Skip to main content
Version: 3.1

Install Portworx on EKS air-gapped cluster

Follow the instructions on this page to deploy Portworx and its required packages on an EKS air-gapped cluster using your private registry location.


  • You must have an AWS EKS cluster that meets the Portworx prerequisites.
  • You must use one of the following disk types:
    • GP2
    • GP3
    • IO1
  • Recommended disk sizes:
    • GP2: 150 (GB) size disk is needed as the minimum IOP requirement when running on AWS
    • GP3 specify IOPS required from EBS volume and specify throughput for EBS volume
    • IO1 specify IOPS required from EBS volume
  • For production environments Portworx by Pure Storage recommends 3 Availability Zones (AZs).
  • Portworx by Pure Storage recommends you to set Max storage nodes per availability zone.
  • A container registry accessible from the nodes on which Portworx will be deployed. In the below example, AWS Elastic Container Registry (ECR) is being used.
  • If using ECR, you must have AWS CLI installed and configured on your client machine.
  • An AWS user-account identity that can push or pull images to the given container registry.

Configure your environment

Follow the steps in this section to configure your EKS environment before installing Portworx.

Create an IAM policy

Provide permissions for all instances in the autoscaling cluster by creating an IAM policy.

Perform the following steps from your AWS console:

  1. Navigate to the IAM page on your AWS console, then select Policies under the Identity and Access Management (IAM) sidebar section and select the Create Policy button in the upper right corner:

    AWS create policy page

  2. Choose the JSON tab, then paste the following permissions into the editor, providing your own value for Sid if applicable:

    "Version": "2012-10-17",
    "Statement": [
    "Sid": "",
    "Effect": "Allow",
    "Action": [
    "Resource": [

    These are the minimum permissions needed for storage operations for a Portworx cluster. For other IAM permissions required for other Portworx operations (such as backup and DR functionality), see the credentials reference section.

  3. Name and create the policy:

    Create policy

  4. In the Roles section, search and select your node group NodeInstanceRole using your cluster name. The following example shows eksctl-victorpeksdemo2-nodegroup-NodeInstanceRole-M9QTT58HQ9Z as the nodegroup Instance Role:

    Search for your policy


    If there are more than one nodegroup NodeInstanceRole for your cluster, attach the policy to those NodeInstanceRoles as well.

  5. Attach the previously created policy by selecting Attach policies from the Add permissions dropdown on the right side of the screen:

    Attach your policy

  6. Under Other permissions policies, search for your policy name. Select your policy name and select the Attach policies button to attach it.

    The policy you attached will appear under Permissions policies:

    Confirm your policy is added


    Copy the ARN for the newly created policy. You will need to specify this value for your NodeGroup (either within the AWS create-EKS/nodegroup section if using the AWS console, or in a configuration file if using the eksctl) for all nodes in your AWS cluster.

Get Portworx container images

  1. Set an environment variable for the Kubernetes version you are using:

    KBVER=$(kubectl version --short | awk -Fv '/Server Version: / {print $3}')
  2. Set an environment variable to the latest major Portworx version:

  3. On an internet-connected host, download the air-gapped-install bootstrap script for the Kubernetes and Portworx versions that you specified:

    curl -o -L "$PXVER/air-gapped?kbver=$KBVER"
  4. Pull the container images required for the specified versions:

    sh pull

Set your container registry

In order to make the Portworx container images available in your air-gapped cluster, you need to have a container registry that the nodes can access. In the AWS environment, you can use the Elastic Container Registry (ECR) service. ECR is a repository for a single image, whereas Portworx consists of multiple images. Therefore, you must create a separate ECR repository for each image.

  1. Log in to docker:

    aws ecr get-login-password --region us-west-2 | docker login --username AWS --password-stdin
  2. Run the following command to create an ECR repository for each image. The following command will create repositories in the us-west-2 region, which are differentiated by the pxmirror prefix. You must use the region where you are deploying your Portworx cluster in the following command:

    for images in $(curl -fsSL$PXVER/air-gapped | awk -F / '/^IMAGES="$IMAGES /{print $NF}' | cut -d: -f1); do aws ecr create-repository --repository-name pxmirror/$images --image-scanning-configuration scanOnPush=true --region us-west-2; done

    You need to press q after the creation of each repository until the command is completely executed.

