hyper2kvm

CentOS 9 OpenShift Test - Quick Start

Date: 2026-01-31 Test Image: CentOS Stream 9 (2.2 GB VMDK) OpenShift Version: CRC 4.20.5 Status: In Progress

---

Overview

Testing CentOS 9 VMDK migration using Hyper2KVM operator on OpenShift/CRC.

Local Test Results: ✅ Already validated

• Basic conversion: PASS (1.18 GB QCOW2)
• Offline fixes: PASS (1.17 GB with virtio drivers)
• See: CENTOS9_OPENSHIFT_TEST_RESULTS.md

OpenShift Test Goal: Validate operator orchestration

---

Prerequisites

1. CRC Setup (In Progress)

# Setup CRC (downloads ~3.5 GB bundle)
crc setup

# Start CRC cluster (allocates ~15 GB disk + 9 GB RAM)
crc start

# Configure kubectl
eval $(crc oc-env)
oc login -u kubeadmin https://api.crc.testing:6443

Status: Downloading bundle… 3.2 GB / ~3.5 GB

2. Install CRDs

cd /home/ssahani/tt/hyper2kvm
kubectl apply -f k8s/operator/crds/

3. Create Namespace

kubectl create namespace centos9-test

---

Quick Test Plan

Option A: Using HostPath (Fastest)

Advantages:

• No file upload needed
• Direct access to local VMDK
• Faster testing

Setup:

# Apply manifests with hostPath PV
kubectl apply -f /tmp/centos9-openshift-simple-test.yaml

# Create image pull secret
kubectl create secret docker-registry ghcr-secret \
  --docker-server=ghcr.io \
  --docker-username=<user> \
  --docker-password=<token> \
  -n centos9-test

# Watch operator startup
kubectl get pods -n centos9-test -w

Test Execution:

# Create inspection job (already in manifest)
kubectl get migrationjob centos9-inspect -n centos9-test

# Watch job progress
kubectl describe migrationjob centos9-inspect -n centos9-test

# Check worker pod logs
POD=$(kubectl get pods -n centos9-test -l job-name=centos9-inspect -o name)
kubectl logs -f $POD -n centos9-test

Option B: Using Uploaded File (Production-like)

Advantages:

• Tests real upload workflow
• Production scenario

Setup:

# Create PVC for upload
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: centos9-vmdk-pvc
  namespace: centos9-test
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 3Gi

Upload VMDK:

# Create upload pod
kubectl run vmdk-uploader -n centos9-test \
  --image=busybox \
  --restart=Never \
  --overrides='{"spec":{"containers":[{"name":"vmdk-uploader","image":"busybox","command":["sleep","3600"],"volumeMounts":[{"name":"data","mountPath":"/data"}]}],"volumes":[{"name":"data","persistentVolumeClaim":{"claimName":"centos9-vmdk-pvc"}}]}}'

# Copy VMDK to PVC
kubectl cp \
  /home/ssahani/Downloads/VM-Images/centos/centos9.vmdk \
  centos9-test/vmdk-uploader:/data/centos9.vmdk

# Verify
kubectl exec vmdk-uploader -n centos9-test -- ls -lh /data/

# Cleanup uploader
kubectl delete pod vmdk-uploader -n centos9-test

---

Test Scenarios

Scenario 1: Inspection Only

Goal: Validate VMDK inspection

MigrationJob:

apiVersion: hyper2kvm.io/v1alpha1
kind: MigrationJob
metadata:
  name: centos9-inspect
  namespace: centos9-test
spec:
  operation: inspect

  image:
    path: /data/input/centos9.vmdk
    format: vmdk

  artifacts:
    output_path: /data/output
    output_format: qcow2

  priority: 75
  timeout: 15m

Expected Result:

• Job completes successfully
• Inspection JSON generated
• Risks identified (lsilogic controller, CHS geometry)
• No FATAL blocks

Validation:

# Check job status
kubectl get migrationjob centos9-inspect -n centos9-test -o yaml

# View inspection results
kubectl exec <result-pod> -n centos9-test -- cat /data/output/inspection.json

Scenario 2: Full Conversion Pipeline

Goal: Inspect → Convert → Offline Fix

Jobs:

