This repository provides automation for deploying and managing two-node OpenShift clusters for development and testing. It supports "Two-Node with Arbiter" (TNA) and "Two-Node with Fencing" (TNF) topologies using either dev-scripts or kcli deployment methods.

If you have AWS access, use the automated workflow. Most lifecycle operations can be performed from the deploy folder using make:

cd deploy/

# Create AWS hypervisor and deploy cluster in one command
make deploy arbiter-ipi    # Two-Node with Arbiter (IPI method)
make deploy arbiter-agent  # Two-Node with Arbiter (Agent method)
make deploy fencing-ipi    # Two-Node with Fencing (IPI method)

# Other useful commands
make ssh                   # SSH into hypervisor
make info                  # Display instance information
make clean                 # Clean OpenShift cluster
make get-tnf-logs          # Collect cluster logs from VMs
make patch-nodes           # Build and patch resource-agents RPM
make help                  # Show all available commands

See deploy/README.md for complete command reference and deploy/aws-hypervisor/README.md for AWS setup instructions.

If you have an existing RHEL 9 server, initialize it and deploy a cluster:

cd deploy/openshift-clusters/

# One-time host initialization (configures RHEL, subscriptions, packages)
cp inventory.ini.sample inventory.ini
# Edit inventory.ini with your server details
ansible-playbook init-host.yml -i inventory.ini

# Deploy OpenShift cluster (choose one method)
ansible-playbook setup.yml -i inventory.ini        # dev-scripts (arbiter or fencing)
ansible-playbook kcli-install.yml -i inventory.ini # kcli (fencing only)

See deploy/openshift-clusters/README-external-host.md for detailed instructions.

dev-scripts: Traditional method supporting both arbiter and fencing topologies with IPI and Agent-based installation options.

• Documentation: deploy/openshift-clusters/README.md

kcli: Modern method with simplified VM management, currently supports fencing topology.

• Documentation: deploy/openshift-clusters/README-kcli.md

Two-Node with Arbiter (TNA): Two master nodes with a separate arbiter node for quorum. See docs/arbiter/README.md

Two-Node with Fencing (TNF): Two master nodes with BMC-based fencing for automated node recovery. See docs/fencing/README.md

If you're using Claude Code, use the /setup command to get interactive help configuring the repository for first-time use. It will guide you through:

• Copying configuration files from templates
• Setting up credentials and authentication
• Installing required dependencies
• Validating your setup

Run /setup to begin, or /setup <method> to configure a specific deployment method (aws, external, kcli, dev-scripts).

The helpers/ directory contains utilities for cluster operations including resource-agents patching, fencing validation, and containerized build validation. To quickly verify a resource-agents branch compiles on CentOS Stream 9 and 10:

make test-resource-agents                          # prompts for repo and ref
make test-resource-agents ARGS="--ref my-branch"   # skip prompts

To test the built RPM on a live cluster, extract it from the Stream 9 image and patch your nodes:

# Extract the RPM from the container image
podman create --name ra-build localhost/tnf-resource-agents-build:stream9
podman cp ra-build:/tmp/resource-agents.rpm ./resource-agents.rpm
podman rm ra-build

# Patch cluster nodes with the extracted RPM
ansible-playbook -i deploy/openshift-clusters/inventory.ini \
  helpers/apply-rpm-patch.yml \
  -l cluster_vms \
  -e rpm_full_path=$(pwd)/resource-agents.rpm

Alternatively, make patch-nodes (from deploy/) clones the resource-agents repo on the EC2 hypervisor, builds the RPM there natively, and patches the cluster nodes — all in one step, without needing a local container build.

See helpers/README.md for full documentation.

If you're using Claude Code, it can help you troubleshoot etcd issues on two-node fencing clusters. Simply ask Claude to diagnose your etcd problems and it will automatically collect diagnostics, analyze the cluster state, and recommend remediation steps. See .claude/commands/etcd/README.md for details.