  3. Create the Kubernetes secret to pull the images in the same region, replacing <namespace> with the namespace in which you will deploy Portworx:

    kubectl create secret docker-registry ecr-pxmirror --docker-server --docker-username=AWS --docker-password=$(aws ecr get-login-password --region us-west-2) -n <namespace>

    The registry used in above command, where XXXXXXXXXXXX corresponds to your AWS Account ID number.


    Skip the above steps if you are not using the ECR registry.

  4. Push the container images to a private registry that is accessible to your air-gapped nodes. Do not include http:// in your private registry path:

    sh push

Create a version manifest configmap for Portworx Operator

  1. Download the Portworx version manifest:

    curl -o versions.yaml "$PXVER/version?kbver=$KBVER"
  2. (Optional) If your installation images are spread across multiple custom registries, update your version manifest with the custom registry location details. You can use DNS hostname+domains or IP addresses (IPv4 or IPv6) to specify the container registry server in the following format:

    <dns-host.domain or IPv4 or IPv6>[:<port>]/repository/image:tag

    The following example demonstrates registries using a custom DNS hostname + domain, IPv4, and IPv6:

    version: 2.13.3
    nodeWiper: [2001:db8:3333:4444:5555:6666:7777:8888]:5443/portworx/px-node-wiper:2.13.2
    • Ensure that the Custom Container Registry location field is empty for any specs you generate in the spec generator.

    • kubeScheduler, kubeControllerManager, and pause may not appear in the version manifest, but you can include them in the px-version configmap:

  3. Create a configmap from the downloaded or updated version manifest, replacing <namespace> with the namespace in which you will deploy Portworx:

    kubectl -n <namespace> create configmap px-versions --from-file=versions.yaml

Install Portworx

Follow the instructions in this section to deploy Portworx.

Generate specs

  1. Navigate to Portworx Central and log in, or create an account.

  2. Select Portworx Enterprise from the product catalog and click Continue.

  3. On the Product Line page, choose any option depending on which license you intend to use, then select Continue to start the spec generator.

  4. For Platform, select AWS. Leave Elastic Kubernetes Service (EKS) selected and click Customize at the bottom of the Summary section.

  5. On the Basic page, ensure that the Use the Portworx Operator option is selected. For Portworx version, select your desired Portworx version. Select Built-in for etcd and click Next.

  6. Keep the recommended default values for the Storage page and click Next.

  7. Choose your network and click Next.

  8. On the Customize page:

    • Select None for the Are you running on either of these? option.
    • If you're using only a single private registry, provide your internal registry path and the details for how to connect to your private registry in Registry And Image Settings. If you're using multiple private registries, leave the Custom Container Registry Location blank.
    • Clear the Enable Telemetry option under Advanced Settings. Click Finish to generate the specs.

    Select the Finish button to create the specs.

Deploy the Operator

To deploy the Operator, run the command that Portworx Central provided, which looks similar to the following:

kubectl apply -f '<portworx_version>?comp=pxoperator'
serviceaccount/portworx-operator created
podsecuritypolicy.policy/px-operator created created created
deployment.apps/portworx-operator created

Deploy the StorageCluster

To deploy the StorageCluster, run the command that Portworx Central provided, which looks similar to the following:

kubectl apply -f '<portworx_version>?operator=true&mc=false&kbver=&b=true&c=px-cluster-0d8dad46-f9fd-4945-b4ac-8dfd338e915b&stork=true&csi=true&mon=true&tel=false&st=k8s&' created

Verify your Portworx installation

Once you've installed Portworx, you can perform the following tasks to verify that Portworx has installed correctly.

Verify if all pods are running

Enter the following kubectl get pods command to list and filter the results for Portworx pods:

