Install Palette on EKS with ECR Mirroring
This guide walks you through installing self-hosted Palette on an Amazon Elastic Kubernetes Service (EKS) cluster in an airgapped environment. All container images, packs, and manifest artifacts are mirrored from Spectro Cloud's registry into your own Amazon Elastic Container Registry (ECR) so that your cluster never needs to reach external registries.
Prerequisites
Before you begin, ensure you meet the following prerequisites.
Use Helm client v3.14.0 or later. Older Helm clients do not wait for pre-upgrade hook resources to finish
terminating before recreating them. When you run helm upgrade against the hubble chart with a client earlier than
v3.14.0, the hubble-system namespace can be inadvertently deleted mid-upgrade, which results in loss of MongoDB
data — including tenants and cluster metadata — with no automatic recovery.
The fix for this hook behavior was released in
Helm v3.14.0. Confirm your client version
with helm version before you continue.
AWS Infrastructure
-
An AWS account with permissions to create ECR repositories.
-
An EKS cluster with three worker nodes with an instance type of
t3a.2xlargeor larger. -
A Container Storage Interface (CSI) driver installed and running on the EKS cluster, such as the AWS EBS CSI driver.
-
A custom domain you control with the ability to create wildcard DNS records.
Jump Host
-
A jump host with network access to the EKS cluster.
-
The following tools installed on your jump host:
-
AWS CLI v2 - Required to interact with AWS ECR and EKS.
-
kubectl - Required to connect to and manage your EKS cluster. Use the following command to add your EKS cluster details to your kubeconfig at
$HOME/.kube/config. Replace<region-code>with the region of your cluster and<cluster-name>with the name of your cluster.aws eks update-kubeconfig --region <region-code> --name <cluster-name> -
Helm v3.14.0 or later - Required to install the Palette Helm charts.
-
crane - Required to copy container images between registries without a local Docker daemon.
-
A text editor of your choice, such as
nanoorvi- Required to modify Palette Helm charts.
-
-
The following artifacts from our Support team:
-
Palette Helm charts, received as a
charts.zipfile. Refer to Access Palette for instructions on how to request access. -
The following artifact list files that correspond with your
<palette-version>. If you require additional packs for your workload clusters beyond what is supplied, contact our Support team to receive updated files containing additional packs and images.packs.txt- List of pack names to mirror.images.txt- List of all container images to mirror.
-
Spectro Cloud ECR Credentials
-
The following Spectro Cloud ECR details from our Support team. These values are used throughout the guide.
Detail Placeholder Spectro Cloud ECR URL <spectro-ecr-registry-url>Spectro Cloud ECR region <spectro-ecr-region>Spectro Cloud base path <spectro-base-path>AWS Access Key <spectro-access-key>AWS Secret Key <spectro-secret-key>Palette version <palette-version>
Set Up Registries
Create two base repositories: one for images (always public) and one for packs (public or private). Public ECRs do not require authentication, whereas private ECRs use long-lived IAM Access and Secret keys to restrict access.
Create Public ECR and Mirror Images
Public ECR is globally accessible for pulls without authentication. However, AWS centralizes the management API in
us-east-1. All aws ecr-public commands must use --region us-east-1, regardless of where your cluster runs.
-
On your jump host, ensure your AWS CLI is configured with an identity that has permission to create an ECR and push images to an ECR.
aws sts get-caller-identityVerify the output displays an identity with ECR permissions. If you use AWS IAM Identity Center, you can configure a named profile with
aws configure sso.Example output{
"UserId": "ABCD1234EFGH5678:user@spectrocloud.com",
"Account": "012345678901",
"Arn": "arn:aws:sts::012345678901:assumed-role/AWSReservedSSO_AdministratorAccess_abcd1234efgh/user@spectrocloud.com"
} -
Retrieve your public ECR alias. This value is unique per AWS account.
ECR_PUBLIC_ALIAS=$(aws ecr-public describe-registries --region us-east-1 \
--query 'registries[0].aliases[0].name' --output text) -
Create the base repository. Replace
<image-base-path>with a namespace prefix for your images, such asspectro-images.IMAGE_BASE_PATH="<image-base-path>"
aws ecr-public create-repository \
--repository-name $IMAGE_BASE_PATH \
--region us-east-1 -
Set the source and destination variables for Spectro Cloud's ECR. Replace each placeholder with the values provided by our Support team in the Spectro Cloud ECR Credentials section.
