install-odf
Overview
Installs OpenShift Data Foundation (ODF) — the Ceph-based storage solution for multi-node
OpenShift clusters. ODF provides block (RBD), file (CephFS), and object (Noobaa/S3) storage.
For single-node clusters, use install-odf-lvm (ODF-LVM) instead.
OCP 5.0 note: ODF on OCP 5.0 requires RHCOS 10 and x86-64-v3+ CPU microarchitecture.
Prerequisites
- ArgoCD bootstrapped: [[bootstrap-gitops]] completed
- At least 3 worker nodes with dedicated storage disks (for Ceph)
- Nodes labeled for ODF storage:
cluster.ocs.openshift.io/openshift-storage="" - For OCP 5.0: nodes must meet x86-64-v3+ CPU requirement
Quick Diagnostics
# Check ODF operator status
oc get subscription odf-operator -n openshift-storage 2>/dev/null || echo "ODF not yet installed"
oc get csv -n openshift-storage | grep odf
# Check storage nodes
oc get nodes -l cluster.ocs.openshift.io/openshift-storage
# Check StorageCluster
oc get storagecluster -n openshift-storage
# Check ODF pods
oc get pods -n openshift-storage | grep -v Running | grep -v Completed
# Check storage classes created by ODF
oc get storageclass | grep -E "ceph|ocs|odf|noobaa"
Steps
Step 1: Select the right overlays for your OCP version and platform
Operator channel (match your OCP version):
| OCP Version | Operator Overlay |
|---|---|
| 4.9 | operator/overlays/stable-4.9 |
| 4.10–4.22 | operator/overlays/stable-4.X (replace X with your version) |
| 5.0 | operator/overlays/stable-5.0 |
Current recommended: stable-4.22 (OCP 4.22, latest 4.x as of Nov 2026)
StorageCluster instance (match your platform):
| Platform | Instance Overlay | Storage Class |
|---|---|---|
| AWS | instance/overlays/aws |
gp2 |
| GCP | instance/overlays/gcp |
standard-csi |
| Azure | instance/overlays/azure |
managed-premium |
| VMware vSphere | instance/overlays/vsphere |
vSphere datastore |
| Bare-metal / Equinix | instance/overlays/bare-metal |
local-storage |
| KVM / libvirt | instance/overlays/kvm |
local-storage |
| SNO | Use ODF-LVM (odf-lvm-operator) instead — ODF Ceph needs 3 nodes |
Step 2: Apply the ODF operator via ArgoCD (or directly)
Via ArgoCD Application (recommended — add to your profile):
apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
name: openshift-data-foundation-operator
namespace: openshift-gitops
annotations:
argocd.argoproj.io/sync-wave: "1"
spec:
destination:
namespace: openshift-storage
server: https://kubernetes.default.svc
project: default
source:
path: gitops/cluster-config/openshift-data-foundation-operator/operator/overlays/stable-4.19
repoURL: https://github.com/tosin2013/sno-quickstarts.git
targetRevision: main
syncPolicy:
automated:
prune: true
selfHeal: true
syncOptions:
- CreateNamespace=true
Directly with kustomize:
oc apply -k gitops/cluster-config/openshift-data-foundation-operator/operator/overlays/stable-4.19
Step 3: Wait for operator to be ready
oc rollout status deployment/ocs-operator -n openshift-storage --timeout=300s
oc get csv -n openshift-storage | grep -i odf
Step 4: Label storage nodes
# Label the nodes that will host ODF storage (minimum 3)
oc label node <node1> cluster.ocs.openshift.io/openshift-storage=""
oc label node <node2> cluster.ocs.openshift.io/openshift-storage=""
oc label node <node3> cluster.ocs.openshift.io/openshift-storage=""
Step 5: Apply a StorageCluster instance
Choose the platform-appropriate overlay:
# For bare metal:
oc apply -k gitops/cluster-config/openshift-data-foundation-operator/instance/overlays/bare-metal
# For AWS:
oc apply -k gitops/cluster-config/openshift-data-foundation-operator/instance/overlays/aws
# For VMware vSphere:
oc apply -k gitops/cluster-config/openshift-data-foundation-operator/instance/overlays/vsphere
Step 6: Enable the ODF console plugin
# The base includes a job that enables the console plugin
# Verify it ran:
oc get jobs -n openshift-storage | grep console
Verification
# StorageCluster should be in Ready phase
oc get storagecluster -n openshift-storage -o jsonpath='{.items[0].status.phase}'
# Expected: Ready
# Check all ODF pods are Running
oc get pods -n openshift-storage | grep -v Running | grep -v Completed
# Verify storage classes
oc get storageclass
# Expected: ocs-storagecluster-ceph-rbd, ocs-storagecluster-cephfs, openshift-storage.noobaa.io
Common Failure Modes
| Symptom | Root Cause | Fix |
|---|---|---|
StorageCluster stuck in Progressing |
Insufficient storage nodes or no raw disks | Ensure 3 nodes labeled, each with an unformatted disk |
Error: LocalVolume not found |
Using bare-metal overlay without local-storage operator | Install install-local-storage skill first |
OSD pods CrashLoopBackOff |
Disks already contain data/partitions | Wipe disks: sgdisk --zap-all /dev/<disk> on each node |
CSV shows Failed |
ODF version incompatible with OCP version | Match overlay version to oc version output |
| Noobaa pod pending | Insufficient memory on nodes | Noobaa requires ~8Gi; check node resources |
OCP 5.0: pod fails with Illegal instruction |
Node CPU doesn't meet x86-64-v3+ requirement | Check CPU: grep -m1 avx512 /proc/cpuinfo |
Related Skills
- [[bootstrap-gitops]] — Must be completed first
- [[install-virtualization]] — Can run in parallel with ODF after operator is Ready
- [[troubleshoot-argocd]] — Use if ArgoCD fails to sync ODF resources