IMPORTANT NOTE: This site is not official Red Hat documentation and is provided for informational purposes only. These guides may be experimental, proof of concept, or early adoption. Officially supported documentation is available at docs.openshift.com and access.redhat.com.

OpenShift Data Foundation with ARO

Kevin Collins

06/28/2022

Note: This guide demonstrates how to setup and configure self-managed OpenShift Data Foundation in Internal Mode on an ARO Cluster and test it out.

Prerequisites

  • An Azure Red Hat OpenShift cluster ( verion 4.10+ )
  • kubectl cli
  • oc cli
  • moreutils (sponge)
  • jq

Install compute nodes for ODF

A best practice for optimal performance is to run ODF on dedicated nodes with a minimum of one per zone. In this guide, we will be provisioning 3 additional compute nodes, one per zone. Run the following script to create the additional nodes:

  1. Log into your ARO Cluster

  2. Create the new compute nodes

     for ZONE in 1 2 3
   do
     item=$((ZONE-1))
     MACHINESET=$(oc get machineset -n openshift-machine-api -o=jsonpath="{.items[$item]}" | jq -r '[.metadata.name] | @tsv')
     oc get machineset -n openshift-machine-api $MACHINESET -o json > default_machineset$ZONE.json
     worker=odf-worker-$ZONE
     jq ".metadata.name = \"$worker\"" default_machineset$ZONE.json | sponge default_machineset$ZONE.json
     jq '.spec.replicas = 1' default_machineset$ZONE.json| sponge default_machineset$ZONE.json
     jq ".spec.selector.matchLabels.\"machine.openshift.io/cluster-api-machineset\" = \"$worker\"" default_machineset$ZONE.json| sponge default_machineset$ZONE.json
     jq ".spec.template.metadata.labels.\"machine.openshift.io/cluster-api-machineset\" = \"$worker\"" default_machineset$ZONE.json| sponge default_machineset$ZONE.json
     jq '.spec.template.spec.providerSpec.value.vmSize = "Standard_D16s_v3"' default_machineset$ZONE.json | sponge default_machineset$ZONE.json
     jq ".spec.template.spec.providerSpec.value.zone = \"$ZONE\"" default_machineset$ZONE.json | sponge default_machineset$ZONE.json
     jq 'del(.status)' default_machineset$ZONE.json | sponge default_machineset$ZONE.json
     oc create -f default_machineset$ZONE.json
 done

  3. Label new compute nodes

    It takes just a couple of minutes for new nodes to appear.

    Check if the nodes are ready:

    oc get nodes | grep odf-worker

    expected output:

    odf-worker-1-jg7db                                  Ready    worker   19h     v1.23.5+3afdacb
odf-worker-2-ktvct                                  Ready    worker   19h     v1.23.5+3afdacb
odf-worker-3-rk22b                                  Ready    worker   19h     v1.23.5+3afdacb

    Once you see the three nodes, the next step we need to do is label and taint the nodes. This will ensure the OpenShift Data Foundation is installed on these nodes and no other workload will be placed on the nodes.

    for worker in $(oc get nodes | grep odf-worker | awk '{print $1}')
do
  oc label node $worker cluster.ocs.openshift.io/openshift-storage=``
  oc adm taint nodes $worker node.ocs.openshift.io/storage=true:NoSchedule
done

Deploy OpenShift Data Foundation

Next, we will install OpenShift Data Foundation via an Operator.

  1. Create the openshift-storage namespace
    cat <<EOF | oc apply -f -
apiVersion: v1
kind: Namespace
metadata:
  labels:
    openshift.io/cluster-monitoring: "true"
  name: openshift-storage
spec: {}
EOF
  2. Create the Operator Group for openshift-storage
    cat <<EOF | oc apply -f -
apiVersion: operators.coreos.com/v1
kind: OperatorGroup
metadata:
  name: openshift-storage-operatorgroup
  namespace: openshift-storage
spec:
  targetNamespaces:
  - openshift-storage
EOF
  3. Subscribe to the ocs-operator
    cat <<EOF | oc apply -f -
apiVersion: operators.coreos.com/v1alpha1
kind: Subscription
metadata:
  name: ocs-operator
  namespace: openshift-storage
spec:
  channel: "stable-4.10"  # <-- Channel should be modified depending on the OCS version to be installed. Please ensure to maintain compatibility with OCP version
  installPlanApproval: Automatic
  name: ocs-operator
  source: redhat-operators  # <-- Modify the name of the redhat-operators catalogsource if not default
  sourceNamespace: openshift-marketplace
EOF
  4. Subscribe to the odf-operator
    cat <<EOF | oc apply -f -
apiVersion: operators.coreos.com/v1alpha1
kind: Subscription
metadata:
  name: odf-operator
  namespace: openshift-storage
spec:
  channel: "stable-4.10" # <-- Channel should be modified depending on the OCS version to be installed. Please ensure to maintain compatibility with OCP version
  installPlanApproval: Automatic
  name: odf-operator
  source: redhat-operators  # <-- Modify the name of the redhat-operators catalogsource if not default
  sourceNamespace: openshift-marketplace
EOF
  5. Create a Storage Cluster
    cat <<EOF | oc apply -f -
apiVersion: ocs.openshift.io/v1
kind: StorageCluster
metadata:
  annotations:
    uninstall.ocs.openshift.io/cleanup-policy: delete
    uninstall.ocs.openshift.io/mode: graceful
  generation: 2
  name: ocs-storagecluster
  namespace: openshift-storage
spec:
  storageDeviceSets:
  - config: {}
    count: 1
    dataPVCTemplate:
      spec:
        accessModes:
        - ReadWriteOnce
        resources:
          requests:
            storage: 2Ti
        storageClassName: managed-premium
        volumeMode: Block
    name: ocs-deviceset-managed-premium
    portable: true
    replica: 3
  version: 4.10.0
EOF

Validate the install

  1. List the cluster service version for the ODF operators

    oc get csv -n openshift-storage

    verify that the operators below have succeeded.