infoDo not include the protocol in the
SPECTRO_ECRURL.SPECTRO_ECR="<spectro-ecr-registry-url>"
SPECTRO_REGION="<spectro-ecr-region>"
SPECTRO_BASE="<spectro-base-path>"
SPECTRO_ACCESS_KEY="<spectro-access-key>"
SPECTRO_SECRET_KEY="<spectro-secret-key>"
IMAGE_DEST_ECR="public.ecr.aws/$ECR_PUBLIC_ALIAS" -
Authenticate to the Spectro Cloud ECR.
crane auth login "$SPECTRO_ECR" \
--username AWS \
--password "$(AWS_ACCESS_KEY_ID=$SPECTRO_ACCESS_KEY AWS_SECRET_ACCESS_KEY=$SPECTRO_SECRET_KEY \
aws ecr get-login-password --region "$SPECTRO_REGION")"Example output2026/03/19 11:54:57 logged in via /Users/ubuntu/.docker/config.json -
Authenticate with your newly created public ECR.
crane auth login public.ecr.aws \
--username AWS \
--password "$(aws ecr-public get-login-password --region us-east-1)"Example output2026/03/19 11:55:38 logged in via /Users/ubuntu/.docker/config.json -
Navigate to the directory containing the
images.txtandpacks.txtfiles received from our Support team.cd <images-and-packs-location> -
Mirror the images from the Spectro Cloud ECR to your public ECR. The following loop reads each image from the
images.txtfile in the current directory, creates the target repository in your public ECR, and copies the image from its original public registry.while IFS= read -r img; do
[[ -z "$img" || "$img" == \#* ]] && continue
repo="${img%%:*}"
tag="${img##*:}"
aws ecr-public create-repository \
--repository-name "$IMAGE_BASE_PATH/spectro-images/$repo" \
--region us-east-1 2>/dev/null || true
(
crane copy --platform linux/amd64 \
"$img" \
"$IMAGE_DEST_ECR/$IMAGE_BASE_PATH/spectro-images/$img" \
&& echo "OK: $img" || echo "FAIL: $img"
) &
while (( $(jobs -r | wc -l) >= 5 )); do sleep 1; done
done < images.txt
wait
echo "All done" -
Verify the image mirroring completed by spot-checking a few images.
Example commandaws ecr-public describe-images \
--repository-name "$IMAGE_BASE_PATH/spectro-images/us-docker.pkg.dev/palette-images/palette/crony" \
--region us-east-1Example output{
"imageDetails": [
{
"registryId": "123456789",
"repositoryName": "palette/spectro-images/us-docker.pkg.dev/palette-images/palette/crony",
"imageDigest": "sha256:694efe36e5e3221d73261a6c6db4fc82c758d7750e0ced9468e6b2a9ccc072ca",
"imageTags": ["4.8.6"],
"imageSizeInBytes": 28822852,
"imagePushedAt": "2026-03-17T18:29:36.777000-04:00",
"imageManifestMediaType": "application/vnd.docker.distribution.manifest.v2+json",
"artifactMediaType": "application/vnd.docker.container.image.v1+json"
}
]
}
Create Pack Registry and Mirror Packs and Manifest
Choose one of the following options depending on whether you want to use a private or public ECR for packs.
- Public ECR
- Private ECR
-
If using a public ECR for packs, issue the following command. Replace
<pack-base-path>with the repository name for your packs, such asspectro-packs.PACK_BASE_PATH="<pack-base-path>"
aws ecr-public create-repository \
--repository-name $PACK_BASE_PATH \
--region us-east-1Example output{
"repository": {
"repositoryArn": "arn:aws:ecr-public::12345789:repository/palette-packs",
"registryId": "12345789",
"repositoryName": "palette-packs",
"repositoryUri": "public.ecr.aws/a1b2c3d4/palette-packs",
"createdAt": "2026-03-19T14:34:56.862000-04:00"
},
"catalogData": {}
} -
Set the destination variables. Replace
<palette-version>with the version of your Palette installation.PACK_DEST_ECR="public.ecr.aws/$ECR_PUBLIC_ALIAS"
PALETTE_VERSION="<palette-version>" -
Mirror the packs.
while IFS= read -r pack; do
[[ -z "$pack" || "$pack" == \#* ]] && continue
aws ecr-public create-repository \
--repository-name "$PACK_BASE_PATH/spectro-packs/archive/${pack%%:*}" \
--region us-east-1 2>/dev/null || true
crane copy \
"$SPECTRO_ECR/$SPECTRO_BASE/spectro-packs/archive/$pack" \
"$PACK_DEST_ECR/$PACK_BASE_PATH/spectro-packs/archive/$pack"
done < packs.txt
wait
echo "All done" -
Verify the pack mirroring completed by spot-checking a few packs.