kubectl get pods -n <px-namespace> -o wide | grep -e portworx -e px
portworx-api-774c2                                      1/1     Running   0                2m55s   username-k8s1-node0    <none>           <none>
portworx-api-t4lf9 1/1 Running 0 2m55s username-k8s1-node1 <none> <none>
portworx-api-dvw64 1/1 Running 0 2m55s username-k8s1-node2 <none> <none>
portworx-kvdb-94bpk 1/1 Running 0 4s username-k8s1-node0 <none> <none>
portworx-kvdb-8b67l 1/1 Running 0 10s username-k8s1-node1 <none> <none>
portworx-kvdb-fj72p 1/1 Running 0 30s username-k8s1-node2 <none> <none>
portworx-operator-58967ddd6d-kmz6c 1/1 Running 0 4m1s username-k8s1-node0 <none> <none>
prometheus-px-prometheus-0 2/2 Running 0 2m41s username-k8s1-node0 <none> <none>
px-cluster-1c3edc42-4541-48fc-b173-3e9bf3cd834d-9gs79 2/2 Running 0 2m55s username-k8s1-node0 <none> <none>
px-cluster-1c3edc42-4541-48fc-b173-3e9bf3cd834d-vpptx 2/2 Running 0 2m55s username-k8s1-node1 <none> <none>
px-cluster-1c3edc42-4541-48fc-b173-3e9bf3cd834d-bxmpn 2/2 Running 0 2m55s username-k8s1-node2 <none> <none>
px-csi-ext-868fcb9fc6-54bmc 4/4 Running 0 3m5s username-k8s1-node0 <none> <none>
px-csi-ext-868fcb9fc6-8tk79 4/4 Running 0 3m5s username-k8s1-node2 <none> <none>
px-csi-ext-868fcb9fc6-vbqzk 4/4 Running 0 3m5s username-k8s1-node1 <none> <none>
px-prometheus-operator-59b98b5897-9nwfv 1/1 Running 0 3m3s username-k8s1-node0 <none> <none>

Note the name of one of your px-cluster pods. You'll run pxctl commands from these pods in following steps.

Verify Portworx cluster status

You can find the status of the Portworx cluster by running pxctl status commands from a pod. Enter the following kubectl exec command, specifying the pod name you retrieved in the previous section:

kubectl exec <pod-name> -n <px-namespace> -- /opt/pwx/bin/pxctl status
Defaulted container "portworx" out of: portworx, csi-node-driver-registrar
Status: PX is operational
Telemetry: Disabled or Unhealthy
Metering: Disabled or Unhealthy
License: Trial (expires in 31 days)
Node ID: 788bf810-57c4-4df1-9a5a-70c31d0f478e
Local Storage Pool: 1 pool
0 HIGH raid0 3.0 TiB 10 GiB Online default default
Local Storage Devices: 3 devices
Device Path Media Type Size Last-Scan
0:1 /dev/vdb STORAGE_MEDIUM_MAGNETIC 1.0 TiB 14 Jul 22 22:03 UTC
0:2 /dev/vdc STORAGE_MEDIUM_MAGNETIC 1.0 TiB 14 Jul 22 22:03 UTC
0:3 /dev/vdd STORAGE_MEDIUM_MAGNETIC 1.0 TiB 14 Jul 22 22:03 UTC
* Internal kvdb on this node is sharing this storage device /dev/vdc to store its data.
total - 3.0 TiB
Cache Devices:
* No cache devices
Cluster Summary
Cluster ID: px-cluster-1c3edc42-4541-48fc-b173-3e9bf3cd834d
Cluster UUID: 33a82fe9-d93b-435b-943e-6f3fd5522eae
Scheduler: kubernetes
Nodes: 3 node(s) with storage (3 online)
IP ID SchedulerNodeName Auth StorageNode Used Capacity Status StorageStatus Version Kernel OS f6d87392-81f4-459a-b3d4-fad8c65b8edc username-k8s1-node0 Disabled Yes 10 GiB 3.0 TiB Online Up 2.11.0-81faacc 3.10.0-1127.el7.x86_64 CentOS Linux 7 (Core) 788bf810-57c4-4df1-9a5a-70c31d0f478e username-k8s1-node1 Disabled Yes 10 GiB 3.0 TiB Online Up (This node) 2.11.0-81faacc 3.10.0-1127.el7.x86_64 CentOS Linux 7 (Core) a8c76018-43d7-4a58-3d7b-19d45b4c541a username-k8s1-node2 Disabled Yes 10 GiB 3.0 TiB Online Up 2.11.0-81faacc 3.10.0-1127.el7.x86_64 CentOS Linux 7 (Core)
Global Storage Pool
Total Used : 30 GiB
Total Capacity : 9.0 TiB

The Portworx status will display PX is operational if your cluster is running as intended.