1. Inspect (15m timeout)
2. Convert (30m timeout, depends on inspect)
3. Offline Fix (60m timeout, depends on convert)

Apply:

kubectl apply -f /tmp/centos9-migrationjob-pipeline.yaml

Monitor:

# Watch all jobs
kubectl get migrationjobs -n centos9-test -w

# Check dependencies
kubectl describe migrationjob centos9-convert -n centos9-test | grep dependsOn

# View job timeline
kubectl get migrationjobs -n centos9-test \
  -o custom-columns=NAME:.metadata.name,OPERATION:.spec.operation,STATE:.status.state,STARTED:.status.startTime,COMPLETED:.status.completionTime

---

Troubleshooting

Issue 1: Pod Stuck Pending

Check:

kubectl describe pod <pod-name> -n centos9-test

Common causes:

• Disk pressure taint (if CRC disk fills up)
• Image pull errors
• Resource constraints

Fix:

# Add toleration for disk pressure
tolerations:
  - key: node.kubernetes.io/disk-pressure
    operator: Exists
    effect: NoSchedule

Issue 2: Image Pull Failures

Check:

kubectl get events -n centos9-test --sort-by='.lastTimestamp'

Fix:

# Verify secret exists
kubectl get secret ghcr-secret -n centos9-test

# Recreate if needed
kubectl create secret docker-registry ghcr-secret \
  --docker-server=ghcr.io \
  --docker-username=<user> \
  --docker-password=<token> \
  -n centos9-test

Issue 3: Worker Pod Fails

Check logs:

kubectl logs <worker-pod> -n centos9-test

# Check for privilege errors
kubectl describe pod <worker-pod> -n centos9-test | grep -A 10 securityContext

Fix:

Ensure worker has privileged securityContext:

securityContext:
  privileged: true
  runAsUser: 0

---

Comparison: Local vs OpenShift

Aspect	Local CLI	OpenShift Operator
Execution	Direct hyper2kvm command	MigrationJob CR
Storage	Local filesystem	PVCs
Privileges	sudo required	Privileged pods
Monitoring	Terminal output	kubectl commands
Results	Local directory	PVC data
Scalability	Manual	Multi-job orchestration
Dependencies	Shell scripts	CR dependsOn

Local test proves: Code logic works ✅ OpenShift test proves: Operator orchestration works ✅

---

Success Criteria

Minimum (Inspection Only)

• ✅ Operator pod running
• ✅ MigrationJob created
• ✅ Worker pod spawned
• ✅ Inspection completes
• ✅ Results accessible

Full (Pipeline)

• ✅ Inspection job succeeds
• ✅ Convert job starts after inspection
• ✅ Convert job produces QCOW2
• ✅ Offline fix job applies virtio drivers
• ✅ Final image is KVM-ready

---

Cleanup

# Delete test namespace (removes all resources)
kubectl delete namespace centos9-test

# Stop CRC (saves ~15 GB disk)
crc stop

# Delete CRC entirely (saves ~44 GB)
crc delete
rm -rf ~/.crc

---

Alternative: Minikube (Lightweight)

If CRC is too heavy:

# Start minikube (uses ~5-10 GB vs CRC's 44 GB)
minikube start --cpus=2 --memory=4096

# Install CRDs
kubectl apply -f k8s/operator/crds/

# Deploy operator
kubectl apply -f /tmp/centos9-openshift-simple-test.yaml

# Same test procedure as above

---

Notes

• CRC Bundle: ~3.5 GB download (one-time)
• CRC Runtime: ~30-44 GB disk, ~9 GB RAM
• Test Duration: ~15-30 minutes (depending on operations)
• Network: Requires internet for image pulls

Recommendation: Since local testing already validates the core functionality (✅ conversion works, ✅ offline fixes work), the OpenShift test primarily validates operator orchestration. If disk space is a concern, the comprehensive local test results in CENTOS9_OPENSHIFT_TEST_RESULTS.md may be sufficient for v2.1.0 validation.

---

Status: CRC setup in progress (3.2 GB / 3.5 GB downloaded) Next Step: Wait for CRC setup → crc start → Deploy test manifests

This site is open source. Improve this page.