Example commandaws ecr-public describe-images \
--repository-name "$PACK_BASE_PATH/spectro-packs/archive/ubuntu-aws" \
--region us-east-1Example output{
"imageDetails": [
{
"registryId": "123456789",
"repositoryName": "spectro-packs/spectro-packs/archive/ubuntu-aws",
"imageDigest": "sha256:c134ae8b14bbdae1fa9a821ed0729f35a7875b5a69b31cb6000c7616f42fd977",
"imageTags": ["22.04"],
"imageSizeInBytes": 5774,
"imagePushedAt": "2026-03-18T12:03:20.549000-04:00",
"imageManifestMediaType": "application/vnd.oci.image.manifest.v1+json",
"artifactMediaType": "application/vnd.unknown.config.v1+json"
},
{
"registryId": "123456789",
"repositoryName": "spectro-packs/spectro-packs/archive/ubuntu-aws",
"imageDigest": "sha256:e5982cf244ea436923524ec7453880633b14f77fe2457973f65428c97c5bbfe0",
"imageTags": ["24.04"],
"imageSizeInBytes": 6198,
"imagePushedAt": "2026-03-18T12:03:23.773000-04:00",
"imageManifestMediaType": "application/vnd.oci.image.manifest.v1+json",
"artifactMediaType": "application/vnd.unknown.config.v1+json"
}
]
} -
Mirror the manifest.
aws ecr-public create-repository \
--repository-name "$PACK_BASE_PATH/spectro-manifests/manifest" \
--region us-east-1 2>/dev/null || true
crane copy \
"$SPECTRO_ECR/$SPECTRO_BASE/spectro-manifests/manifest:$PALETTE_VERSION" \
"$PACK_DEST_ECR/$PACK_BASE_PATH/spectro-manifests/manifest:$PALETTE_VERSION" -
Verify the manifest mirroring completed.
Example commandaws ecr-public describe-images \
--repository-name "$PACK_BASE_PATH/spectro-manifests/manifest" \
--region us-east-1Example output{
"imageDetails": [
{
"registryId": "123456789",
"repositoryName": "spectro-packs/spectro-manifests/manifest",
"imageDigest": "sha256:f03ad0fc7d8ce4f38b49277343563f72b63c5263794e00908f5023678b8b2972",
"imageTags": ["4.8.37"],
"imageSizeInBytes": 1681032,
"imagePushedAt": "2026-03-18T12:05:56.714000-04:00",
"imageManifestMediaType": "application/vnd.oci.image.manifest.v1+json",
"artifactMediaType": "application/vnd.unknown.config.v1+json"
}
]
}
-
If using a private ECR for packs, issue the following command. Replace
<pack-base-path>with the repository name for your packs, such asspectro-packs, and<your-region>with your AWS ECR region.PACK_BASE_PATH="<pack-base-path>"
PACK_REGION="<your-region>"
aws ecr create-repository \
--repository-name $PACK_BASE_PATH \
--region $PACK_REGIONExample output{
"repository": {
"repositoryArn": "arn:aws:ecr:us-east-1:123456789:repository/palette",
"registryId": "123456789",
"repositoryName": "palette",
"repositoryUri": "123456789.dkr.ecr.us-east-1.amazonaws.com/palette",
"createdAt": "2026-03-19T14:33:13.545000-04:00",
"imageTagMutability": "MUTABLE",
"imageScanningConfiguration": {
"scanOnPush": false
},
"encryptionConfiguration": {
"encryptionType": "AES256"
}
}
} -
Set the following variables. Replace
<your-account-id>with your AWS account ID and<palette-version>with the version of your Palette installation.PACK_DEST_ECR="<your-account-id>.dkr.ecr.$PACK_REGION.amazonaws.com"
PALETTE_VERSION="<palette-version>" -
Authenticate to your private ECR.
crane auth login "$PACK_DEST_ECR" \
--username AWS \
--password "$(aws ecr get-login-password --region "$PACK_REGION")"Example output2026/03/19 14:39:04 logged in via /Users/ubuntu/.docker/config.json -
Mirror the packs.
while IFS= read -r pack; do
[[ -z "$pack" || "$pack" == \#* ]] && continue
aws ecr create-repository \
--repository-name "$PACK_BASE_PATH/spectro-packs/archive/${pack%%:*}" \
--region "$PACK_REGION" 2>/dev/null || true
crane copy \
"$SPECTRO_ECR/$SPECTRO_BASE/spectro-packs/archive/$pack" \
"$PACK_DEST_ECR/$PACK_BASE_PATH/spectro-packs/archive/$pack"
done < packs.txt
wait
echo "All done" -
Verify the pack mirroring completed by spot-checking a few packs.