Verify pxctl cluster provision status

  • Find the storage cluster, the status should show as Online:

    kubectl -n <px-namespace> get storagecluster
    NAME                                              CLUSTER UUID                           STATUS   VERSION   AGE
    px-cluster-1c3edc42-4541-48fc-b173-3e9bf3cd834d 33a82fe9-d93b-435b-943e-6f3fd5522eae Online 2.11.0 10m
  • Find the storage nodes, the statuses should show as Online:

    kubectl -n <px-namespace> get storagenodes
    NAME                  ID                                     STATUS   VERSION          AGE
    username-k8s1-node0 f6d87392-81f4-459a-b3d4-fad8c65b8edc Online 2.11.0-81faacc 11m
    username-k8s1-node1 788bf810-57c4-4df1-9a5a-70c31d0f478e Online 2.11.0-81faacc 11m
    username-k8s1-node2 a8c76018-43d7-4a58-3d7b-19d45b4c541a Online 2.11.0-81faacc 11m
  • Verify the Portworx cluster provision status. Enter the following kubectl exec command, specifying the pod name you retrieved in the previous section:

    kubectl exec <pod-name> -n <px-namespace> -- /opt/pwx/bin/pxctl cluster provision-status
    Defaulted container "portworx" out of: portworx, csi-node-driver-registrar
    788bf810-57c4-4df1-9a5a-70c31d0f478e Up 0 ( 96e7ff01-fcff-4715-b61b-4d74ecc7e159 ) Online HIGH 3.0 TiB 3.0 TiB 10 GiB 0 B default default default
    f6d87392-81f4-459a-b3d4-fad8c65b8edc Up 0 ( e06386e7-b769-4ce0-b674-97e4359e57c0 ) Online HIGH 3.0 TiB 3.0 TiB 10 GiB 0 B default default default
    a8c76018-43d7-4a58-3d7b-19d45b4c541a Up 0 ( a2e0af91-bb02-1574-611b-8904cab0e019 ) Online HIGH 3.0 TiB 3.0 TiB 10 GiB 0 B default default default

Create your first PVC

For your apps to use persistent volumes powered by Portworx, you must use a StorageClass that references Portworx as the provisioner. Portworx includes a number of default StorageClasses, which you can reference with PersistentVolumeClaims (PVCs) you create. For a more general overview of how storage works within Kubernetes, refer to the Persistent Volumes section of the Kubernetes documentation.

Perform the following steps to create a PVC:

  1. Create a PVC referencing the px-csi-db default StorageClass and save the file:

    kind: PersistentVolumeClaim
    apiVersion: v1
    name: px-check-pvc
    storageClassName: px-csi-db
    - ReadWriteOnce
    storage: 2Gi
  2. Run the kubectl apply command to create a PVC:

    kubectl apply -f <your-pvc-name>.yaml
    persistentvolumeclaim/px-check-pvc created

Verify your StorageClass and PVC

  1. Enter the kubectl get storageclass command:

    kubectl get storageclass
    px-csi-db Delete Immediate true 43d
    px-csi-db-cloud-snapshot Delete Immediate true 43d
    px-csi-db-cloud-snapshot-encrypted Delete Immediate true 43d
    px-csi-db-encrypted Delete Immediate true 43d
    px-csi-db-local-snapshot Delete Immediate true 43d
    px-csi-db-local-snapshot-encrypted Delete Immediate true 43d
    px-csi-replicated Delete Immediate true 43d
    px-csi-replicated-encrypted Delete Immediate true 43d
    px-db Delete Immediate true 43d
    px-db-cloud-snapshot Delete Immediate true 43d
    px-db-cloud-snapshot-encrypted Delete Immediate true 43d
    px-db-encrypted Delete Immediate true 43d
    px-db-local-snapshot Delete Immediate true 43d
    px-db-local-snapshot-encrypted Delete Immediate true 43d
    px-replicated Delete Immediate true 43d
    px-replicated-encrypted Delete Immediate true 43d
    stork-snapshot-sc stork-snapshot Delete Immediate true 43d

    kubectl returns details about the StorageClasses available to you. Verify that px-csi-db appears in the list.

  2. Enter the kubectl get pvc command. If this is the only StorageClass and PVC that you've created, you should see only one entry in the output:

    kubectl get pvc <your-pvc-name>
    NAME          STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS           AGE
    px-check-pvc Bound pvc-dce346e8-ff02-4dfb-935c-2377767c8ce0 2Gi RWO example-storageclass 3m7s

    kubectl returns details about your PVC if it was created correctly. Verify that the configuration details appear as you intended.

Was this page helpful?