Cluster Setup

​

Setting Up Your Kubernetes Cluster with EKS, IAM, and Trust Policies for Galileo Applications

This guide provides a comprehensive walkthrough for configuring and deploying an EKS (Elastic Kubernetes Service) environment to support Galileo applications. Galileo applications are designed to operate efficiently on managed Kubernetes services like EKS (Amazon Elastic Kubernetes Service) and GKE (Google Kubernetes Engine). This document, however, will specifically address the setup process within an EKS environment, including the integration of IAM (Identity and Access Management) roles and Trust Policies, alongside configuring the necessary Galileo DNS endpoints.

​

Prerequisites

Before you begin, ensure you have the following:

• An AWS account with administrative access

• kubectl installed on your local machine

• aws-cli version 2 installed and configured

• Basic knowledge of Kubernetes, AWS EKS, and IAM policies

Below lists the 4 steps to set deploy Galileo onto a an EKS environment.

​

Setting Up the EKS Cluster

1. Create an EKS Cluster: Use the AWS Management Console or AWS CLI to create an EKS cluster in your preferred region. For CLI, use the command aws eks create-cluster with the necessary parameters.

2. Configure kubectl: Once your cluster is active, configure kubectl to communicate with your EKS cluster by running aws eks update-kubeconfig --region <region> --name <cluster_name>.

​

Configuring IAM Roles and Trust Policies

1. Create IAM Roles for EKS: Navigate to the IAM console and create a new role. Select “EKS” as the trusted entity and attach policies that grant required permissions for managing the cluster.

2. Set Up Trust Policies: Edit the trust relationship of the IAM roles to allow the EKS service to assume these roles on behalf of your Kubernetes pods.

​

Integrating Galileo DNS Endpoints

1. Determine Galileo DNS Endpoints: Identify the four DNS endpoints required by Galileo applications to function correctly. These typically include endpoints for database connections, API gateways, telemetry services, and external integrations.

2. Configure DNS in Kubernetes: Utilize ConfigMaps or external-dns controllers in Kubernetes to route your applications to the identified Galileo DNS endpoints effectively.

​

Deploying Galileo Applications

1. Prepare Application Manifests: Ensure your Galileo application Kubernetes manifests are correctly set up with the necessary configurations, including environment variables pointing to the Galileo DNS endpoints.

2. Deploy Applications: Use kubectl apply to deploy your Galileo applications onto the EKS cluster. Monitor the deployment status to ensure they are running as expected.

​

Step 0: (Optional) Deploying via AWS CloudShell

To use eksctl via CloudShell in the AWS console, open a CloudShell session and do the following:

# Create directory
mkdir -p $HOME/.local/bin
cd $HOME/.local/bin

# eksctl
curl --silent --location "https://github.com/weaveworks/eksctl/releases/latest/download/eksctl_$(uname -s)_amd64.tar.gz" | tar xz -C /tmp
sudo mv /tmp/eksctl $HOME/.local/bin

The rest of the installation deployment can now be run from the CloudShell session. You can use vim to create/edit the required yaml and json files within the shell session.

​

Recommended Cluster Configuration

Galileo recommends the following Kubernetes deployment configuration:

Here’s an example EKS cluster configuration.

​

Step 1: Creating Roles and Policies for the Cluster

• Galileo IAM Policy: This policy is attached to the Galileo IAM Role. Add the following to a file called galileo-policy.json

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "eks:AccessKubernetesApi",
                "eks:DescribeCluster"
            ],
            "Resource": "arn:aws:eks:CLUSTER_REGION:ACCOUNT_ID:cluster/CLUSTER_NAME"
        }
    ]
}

• Galileo IAM Trust Policy: This trust policy enables an external Galileo user to assume your Galileo IAM Role to deploy changes to your cluster securely. Add the following to a file called galileo-trust-policy.json

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Principal": {
        "AWS": ["arn:aws:iam::273352303610:role/GalileoConnect"],
        "Service": "ec2.amazonaws.com"
      },
      "Action": "sts:AssumeRole"
    }
  ]
}

• Galileo IAM Role with Policy: Role should only include the Galileo IAM Policy mentioned in this table. Create a file called create-galileo-role-and-policies.sh, make it executable with chmod +x create-galileo-role-and-policies.sh and run it. Make sure to run in the same directory as the json files created in the above steps.

#!/bin/sh -ex

aws iam create-policy --policy-name Galileo --policy-document file://galileo-policy.json
aws iam create-role --role-name Galileo --assume-role-policy-document file://galileo-trust-policy.json
aws iam attach-role-policy --role-name Galileo --policy-arn $(aws iam list-policies | jq -r '.Policies[] | select (.PolicyName == "Galileo") | .Arn')

​

Step 2: Deploying the EKS Cluster

With the role and policies created, the cluster itself can be deployed in a single command using eksctl. Using the cluster template here, create a galileo-cluster.yaml file and edit the contents to replace CUSTOMER_NAME with your company name like galileo. Also check and update all availabilityZones as appropriate.