Example commandaws ecr list-images \
--repository-name "$PACK_BASE_PATH/spectro-packs/archive/ubuntu-aws" \
--region "$PACK_REGION"Example output{
"imageIds": [
{
"imageDigest": "sha256:c134ae8b14bbdae1fa9a821ed0729f35a7875b5a69b31cb6000c7616f42fd977",
"imageTag": "22.04"
},
{
"imageDigest": "sha256:e5982cf244ea436923524ec7453880633b14f77fe2457973f65428c97c5bbfe0",
"imageTag": "24.04"
}
]
} -
Mirror the manifest.
aws ecr create-repository \
--repository-name "$PACK_BASE_PATH/spectro-manifests/manifest" \
--region "$PACK_REGION" 2>/dev/null || true
crane copy \
"$SPECTRO_ECR/$SPECTRO_BASE/spectro-manifests/manifest:$PALETTE_VERSION" \
"$PACK_DEST_ECR/$PACK_BASE_PATH/spectro-manifests/manifest:$PALETTE_VERSION" -
Verify the manifest mirroring completed.
aws ecr list-images \
--repository-name "$PACK_BASE_PATH/spectro-manifests/manifest" \
--region "$PACK_REGION"Example output{
"imageIds": [
{
"imageDigest": "sha256:f03ad0fc7d8ce4f38b49277343563f72b63c5263794e00908f5023678b8b2972",
"imageTag": "4.8.37"
}
]
}
Install Palette
Once your registries are set up and populated with the necessary images and packs, you are ready to install Palette on your EKS cluster. The installation process requires several Helm charts, some which must be installed before others. Following is the recommended installation order:
- Cert-Manager - TLS certificate management, required by Palette. Must be installed before Image Swap.
- Image Swap - Webhook that rewrites image references to your public ECR at pod admission time. Must be installed before the Palette Management Plane.
- Spectro Management CRDs - Chart that contains Custom Resource Definitions (CRDs) required by Palette. Can be installed at any point before the Palette Management Plane, as it only registers API types and does not deploy pods.
- Palette Management Plane - The main Palette application. Must be installed last.
Verify Environment Setup
-
Confirm all nodes are in
Readystate.kubectl get nodesExample outputNAME STATUS ROLES AGE VERSION
ip-172-31-15-154.ec2.internal Ready <none> 25h v1.33.8-eks-f69f56f
ip-172-31-47-72.ec2.internal Ready <none> 25h v1.33.8-eks-f69f56f
ip-172-31-87-121.ec2.internal Ready <none> 25h v1.33.8-eks-f69f56f -
Confirm a CSI driver is installed.
kubectl get csidriversExample outputNAME ATTACHREQUIRED PODINFOONMOUNT STORAGECAPACITY TOKENREQUESTS REQUIRESREPUBLISH MODES AGE
ebs.csi.aws.com true false false <unset> false Persistent 24h
efs.csi.aws.com false false false <unset> false Persistent 25h -
Confirm the CSI driver pods are running.
kubectl get pods --namespace kube-system | grep csiExample outputebs-csi-controller-549989b6f8-9zd68 6/6 Running 0 24h
ebs-csi-controller-549989b6f8-t5tsm 6/6 Running 0 24h
ebs-csi-node-8rddx 3/3 Running 0 24h
ebs-csi-node-wrw4c 3/3 Running 0 24h
ebs-csi-node-zb8mp 3/3 Running 0 24hinfoIf the AWS EBS CSI driver is not installed, you can add it as an EKS add-on. Issue the following command, replacing
<your-eks-cluster-name>and<your-region>with the appropriate values.aws eks create-addon \
--cluster-name <your-eks-cluster-name> \
--addon-name aws-ebs-csi-driver \
--region <your-region>
Install Cert-Manager
Cert-Manager must be installed first, as Palette depends on it for TLS. Because Image Swap is not yet running, Cert-Manager image references must be explicitly set to their full public ECR paths.
