Planning your Portworx setup
Before you create your cluster and install Portworx, review the following planning steps.
- Decide where you want to store the Portworx metadata. You can use KVDB or an external Database instance. For more information, see Understanding the key-value store. To learn more about what the key value does, see the Portworx documentation.
- Decide whether you want encryption. You can use Hyper Protect Crypto Services or IBM Key Protect. For more information, see Understanding encryption for Portworx.
- Decide whether you want to use journal devices. Journal devices allow Portworx to write logs directly to a local disk on your worker node.
- VPC or Satellite clusters only - Decide whether you want to use cloud drives. Cloud drives allow you to dynamically provision the Portworx volumes. If you don’t want to use cloud drives, you must manually attach volumes to worker nodes.
- Review the Limitations.
Limitations
Review the following Portworx limitations.
Limitation | Description |
---|---|
Classic clusters Pod restart required when adding worker nodes. | Because Portworx runs as a DaemonSet in your cluster, existing worker nodes are automatically inspected for raw block storage and added to the Portworx data layer when you deploy Portworx. If you add or update worker nodes to your cluster and add raw block storage to those workers, restart the Portworx pods on the new or updated worker nodes so that your storage volumes are detected by the DaemonSet. |
VPC clusters Storage volume reattachment required when updating worker nodes. | When you update a worker node in a VPC cluster, the worker node is removed from your cluster and replaced with a new worker node. If Portworx volumes are attached to the worker node that is replaced, you must attach the volumes to the new worker node. You can attach storage volumes with the API or the CLI. Note this limitation does not apply to Portworx deployments that are using cloud drives. |
The Portworx experimental InitializerConfiguration feature is not supported. |
Red Hat OpenShift on IBM Cloud does not support the Portworx experimental InitializerConfiguration admission controller. |
Private clusters | To install Portworx in a cluster that doesn't have VRF or access to private cloud service endpoints (CSEs), you must create a rule in the default security group to allow inbound and outbound traffic for the following IP addresses: 166.9.24.81 ,
166.9.22.100 , and 166.9.20.178 . For more information, see Updating the default security group. |
Portworx Backup | Portworx backup is not supported for Satellite clusters. |
Overview of the Portworx lifecycle
- Create a multizone cluster.
- Infrastructure provider: For Satellite clusters, make sure to add block storage volumes to your hosts before attaching them to your location. If you use classic infrastructure, you must choose a bare metal flavor for the worker nodes. For classic clusters, virtual machines have only 1000 Mbps of networking speed, which is not sufficient to run production workloads with Portworx. Instead, provision Portworx on bare metal machines for the best performance.
- Worker node flavor: Choose an SDS or bare metal flavor. If you want to use virtual machines, use a worker node with 8 vCPU and 8 GB memory or more.
- Minimum number of workers: Two worker nodes per zone across three zones, for a minimum total of six worker nodes.
- VPC and non-SDS classic worker nodes only: Create raw, unformatted, and unmounted block storage.
- For production workloads, create an external Databases for etcd instance for your Portworx metadata key-value store.
- Optional Set up encryption.
- Install Portworx.
- Maintain the lifecycle of your Portworx deployment in your cluster.
- When you update worker nodes in VPC clusters, you must take additional steps to re-attach your Portworx volumes. You can attach your storage volumes by using the API or CLI.
- To remove a Portworx volume, storage node, or the entire Portworx cluster, see Portworx cleanup.
Creating a secret to store the KMS credentials
Before you begin: Set up encryption
- Encode the credentials that you retrieved in the previous section to base64 and note all the base64 encoded values. Repeat this command for each parameter to retrieve the base64 encoded value.
echo -n "<value>" | base64
- Create a project in your cluster called
portworx
.oc create ns portworx
- Create a Kubernetes secret named
px-ibm
in theportworx
project of your cluster to store your IBM Key Protect information.-
Create a configuration file for your Kubernetes secret with the following content.
apiVersion: v1 kind: Secret metadata: name: px-ibm namespace: portworx type: Opaque data: IBM_SERVICE_API_KEY: <base64_apikey> IBM_INSTANCE_ID: <base64_guid> IBM_CUSTOMER_ROOT_KEY: <base64_rootkey> IBM_BASE_URL: <base64_endpoint>
metadata.name
- Enter
px-ibm
as the name for your Kubernetes secret. If you use a different name, Portworx does not recognize the secret during installation. data.IBM_SERVICE_API_KEY
- Enter the base64 encoded IBM Key Protect or Hyper Protect Crypto Services API key that you retrieved earlier.
data.IBM_INSTANCE_ID
- Enter the base64 encoded service instance GUID that you retrieved earlier.
data.IBM_CUSTOMER_ROOT_KEY
- Enter the base64 encoded root key that you retrieved earlier.
data.IBM_BASE_URL
- IBM Key Protect: Enter the base64 encoded API endpoint of your service instance.
- Hyper Protect Crypto Services: Enter the base64 encoded Key Management public endpoint.
-
Create the secret in the
portworx
project of your cluster.oc apply -f secret.yaml
-
Verify that the secret is created successfully.
oc get secrets -n portworx
-
- If you set up encryption before your installed Portworx, you can now install Portworx in your cluster. To add encryption to your cluster after you installed Portworx, update
the Portworx DaemonSet to add
"-secret_type"
and"ibm-kp"
as additional options to the Portworx container definition.-
Update the Portworx DaemonSet.
oc edit daemonset portworx -n kube-system
Example updated DaemonSet
containers: - args: - -c - testclusterid - -s - /dev/sdb - -x - kubernetes - -secret_type - ibm-kp name: portworx
After you edit the DaemonSet, the Portworx pods are restarted and automatically update the
config.json
file on the worker node to reflect that change. -
List the Portworx pods in your
kube-system
project.oc get pods -n kube-system | grep portworx
-
Log in to one of your Portworx pods.
oc exec -it <pod_name> -it -n kube-system
-
Navigate in to the
pwx
directory.cd etc/pwx
-
Review the
config.json
file to verify that"secret_type": "ibm-kp"
is added to the secret section of your CLI output.cat config.json
Example output
{ "alertingurl": "", "clusterid": "px-kp-test", "dataiface": "", "kvdb": [ "etcd:https://portal-ssl748-34.bmix-dal-yp-12a2312v5-123a-44ac-b8f7-5d8ce1d123456.123456789.composedb.com:56963", "etcd:https://portal-ssl735-35.bmix-dal-yp-12a2312v5-123a-44ac-b8f7-5d8ce1d123456.12345678.composedb.com:56963" ], "mgtiface": "", "password": "ABCDEFGHIJK", "scheduler": "kubernetes", "secret": { "cluster_secret_key": "", "secret_type": "ibm-kp" }, "storage": { "devices": [ "/dev/sdc1" ], "journal_dev": "", "max_storage_nodes_per_zone": 0, "system_metadata_dev": "" }, "username": "root", "version": "1.0" }
-
Exit the pod.
-
Check out how to encrypt the secrets in your cluster, including the secret where you stored your Key Protect CRK for your Portworx storage cluster.