE: Make sure you manually create a Google Cloud firewall rule to allow SSH connections to your bastion server over port 22

    cd asm-private-multiclusters-intranet
source vars.sh

  2. Source scripts/main.sh

    source scripts/main.sh

  3. Run install_asm_mesh

    install_asm_mesh

    or, you can run the commands in install_asm_mesh step by step manually

    # Navigate to your working directory. Binaries will be downloaded to this directory.
cd ${WORK_DIR}

# Set up K8s config and context
set_up_credential ${CLUSTER1_CLUSTER_NAME} ${CLUSTER1_LOCATION} ${CLUSTER1_CLUSTER_CTX} ${TF_VAR_project_id}
set_up_credential ${CLUSTER2_CLUSTER_NAME} ${CLUSTER2_LOCATION} ${CLUSTER2_CLUSTER_CTX} ${TF_VAR_project_id}

# Download ASM Installer
download_asm_installer ${ASM_MAJOR_VER} ${ASM_MINOR_VER}

#Install ASM
install_asm ${CLUSTER1_CLUSTER_NAME} ${CLUSTER1_LOCATION} ${TF_VAR_project_id}
install_asm ${CLUSTER2_CLUSTER_NAME} ${CLUSTER2_LOCATION} ${TF_VAR_project_id}

# Register clusters
grant_role_to_connect_agent ${TF_VAR_project_id}
register_cluster ${CLUSTER1_CLUSTER_CTX} ${CLUSTER1_LOCATION}
register_cluster ${CLUSTER2_CLUSTER_CTX} ${CLUSTER2_LOCATION}

# Add clusters to mesh
cross_cluster_service_secret ${CLUSTER1_CLUSTER_NAME} ${CLUSTER1_CLUSTER_CTX} ${CLUSTER2_CLUSTER_CTX}
cross_cluster_service_secret ${CLUSTER2_CLUSTER_NAME} ${CLUSTER2_CLUSTER_CTX} ${CLUSTER1_CLUSTER_CTX}

Deploy test helloworld application

Run install_test_app

install_test_app

Prepare for verification

export CTX1=$CLUSTER1_CLUSTER_CTX
export CTX2=$CLUSTER2_CLUSTER_CTX

Follow the instruction in "Verify cross-cluster load balancing" section of Add clusters to an Anthos Service Mesh to verify.

Please Note: You don't need to install Helloworld application, it has been installed for you already.

Internal Load Balancer

Anthos ASM deploys ingress gateway using exernal load balancer by default. If we need to change the ingress gateway to be internal load balancer, we can use --option or --custom-overlay parameter along with out load balancer yaml (./istio-profiles/internal-load-balancer.yaml).

Please note that we need to specify out "targetPort" for https and http2 ports for current ASM version.

Twistlock PoC

  • Pod traffic security scanning, using ASM, Docker and Google Artifact Registry (GAR)
  • Please see the twistlock folder readme

Cloud SQL for PostgreSQL PoC

  • Connecting GKE clusters and ASM to a database that is external database to the Kubernetes clusters
  • Please see the postgres folder readme