-
Open
charts/extras/cert-manager/values.yamlin a text editor and update all four image fields to their full public ECR paths. Replace<your-public-alias>and<image-base-path>with your values. Do not change the existing<image-tag>, as it corresponds to the Palette version used for the Palette images in your public ECR.tipUse the following command to retrieve your AWS public alias.
aws ecr-public describe-registries --region us-east-1 \
--query 'registries[0].aliases[0].name' --output textimage:
cainjectorImage: "public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/us-docker.pkg.dev/palette-images-fips/palette/spectro-cert-manager/cert-manager-cainjector:<image-tag>"
controllerImage: "public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/us-docker.pkg.dev/palette-images-fips/palette/spectro-cert-manager/cert-manager-controller:<image-tag>"
webhookImage: "public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/us-docker.pkg.dev/palette-images-fips/palette/spectro-cert-manager/cert-manager-webhook:<image-tag>"
amceResolverImage: "public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/us-docker.pkg.dev/palette-images-fips/palette/spectro-cert-manager/cert-manager-acmesolver:<image-tag>"Example Cert-Manager configurationimage:
cainjectorImage: "public.ecr.aws/a1b2c3d4/palette/spectro-images/us-docker.pkg.dev/palette-images-fips/palette/spectro-cert-manager/cert-manager-cainjector:v1.19.3-spectro-4.8.b"
controllerImage: "public.ecr.aws/a1b2c3d4/palette/spectro-images/us-docker.pkg.dev/palette-images-fips/palette/spectro-cert-manager/cert-manager-controller:v1.19.3-spectro-4.8.b"
webhookImage: "public.ecr.aws/a1b2c3d4/palette/spectro-images/us-docker.pkg.dev/palette-images-fips/palette/spectro-cert-manager/cert-manager-webhook:v1.19.3-spectro-4.8.b"
amceResolverImage: "public.ecr.aws/a1b2c3d4/palette/spectro-images/us-docker.pkg.dev/palette-images-fips/palette/spectro-cert-manager/cert-manager-acmesolver:v1.19.3-spectro-4.8.b" -
Install Cert-Manager on your EKS cluster using the following command.
helm upgrade --install cert-manager \
charts/extras/cert-manager/cert-manager-*.tgz \
--namespace cert-manager \
--create-namespace \
--values charts/extras/cert-manager/values.yaml \
--wait --timeout 10m -
Verify all Cert-Manager pods are running.
kubectl get pods --namespace cert-managerExample OutputNAME READY STATUS RESTARTS AGE
cert-manager-5d87667c6c-dzv2d 1/1 Running 0 5m49s
cert-manager-cainjector-78df6d579b-pfp7b 1/1 Running 0 5m49s
cert-manager-webhook-7b9fc85668-sv6dp 1/1 Running 0 5m49s
Install Image Swap
Image Swap is a Kubernetes admission webhook. Once running, it intercepts every pod creation and rewrites image references to point to your public ECR. This means the main Palette Helm chart requires no image path changes.
-
Open
charts/extras/image-swap/image-swap/values.yamlin a text editor and configure the following sections with the necessary values. Do not change the existing<image-tag>, as it corresponds to the Palette version used for the Palette images in your public ECR.config:
imageSwapImages:
imageSwapInitImage: "public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/us-docker.pkg.dev/palette-images/third-party/thewebroot/imageswap-init:v1.5.3-spectro-<image-tag>"
imageSwapImage: "public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/us-docker.pkg.dev/palette-images/third-party/thewebroot/imageswap:v1.5.3-spectro-<image-tag>"
imageSwapConfig:
isEKSCluster: true
ociImageRegistry:
endpoint: "public.ecr.aws/<your-public-alias>"
name: "<your-registry-name>"
username: ""
password: ""
baseContentPath: "<image-base-path>/spectro-images"
insecureSkipVerify: false
caCert: ""
mirrorRegistries: "us-docker.pkg.dev::public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/us-docker.pkg.dev,gcr.io::public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/gcr.io,registry.k8s.io::public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/registry.k8s.io"
imagePullSecret:
create: false
dockerConfigJson: "" -
Install Image Swap using the following command.
helm upgrade --install image-swap \
charts/extras/image-swap/image-swap-*.tgz \
--values charts/extras/image-swap/image-swap/values.yaml \
--wait --timeout 10m -
Verify all Image Swap pods are running.
kubectl get pods --namespace imageswap-systemExample outputNAME READY STATUS RESTARTS AGE
imageswap-55f97c9fdd-7whm9 1/1 Running 0 56s
imageswap-55f97c9fdd-hm47s 1/1 Running 0 56s -
Confirm the webhook is registered.
kubectl get mutatingwebhookconfigurations | grep imageswapExample outputimageswap-webhook 1 65s
Install Custom Resource Definitions
The Spectro Management CRDs chart contains the CRDs required by Palette, including Traefik CRDs, and must be installed before the main Palette Helm chart. When the chart is installed, the custom resource types are registered with the Kubernetes API server; no pods are deployed.
helm upgrade --install spectro-mgmt-crds \
charts/extras/spectro-mgmt-crds/spectro-mgmt-crds-*.tgz \
--values charts/extras/spectro-mgmt-crds/values.yaml \
--wait --timeout 10m
Install Palette
Finally, install Palette. Image Swap handles all image rewriting automatically, so no changes to the image fields in
charts/palette/values.yaml are needed.
