Deploying your Kubernetes cluster v1.3.4

Overview

Role focus: Kubernetes engineer / CSP Administrator

Prerequisites:

• Phase 1: Planning your architecture (Completed)
• Phase 2: Gathering your system requirements (Completed)

Outcomes

• A running, reachable Kubernetes cluster that meets all compute, storage, and networking requirements for Hybrid Manager (HM).

Next phase: Phase 4: Preparing your environment

Start the provisioning process

With your system requirements defined, you are ready to provision the infrastructure. This phase covers deploying the base Kubernetes cluster that hosts the Hybrid Manager (HM) platform.

HM is largely platform-agnostic, supporting major cloud providers and on-premises distributions. For platform-specific variabilities,

Set up your management workstation

Provision and configure your management workstation (Bastion or other) as planned in Gathering your System requirements the previous phase.

Install core tooling

You must install the following CLI tools to orchestrate the deployment.

1. Install Kubernetes Tools (kubectl & helm): Follow the official guides to install kubectl and Helm.

2. Install edbctl (EDB Hybrid Manager CLI): Run the installation script to download the binary:

   → WrapCopy

   curl -sfL [https://get.enterprisedb.com/edbctl/install.sh](https://get.enterprisedb.com/edbctl/install.sh) | sh -

3. Install utilities:

   Ensure yq (v4+), curl, and openssl are installed via your package manager.

   → WrapCopy

   # Example for Ubuntu/Debian
   sudo apt-get update && sudo apt-get install -y curl openssl
   
   # Example for MacOS
   brew install yq curl openssl

4. Install platform CLIs (Environment dependent):

   If you are deploying to a public cloud, install the relevant CLI for authentication.

   • AWS: Install AWS CLI
   • Google Cloud: Install gcloud SDK
   • OpenShift: Install oc CLI

Verify connectivity

Before proceeding to the configuration phase, validate that your workstation can reach the necessary endpoints.

1. Verify Internet/Registry access

   → WrapCopy

   curl -I [https://docker.enterprisedb.com/v2/](https://docker.enterprisedb.com/v2/)

2. Verify Cloud Identity (if using AWS/GCP)

   for AWS:

   → WrapCopy

   aws sts get-caller-identity

   or for GCP:

   → WrapCopy

   gcloud auth list

3. Verify local tools

   → WrapCopy

   edbctl version
   kubectl version --client

Deploy your Kubernetes cluster

With your management workstation configured, you are ready to deploy your cluster according to the relevant specifications defined in the previous Phase.

Select your deployment path

Select the guide below that matches the Kubernetes flavor you selected in Phase 1 or 2 and proceed by deploying according to your system requirements defined in Gathering your system requirements.

Public cloud providers

Private cloud and on-premises

Set your node abstractions

Control Plane

Required labels & taints:

• Label: edbaiplatform.io/control-plane: "true"
• Taint:
  • Key: edbaiplatform.io/control-plane
  • Value: "true"
  • Effect: NoSchedule

spec:
    replicas: 3
    template:
    spec:
        metadata:
        labels:
            edbaiplatform.io/control-plane: "true"
        taints:
        - key: edbaiplatform.io/control-plane
            value: "true"
            effect: NoSchedule

Data plane

Required labels & taints:

• Label: edbaiplatform.io/postgres: "true"
• Taint:
  • Key: edbaiplatform.io/postgres
  • Value: "true"
  • Effect: NoSchedule

# Example Node Pool Specification
spec:
    replicas: 3 # Minimum recommended for High Availability
    template:
      metadata:
        labels:
          edbaiplatform.io/postgres: "true"
      spec:
        taints:
        - key: edbaiplatform.io/postgres
          value: "true"
          effect: NoSchedule

AI model nodes

• Sizing: Current recommendations require Nvidia B200 GPUs (or equivalent supported hardware).

Required labels & taints:

• Label: nvidia.com/gpu: "true"
• Taint:
  • Key: nvidia.com/gpu
  • Value: "true"
  • Effect: NoSchedule

Validate cluster specifications

Now that you have deployed your cluster, before proceeding:

1. Confirm that your management workstation can reach all relavent endpoints for configuration in the next phase: Preparing your environment.

2. Confirm that your deployed cluster matches the relevant specifications you defined in Phase 2: Gathering your system requirements:

Compute resources

Source: Gathering your system requirements>Compute (Node requirements)

• Node count: Ensure your node pools match what was decided in the previous phases.
• Architecture: Verify that all nodes use the AMD64 (x86_64) architecture. ARM-based nodes are not currently supported for the Control Plane.

Validation of node abstractions

After provisioning your infrastructure, run these commands to verify the nodes are correctly labeled and tainted. This ensures the Hybrid Manager installer can schedule pods correctly.

1. Validate Control Plane nodes:

   → WrapCopy

   kubectl get nodes -l edbaiplatform.io/control-plane="true"

2. Verify taints

   → WrapCopy

   kubectl get nodes -o json | jq '.items[] | select(.metadata.labels["edbaiplatform.io/control-plane"]=="true") | {name: .metadata.name, taints: (.spec.taints // [] | map(select(.key=="edbaiplatform.io/control-plane")))}'

3. Validate Data Plane nodes:

   → WrapCopy

   kubectl get nodes -l edbaiplatform.io/postgres="true"

4. Verify taints

   → WrapCopy

   kubectl get nodes -o json | jq '.items[] | {name: .metadata.name, labels: .metadata.labels["edbaiplatform.io/postgres"], taints: (.spec.taints // [] | map(select(.key=="edbaiplatform.io/postgres")))}'

Block storage configuration

Source: Gathering your system requirements>6. Block storage (Database & logs)

• Storage class: Ensure the cluster has the Storage Class default you identified in Phase 2 for the Control Plane (e.g., gp2).
• Capability: If desired, you may establish any number of storage classes for your Data Plane (Postgres workloads), and they are then options to use when provisioning data base clusters using HM.

Ingress

Source: Gathering your system requirements>5. Ingress

• Ingress controller: Verify that your chosen Ingress mechanism (LoadBalancer or NodePort) is active and exposing the required ports.

• DNS: Ensure CoreDNS (or equivalent) is healthy and resolving internal service names.

Post-deployment validation

Once your cluster is Active or Ready, perform these quick checks to ensure it is ready for the Preparing your environment phase.

Verify API Access

From your management workstation, ensure you can reach the API server.

kubectl cluster-info

• Success: Returns the Kubernetes control plane URL.

Verify Storage Capability

Confirm a valid storage class is present and set to (default).

kubectl get sc

• Success: Output lists a class (e.g., gp3) with (default) in the name.

Verify Identity (Cloud Only)

If you are on EKS/GKE/AKS, verify that your current IAM identity has administrative rights.

kubectl auth can-i create secrets --all-namespaces

• Success: Output is yes.

Next phase

Your infrastructure is provisioned. You can now connect to this cluster.

Next sync images to your local registry, stage your secrets, configure TLS, set up any relevant advanced features, and apply the HM configuration file: values.yaml.

Proceed to Phase 4: Preparing your environment →