    NAME                  DISPLAY                       VERSION   REPLACES   PHASE
mcg-operator.v4.10.4   NooBaa Operator               4.10.4                Succeeded
ocs-operator.v4.10.4   OpenShift Container Storage   4.10.4                Succeeded
odf-operator.v4.10.4   OpenShift Data Foundation     4.10.4                Succeeded

  2. Check that the ocs storage classes have been created

    note: this can take around 5 minutes

    oc get sc

    NAME                          PROVISIONER                             RECLAIMPOLICY   VOLUMEBINDINGMODE      ALLOWVOLUMEEXPANSION   AGE
managed-csi                   disk.csi.azure.com                      Delete          WaitForFirstConsumer   true                   118m
managed-premium (default)     kubernetes.io/azure-disk                Delete          WaitForFirstConsumer   true                   119m
ocs-storagecluster-ceph-rbd   openshift-storage.rbd.csi.ceph.com      Delete          Immediate              true                   7s
ocs-storagecluster-cephfs     openshift-storage.cephfs.csi.ceph.com   Delete          Immediate              true                   7s

Test it out

To test out ODF, we will create ‘writer’ pods on each node across all zones and then a reader pod to read the data that is written. This will prove both regional storage along with “read write many” mode is working correctly.

  1. Create a new project

    oc new-project odf-demo

  2. Create a RWX Persistent Volume Claim for ODF

    cat <<EOF | kubectl apply -f -
kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: standard
spec:
  accessModes:
    - ReadWriteMany
  resources:
    requests:
      storage: 400Gi
  storageClassName: ocs-storagecluster-cephfs
EOF

  3. Create writer pods via a DaemonSet Using a deamonset will ensure that we have a ‘writer pod’ on each worker node and will also prove that we correctly set a taint on the ‘ODF Workers’ where which we do not want workload to be added to.

    The writer pods will write out which worker node the pod is running on, the data, and a hello message.

    cat <<EOF | oc apply -f -
apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: test-odf
  labels:
    app: test-odf
spec:
  selector:
    matchLabels:
      name: test-odf
  template:
    metadata:
      labels:
        name: test-odf
    spec:
      containers:
        - name: azodf
          image: centos:latest
          command: ["sh", "-c"]
          resources:
            limits:
              cpu: 1
              memory: "1Gi"
          args:
            [
              "while true; do printenv --null NODE_NAME | tee -a /mnt/odf-data/verify-odf; echo ' says hello '$(date) | tee -a /mnt/odf-data/verify-odf; sleep 15; done;",
            ]
          volumeMounts:
            - name: odf-vol
              mountPath: "/mnt/odf-data"
          env:
            - name: NODE_NAME
              valueFrom:
                fieldRef:
                  fieldPath: spec.nodeName
      volumes:
        - name: odf-vol
          persistentVolumeClaim:
            claimName: standard
EOF

  4. Check the writer pods are running.

    note: there should be 1 pod per non-ODF worker node

    oc get pods

    expected output

    NAME             READY   STATUS    RESTARTS   AGE    IP            NODE                                   NOMINATED NODE   READINESS GATES
test-odf-47p2g   1/1     Running   0          107s   10.128.2.15   aro-kmobb-7zff2-worker-eastus1-xgksq   <none>           <none>
test-odf-p5xk6   1/1     Running   0          107s   10.131.0.18   aro-kmobb-7zff2-worker-eastus3-h4gv7   <none>           <none>
test-odf-ss8b5   1/1     Running   0          107s   10.129.2.32   aro-kmobb-7zff2-worker-eastus2-sbfpm   <none>           <none>

  5. Create a reader pod The reader pod will simply log data written by the writer pods.

    cat <<EOF | oc apply -f -
apiVersion: v1
kind: Pod
metadata:
  name: test-odf-read
spec:
  containers:
    - name: test-odf-read
      image: centos:latest
      command: ["/bin/bash", "-c", "--"]
      resources:
        limits:
          cpu: 1
          memory: "1Gi"
      args: ["tail -f /mnt/odf-data/verify-odf"]
      volumeMounts:
        - name: odf-vol
          mountPath: "/mnt/odf-data"
  volumes:
    - name: odf-vol
      persistentVolumeClaim:
        claimName: standard
EOF

  6. Now let’s verify the POD is reading from shared volume.

    oc logs test-odf-read
aro-kmobb-7zff2-worker-eastus1-xgksq says hello Wed Jun 29 10:41:06 EDT 2022
aro-kmobb-7zff2-worker-eastus3-h4gv7 says hello Wed Jun 29 10:41:06 EDT 2022
aro-kmobb-7zff2-worker-eastus2-sbfpm says hello Wed Jun 29 10:41:06 EDT 2022
aro-kmobb-7zff2-worker-eastus1-xgksq says hello Wed Jun 29 10:41:06 EDT 2022
aro-kmobb-7zff2-worker-eastus3-h4gv7 says hello Wed Jun 29 10:41:06 EDT 2022
aro-kmobb-7zff2-worker-eastus2-sbfpm says hello Wed Jun 29 10:41:06 EDT 2022
aro-kmobb-7zff2-worker-eastus1-xgksq says hello Wed Jun 29 10:41:06 EDT 2022
aro-kmobb-7zff2-worker-eastus3-h4gv7 says hello Wed Jun 29 10:41:06 EDT 2022

    Notice that pods in different zones are writing to the PVC which is managed by ODF.