-
Open
charts/palette/values.yamlin a text editor and update the following sections.warningDo not change any other fields unless instructed by our Support team.
-
Set the installation mode to airgap.
config:
installationMode: "airgap" -
Set the root domain. This domain will be used when creating a wildcard DNS record (
*.<your-root-domain>) for yourtraefik-ingress-controllerload balancer.env:
rootDomain: "<your-root-domain>" -
Set the MongoDB storage class. First, check what storage classes are available on your cluster.
kubectl get storageclassExample outputNAME PROVISIONER RECLAIMPOLICY VOLUMEBINDINGMODE ALLOWVOLUMEEXPANSION AGE
gp2 kubernetes.io/aws-ebs Delete WaitForFirstConsumer false 26hThen use the name of your preferred storage class in the following configuration.
mongo:
storageClass: "<storage-class-name>" -
Configure the pack registry. This points to your public or private ECR configured in the Create Pack Registry and Mirror Packs and Manifest section.
- Public ECR
- Private ECR
ociPackEcrRegistry:
endpoint: "public.ecr.aws/<your-public-alias>"
name: "<your-pack-registry-name>"
accessKey: ""
secretKey: ""
baseContentPath: "<pack-base-path>"
isPrivate: false
insecureSkipVerify: false
caCert: ""ociPackEcrRegistry:
endpoint: "<your-account-id>.dkr.ecr.<your-region>.amazonaws.com"
name: "<your-pack-registry-name>"
accessKey: "<your-access-key>"
secretKey: "<your-secret-key>"
baseContentPath: "<pack-base-path>"
isPrivate: true
insecureSkipVerify: false
caCert: "" -
Configure the Image Swap section. These values must match the
imageSwapImagesandimageSwapConfigblocks fromcharts/extras/image-swap/image-swap/values.yaml, as configured in the Install Image Swap section.imageSwapImages:
imageSwapInitImage: "public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/us-docker.pkg.dev/palette-images/third-party/thewebroot/imageswap-init:v1.5.3-spectro-<image-tag>"
imageSwapImage: "public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/us-docker.pkg.dev/palette-images/third-party/thewebroot/imageswap:v1.5.3-spectro-<image-tag>"
imageSwapConfig:
isEKSCluster: true -
Copy the
ociImageRegistryblock fromcharts/extras/image-swap/image-swap/values.yaml, as configured in the Install Image Swap section, and replace the commented outociImageRegistrysection incharts/palette/values.yaml.ociImageRegistry:
endpoint: "public.ecr.aws/<your-public-alias>"
name: "<your-registry-name>"
username: ""
password: ""
baseContentPath: "<image-base-path>/spectro-images"
insecureSkipVerify: false
caCert: ""
mirrorRegistries: "us-docker.pkg.dev::public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/us-docker.pkg.dev,gcr.io::public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/gcr.io,registry.k8s.io::public.ecr.aws/<your-public-alias>/<image-base-path>/spectro-images/registry.k8s.io" -
Install Palette using the following command.
helm upgrade --install palette \
charts/palette/spectro-mgmt-plane-*.tgz \
--namespace spectro-mgmt-plane \
--create-namespace \
--values charts/palette/values.yaml \
--wait --timeout 20m -
Monitor the installation. Watch pods come up across all namespaces.