With the yaml file saved, run the following command to deploy the cluster:

eksctl create cluster -f galileo-cluster.yaml

​

Step 3: EKS IAM Identity Mapping

This ensures that only users who have access to this role can deploy changes to the cluster. Account owners can also make changes. This is easy to do with eksctl with the following command:

eksctl create iamidentitymapping
--cluster customer-cluster
--region your-region-id
--arn "arn:aws:iam::CUSTOMER-ACCOUNT-ID:role/Galileo"
--username galileo
--group system:masters

​

Step 4: Required Configuration Values

Customer specific cluster values (e.g. domain name, slack channel for notifications etc) will be placed in a base64 encoded string, stored as a secret in GitHub that Galileo’s deployment automation will read in and use when templating a cluster’s resource files.\

​

Step 5: Access to Deployment Logs

As a customer, you have full access to the deployment logs in Google Cloud Storage. You (customer) are able to view all configuration there. A customer email address must be provided to have access to this log.

​

Step 6: Customer DNS Configuration

Galileo has 4 main URLs (shown below). In order to make the URLs accessible across the company, you have to set the following DNS addresses in your DNS provider after the platform is deployed.

| Service | URL | | --- | --- | | API | api.galileo.company.[com|ai|io…] | | UI | console.galileo.company.[com|ai|io…] | | Grafana | grafana.galileo.company.[com|ai|io…] |

Each URL must be entered as a CNAME record into your DNS management system as the ELB address. You can find this address by listing the kubernetes ingresses that the platform has provisioned.

​

Creating a GPU-enabled Node Pool

For specialized tasks that require GPU processing, such as machine learning workloads, Galileo supports the configuration of GPU-enabled node pools. Here’s how you can set up and manage a node pool with GPU-enabled nodes using eksctl, a command line tool for creating and managing Kubernetes clusters on Amazon EKS.

1. Node Pool Creation: Use eksctl to create a node pool with an Amazon Machine Image (AMI) that supports GPUs. This example uses the g6.2xlarge instances and specifies a GPU-compatible AMI.

   eksctl create nodegroup --cluster your-cluster-name --name galileo-ml --node-type g6.2xlarge --nodes-min 1 --nodes-max 5 --node-ami ami-0656ebce2c7921ec0 --node-labels "galileo-node-type=galileo-ml" --region your-region-id
   

   In this command, replace your-cluster-name and your-region-id with your specific details. The --node-ami option is used to specify the exact AMI that supports CUDA and GPU workloads.

2. If the cluster has low usage and you want to save costs, you may also choose to use cheaper GPU like g4dn.2xlarge . Note that it only saves costs when the usage is too low to saturate one GPU, otherwise it would even cost more. And don’t choose this option if you use Protect that requires low real-time latency.

​

Using Managed RDS Postgres DB server

To use Managed RDS Postgres DB Server. You should create RDS Aurora directly in AWS console and Create K8s Secret and config map in kubernetes so that Galileo app can use it to connect to the DB server

​

Creating RDS Aurora cluster

1. Go to AWS Console —> RDS Service and create a RDS Subnet group.

• Select the VPC in which EKS cluster is running.

• Select AZs A and B and the respective private subnets

1. Next Create a RDS aurora Postgres Cluster. Config for the cluster are listed below. General fields like cluster name, username, password etc can we enter as per cloud best practice.

1. Create K8s Secret

• Kubernetes resources: Add the following to a file called galileo-rds-details.yaml. Update all marker $ text with appropriate values. Then run kubectl apply -f galileo-rds-details.yaml

---
apiVersion: v1
kind: Namespace
metadata:
  name: galileo
---
apiVersion: v1
kind: Secret
metadata:
  name: postgres
  namespace: galileo
type: Opaque
data:
  GALILEO_POSTGRES_USER: "${db_username}"
  GALILEO_POSTGRES_PASSWORD: "${db_username}"
  GALILEO_POSTGRES_REPLICA_PASSWORD: "${db_master_password}"
  GALILEO_DATABASE_URL_WRITE: "postgresql+psycopg2://${db_username}:${db_master_password}@${db_endpoint}/${database_name}"
  GALILEO_DATABASE_URL_READ: "postgresql+psycopg2://${db_username}:${db_master_password}@${db_endpoint}/${database_name}"
---
apiVersion: v1
kind: ConfigMap
metadata:
  name: grafana-datasources
  namespace: galileo
  labels:
    app: grafana
data:
  datasources.yaml: |
    apiVersion: 1
    datasources:
    - access: proxy
      isDefault: true
      name: prometheus
      type: prometheus
      url: "http://prometheus.galileo.svc.cluster.local:9090"
      version: 1
    - name: postgres
      type: postgres
      url: "${db_endpoint}"
      database: ${database_name}
      user: ${db_username}
      secureJsonData:
        password: ${db_master_password}
      jsonData:
        sslmode: "disable"
---