Skip to main content

Deploy a PCG to an Existing Kubernetes Cluster

For infrastructure environments that the Palette CLI does not support, you can deploy a Private Cloud Gateway (PCG) to an existing Kubernetes cluster. The PCG can be installed on any Kubernetes cluster that meets the prerequisites, regardless of the underlying infrastructure. For example, you can host the PCG on an EKS cluster, a self-managed cluster on EC2, or any other cluster that is not managed by Palette.

Supported Cloud Types​

During PCG creation, you select a Cloud type that determines which cloud environment you can deploy workload clusters into through the PCG, including:

  • AWS

  • Azure

  • Technical preview feature badgeTechnical preview feature badge Nutanix

    info

    To select Nutanix as the PCG cloud type, you must Register Nutanix Cloud with Palette.

PCG Sizing​

The following table provides the recommended sizing for the PCG based on the number of nodes, CPU, memory, storage, and the maximum concurrent cluster deployments. You can continue to deploy additional clusters once the current clusters deployment batch is complete.

We recommend using a minimum of three nodes for production environments. Single node clusters are better suited for development and testing environments.

Single-Node Cluster​

SizeNodesCPUMemoryStorageMaximum concurrent cluster deployments
Small144 GB60 GB1-3
Medium188 GB100 GB4-6
Large11616 GB120 GB7-10

High-Availability (HA) Cluster​

SizeNodesCPUMemoryStorageMaximum concurrent cluster deployments
Small344 GB60 GB4-6
Medium388 GB100 GB7-10
Large31616 GB120 GB10-15

Prerequisites​

  • An existing Kubernetes cluster that has network connectivity with Palette. The Kubernetes cluster must be able to connect to the internet to download the required images and packages and connect with Palette on port 443. To learn more about PCG network requirements, refer to the Network Communication section.

    warning

    Do not install the PCG in a Kubernetes cluster managed by Palette. The installation will fail due to resource naming conflicts.

  • Sufficient permissions to deploy the PCG services in the cluster. The default Kubernetes cluster roles cluster-admin and admin have the necessary permissions to deploy the PCG services.

  • The Kubernetes cluster must have at least the following resources:

    • 4 vCPUs
    • 4 GB of memory
    • 60 GB of disk space
    • One control plane node with worker capabilities or three nodes for production environments. Refer to the PCG Sizing section for more information.
    • A Container Network Interface (CNI) plugin installed.
    • A Container Storage Interface (CSI) plugin installed.
    • A version of Kubernetes that is compatible to your Palette version. Refer to the Kubernetes Requirements section to find the required Kubernetes version.

  • PCG IP address requirements:

    • One IP address for a single-node PCG or three IP addresses for a three-node PCG. Refer to the PCG Sizing section for more information on sizing.
    • One IP address reserved for cluster repave operations.
    • One IP address for the virtual IP (VIP).
    • DNS can resolve the domain api.spectrocloud.com or the domain that you have configured for a self-hosted Palette installation.

  • (Optional) If you want Palette to display cluster metrics, ensure that the metrics server is installed in the cluster. You can install the metrics server using the following command.

    helm repo add metrics-server https://kubernetes-sigs.github.io/metrics-server/
    helm upgrade --install metrics-server metrics-server/metrics-server \
     --namespace kube-system \
     --set args='{--kubelet-insecure-tls}'

    Add the line --set args='{--kubelet-insecure-tls}' to the command if your cluster uses self-signed Kubelet certificates, which is common for self-managed clusters. If you do not use the --kubelet-insecure-tls argument and you have self-signed Kubelet certificates, the metrics API server will not respond.

Install PCG​

Use the following steps to install the PCG in your existing Kubernetes cluster.

  1. Log in to Palette as a tenant admin.

  2. From the left main menu, select Tenant Settings.

  3. From the Tenant Settings menu, select Private Cloud Gateways > Add New Private Cloud Gateway.

  4. On the Private Cloud Gateway installation instructions modal window, select Self Hosted.

  5. Enter a Private cloud gateway name and use the Cloud type drop-down menu to select the target infrastructure provider.

  6. Create your PCG when finished. A set of instructions with commands is displayed on the drawer.

    View of the cluster details page with the side drawer extended that contains the kubectl commands

  7. Select Download manifest.

    info

    Once you select Download manifest, the cluster status transitions to Importing. If you close the drawer and later discover the download was blocked or did not complete, the drawer will not reopen. In this case, you must fetch the manifests using the Palette API endpoints /v1/pcg/<pcg-uid>/services/jet/manifest and /v1/pcg/<pcg-uid>/services/ally/manifest.

    Retrieve Manifests Manually

    Replace the placeholders with the appropriate values.

    ParameterDescription
    <palette-endpoint>The endpoint of your Palette instance.
    <cluster-uid>The unique ID assigned to imported cluster being used as a PCG. Found at Tenant Settings > Private Cloud Gateways > [PCG] > Cluster ID.
    <pcg-uid>The unique ID assigned to the PCG. Must be retrieved using the cluster UID.
    <api-key>Your Palette API Key.
    <jwt-token>Your Palette Authorization Token.

    1. Fetch the <pcg-uid> associated with the <cluster-uid>.

      curl --silent --request GET \
        "https://<palette-endpoint>/v1/overlords" \
        --header "ApiKey: <api-key>" | \
        jq -r --arg cid "<cluster-uid>" \
        '.items[] | select(.spec.spectroClusterUid == $cid) | .metadata.uid'

    2. Retrieve the Jet manifest.

      curl --silent --show-error --request GET \
          "https://<palette-endpoint>/v1/pcg/<pcg-uid>/services/jet/manifest" \
          --header "ApiKey: <api-key>" \
          --output "cluster-<pcg-uid>-jet-manifest.yaml"

    3. Retrieve the Ally manifest.

      curl --silent --show-error --request GET \
        "https://<palette-endpoint>/v1/pcg/<pcg-uid>/services/ally/manifest" \
        --header "ApiKey: <api-key>" \
        --output "cluster-<pcg-uid>-ally-manifest.yaml"

  8. Apply the downloaded manifests to your Kubernetes cluster using the appropriate download paths.

    kubectl apply --filename "<path-to-download>/cluster-*-jet-manifest.yaml"
    kubectl apply --filename "<path-to-download>/cluster-*-ally-manifest.yaml"

  9. When the agents initialize, the drawer disappears, and your Cluster Status transitions to Running. Within a few minutes, your cluster's Health status changes to Healthy.

    tip

    You can provide network proxy configurations to your Kubernetes clusters deployed through Palette. To provide network proxy configurations to your host clusters, update the PCG with the proxy server details. To learn more, check out the Proxy Configuration guide.

Validate​

When deployed, the PCG registers itself with Palette. Use the steps below to verify if the PCG registration is successful.

  1. Log in to Palette as a tenant admin.

  2. From the left main menu, select Tenant Settings.

  3. From the Tenant Settings menu, select Private Cloud Gateways, and choose your PCG.

  4. On the Private Cloud Gateway Overview screen, verify the Cluster Status is Running.

Next Steps​

Once your PCG is in the Running state, you can add an AWS, Azure, or Nutanix cloud account to Palette with the Connect Private Cloud Gateway option toggled and your self-hosted PCG selected. You can then use the PCG to deploy and delete clusters in the cloud environment connected to the PCG.

If you need to configure a proxy server for your PCG, refer to Enable and Manage Proxy Configurations. The proxy must be configured before deploying workload clusters through the PCG in order for the workload clusters to inherit the PCG proxy configurations.