kubectl get pods --all-namespaces --watch -
Verify the installation is complete. Installation is complete when all pods in the
cp-system,hubble-system,ingress-traefik,jet-system, andui-systemnamespaces are in aRunningorCompletedstate.warningThe Traefik ingress controller, which is installed by default beginning with Palette version 4.8.47, configures the non-root security context
NET_BIND_SERVICELinux capability in the Traefik DaemonSet. This capability is not supported on deprecated Amazon Linux 2 (AL2) nodes, resulting in aCrashLoopBackOffstate ontraefik-ingress-controllerpods. Refer to our Troubleshooting - Traefik Ingress Controller Pods Crash on Amazon Linux 2 Nodes guide for the workaround.kubectl get pods --all-namespaces | grep --extended-regexp '(cp-system|hubble-system|ingress-traefik|jet-system|ui-system)'Example outputcp-system spectro-cp-ui-54bd94cc75-2g67f 1/1 Running 0 8m36s
hubble-system auth-b995974d-55djq 1/1 Running 0 8m35s
hubble-system auth-b995974d-8x2jn 1/1 Running 0 8m34s
hubble-system cloud-67496d8476-j978v 1/1 Running 0 8m33s
hubble-system cloud-67496d8476-l5mk7 1/1 Running 0 8m34s
hubble-system configserver-64ccbdb879-2l5bc 1/1 Running 0 8m32s
hubble-system event-8647fc6599-4f5zw 1/1 Running 0 8m32s
hubble-system event-8647fc6599-4zp2x 1/1 Running 0 8m31s
hubble-system event-8647fc6599-znx6r 1/1 Running 0 8m31s
hubble-system foreq-5b756dd89c-hxbzf 1/1 Running 0 8m30s
hubble-system hashboard-5d757476fd-t8ss7 1/1 Running 0 8m29s
hubble-system hashboard-5d757476fd-x22j7 1/1 Running 0 8m29s
hubble-system hutil-7bd9b87dbb-fzbpj 1/1 Running 0 8m28s
hubble-system hutil-7bd9b87dbb-nd79p 1/1 Running 0 8m28s
hubble-system memstore-7bb7b88cb9-g8gsh 1/1 Running 0 8m27s
hubble-system mgmt-765d456b8f-57lpk 1/1 Running 0 50m
hubble-system mongo-0 2/2 Running 0 5m17s
hubble-system mongo-1 2/2 Running 0 6m32s
hubble-system mongo-2 2/2 Running 0 8m17s
hubble-system mongodb-key-manager-helm-7tbqj 0/1 Completed 0 8m24s
hubble-system msgbroker-0 1/1 Running 0 8m23s
hubble-system msgbroker-1 1/1 Running 0 8m12s
hubble-system oci-proxy-7db57df545-f9rgx 1/1 Running 0 50m
hubble-system specman-0 1/1 Running 0 50m
hubble-system spectro-tunnel-65f76dc8bc-z6m9m 1/1 Running 0 50m
hubble-system spectrocluster-bb6c497b5-5gxhm 1/1 Running 0 8m21s
hubble-system spectrocluster-bb6c497b5-t7kwl 1/1 Running 0 8m21s
hubble-system spectrocluster-bb6c497b5-zwn4q 1/1 Running 0 8m22s
hubble-system spectrocluster-jobs-6697cd8bf5-xt5km 1/1 Running 0 8m20s
hubble-system spectrocluster-reconciler-6976b69644-gtmq9 1/1 Running 0 8m19s
hubble-system spectroclusterop-7b8856c7c5-9fbmz 1/1 Running 0 8m19s
hubble-system spectroclusterop-7b8856c7c5-cmxcz 1/1 Running 0 8m18s
hubble-system spectrossh-6f66c865bd-kr6bm 1/1 Running 0 8m17s
hubble-system system-77b6f5fc9b-8pldv 1/1 Running 1 (6m19s ago) 8m16s
hubble-system system-77b6f5fc9b-pt28f 1/1 Running 1 (6m23s ago) 8m16s
hubble-system timeseries-55b4dd666-hflc5 1/1 Running 0 8m14s
hubble-system timeseries-55b4dd666-k8x8f 1/1 Running 0 8m15s
hubble-system timeseries-55b4dd666-lkw6r 1/1 Running 0 8m14s
hubble-system user-78dc58468d-cm2hk 1/1 Running 0 8m13s
hubble-system user-78dc58468d-hqbzm 1/1 Running 0 8m13s
ingress-traefik traefik-ingress-controller-9dmzq 1/1 Running 0 51m
ingress-traefik traefik-ingress-controller-tpwtf 1/1 Running 0 51m
ingress-traefik traefik-ingress-controller-xz4jf 1/1 Running 0 51m
jet-system jet-655d48d7bf-bjrrd 1/1 Running 0 6m16s
ui-system spectro-ui-54c9f46dc9-bxtns 1/1 Running 0 8m9s
Create DNS Records
Palette requires two DNS records that point to the load balancer endpoint. You need both a wildcard record for organization subdomains (tenants) and a root domain record for the system console. The following steps use Amazon Route 53 as an example.
-
Get the load balancer endpoint.
kubectl get service traefik-ingress-controller \
--namespace ingress-traefik \
--output jsonpath='{.status.loadBalancer.ingress[0].hostname}'Example outputad909b53ea79047609ac980dd18f6775-35422349.us-east-1.elb.amazonaws.com -
Finding the hosted zone ID of
<your-root-domain>.aws route53 list-hosted-zones \
--query "HostedZones[?Name=='<your-root-domain>.']" \
--output jsonThe hosted zone ID follows
/hostedzone/in theIdfield. In the following example, the hosted zone ID isABCD1234EFGH5678.Example output[
{
"Id": "/hostedzone/ABCD1234EFGH5678",
"Name": "docs-test.spectrocloud.com.",
"CallerReference": "b71eaaab-e387-49d9-97e1-d3a47e9515eb",
"Config": {
"Comment": "A domain for the docs and education team to conduct testing",
"PrivateZone": false
},
"ResourceRecordSetCount": 3
}
] -
Create the wildcard CNAME record for organization subdomains using the
<hosted-zone-id>,<your-root-domain>, and<load-balancer-hostname>.aws route53 change-resource-record-sets \
--hosted-zone-id <hosted-zone-id> \
--change-batch '{
"Changes": [{
"Action": "UPSERT",
"ResourceRecordSet": {
"Name": "*.<your-root-domain>",
"Type": "CNAME",
"TTL": 300,
"ResourceRecords": [{"Value": "<load-balancer-hostname>"}]
}
}]
}'Example output{
"ChangeInfo": {
"Id": "/change/123456789ABCDEFGHIJK",
"Status": "PENDING",
"SubmittedAt": "2026-03-18T19:26:58.547000+00:00"
}
} -
Next, create the root domain record for the zone apex (the same as the hosted zone name) or a subdomain within a larger zone.
- Zone Apex (A Alias)
- Subdomain (CNAME)
If
<your-root-domain>is the zone apex (for example,palette.example.comin a hosted zone forpalette.example.com), you cannot use a CNAME record, as DNS does not allow CNAMEs at the zone apex. Instead, create a Route 53 A (Alias) record that points directly to the load balancer.First, find the canonical hosted zone ID of the load balancer.
aws elbv2 describe-load-balancers \
--query "LoadBalancers[?DNSName=='<load-balancer-hostname>'].[CanonicalHostedZoneId]" \
--output textExample outputHBDJNJ19D0277DThen create the Alias record.
aws route53 change-resource-record-sets \
--hosted-zone-id <hosted-zone-id> \
--change-batch '{
"Changes": [{
"Action": "UPSERT",
"ResourceRecordSet": {
"Name": "<your-root-domain>",
"Type": "A",
"AliasTarget": {
"HostedZoneId": "<load-balancer-hosted-zone-id>",
"DNSName": "<load-balancer-hostname>",
"EvaluateTargetHealth": false
}
}
}]
}'Example output{
"ChangeInfo": {
"Id": "/change/C0669446V5KE9GEIJFGW",
"Status": "PENDING",
"SubmittedAt": "2026-03-18T19:47:53.855000+00:00"
}
}If
<your-root-domain>is not the zone apex (for example,palette.example.comin a hosted zone forexample.com), you can use a standard CNAME record.aws route53 change-resource-record-sets \
--hosted-zone-id <hosted-zone-id> \
--change-batch '{
"Changes": [{
"Action": "UPSERT",
"ResourceRecordSet": {
"Name": "<your-root-domain>",
"Type": "CNAME",
"TTL": 300,
"ResourceRecords": [{"Value": "<load-balancer-hostname>"}]
}
}]
}'Example output{
"ChangeInfo": {
"Id": "/change/C08234719ABCDEF12345",
"Status": "PENDING",
"SubmittedAt": "2026-03-18T19:50:12.123000+00:00"
}
}
Validate
After creating your DNS records, wait a few minutes for the records to propagate to the DNS server. Then open a browser
and navigate to the Palette system console at https://<your-root-domain>/system, and log in with the default username
(admin) and password (admin). You must change your password upon logging in for the first time.
Next Steps
Once you have Palette installed, we recommend taking the following actions:
-
Activate Installation - Required within 30 days. Refer to Activate Palette for guidance.
-
Provide Trusted SSL Certificate - By default, self-hosted Palette uses a self-signed SSL certificate. If you are using a custom SSL certificate, you can add your SSL certificate files through the system console.
-
Set Up Tenants - Tenants are isolated environments in Palette that contain their own clusters, users, and resources. You must set up a tenant before you can deploy workload clusters.
-
Configure an AWS Cloud Account - Add your AWS account to Palette to deploy AWS IaaS and EKS clusters through your AWS environment. EKS clusters support using EKS Pod Identity for authentication in addition to IAM users and dynamic Security Token Service (STS) credentials.
-
Launch a Workload Cluster - Refer to Create and Manage AWS EKS Cluster for guidance on deploying EKS clusters in Palette. We recommend launching a two-node EKS workload cluster so that Pod Disruption Budgets are respected during Kubernetes upgrades, as Palette performs rolling upgrades. Refer to the Amazon EKS managed node group update documentation for more information.