On-Prem Prerequisites

This page covers only the prerequisites that differ from the AWS path. For the shared Red Hat, Microsoft, and lab-input guidance, keep AWS PREREQUISITES nearby.

What You Need On The Operator Workstation

• a local checkout of this repo
• ansible-core
• rsync
• ssh and a working SSH keypair
• enough local disk to stage the repo, pull secret, and generated artifacts

Unlike the AWS path, the on-prem target does not require:

• aws CLI
• CloudFormation permissions
• EBS quota
• Elastic IP allocation

Install the required collection with:

cd <project-root>ansible-galaxy collection install -r aws-metal-openshift-demo/requirements.yml

What You Need On The On-Prem Hypervisor

Assume you are starting from a freshly installed RHEL host that will act as virt-01.

Required host contract:

• RHEL host with virtualization support
• nested KVM available and stable
• libvirt / KVM capable hardware
• an uplink that OVS can build around
• SSH reachability from the operator workstation
• passwordless SSH access for root on the hypervisor using the same project keypair you will configure in:
  • inventory/hosts.yml via ansible_ssh_private_key_file
  • the matching public key already present in /root/.ssh/authorized_keys on virt-01
• enough local storage and RAM for the selected guest footprint
• a lab-management path from bastion back to the hypervisor, with explicit on-prem runtime settings for:
  • on_prem_bastion_hypervisor_host
  • on_prem_bastion_hypervisor_user

The current inventory example is:

• on-prem-openshift-demo/inventory/hosts.yml

The current group-wide on-prem settings live in:

• on-prem-openshift-demo/inventory/group_vars/all.yml

For a smaller host that should stop before the cluster build, start from the example override at:

• on-prem-openshift-demo/inventory/overrides/precluster-64g.yml.example

For the current ~128 GiB host class running the core-services+AD footprint, start from:

• on-prem-openshift-demo/inventory/overrides/core-services-ad-128g.yml.example

For the current cluster-capable external Ceph profile, start from:

• on-prem-openshift-demo/inventory/overrides/core-services-ad-plus-openshift-3node-external-ceph.yml.example

The override contract is documented in:

• OVERRIDE MECHANISM

Keep the access paths separate:

• inventory/hosts.yml describes how you reach the hypervisor from the operator workstation
• on_prem_bastion_hypervisor_host and on_prem_bastion_hypervisor_user describe how bastion reaches that same hypervisor on the lab network

For the current on-prem flow, treat root SSH trust as a prerequisite, not a bootstrap output. The workstation-side plays expect to connect to the hypervisor non-interactively with the configured project key before they can prepare the host.

The LVM Storage Contract

You must provide one LVM2 volume group that backs the guest logical volumes.

Required inputs:

• on_prem_lvm_volume_group
• optional on_prem_lvm_lv_name_prefix
• optional on_prem_lvm_guest_symlink_root

Optional seed inputs for a dedicated lab disk:

• on_prem_lvm_seed_enabled
• on_prem_lvm_seed_device
• on_prem_lvm_seed_force

The on-prem bootstrap validates that volume group and checks free space before any lvcreate runs. When the optional seed path is enabled, bootstrap can create the requested guest volume group directly from one explicit block device. That path is disabled by default and fails closed unless you opt into it. When forced, it uses the same destructive whole-device wipe profile the project uses for ODF backing-disk recovery: discard if possible, remove signatures, clear the first 2 GiB, wipe fixed deeper offsets, and clear the device tail before creating the guest VG.

Current full-footprint guest storage requirement from the stock volume inventory is approximately:

• 5950 GiB for all current guest logical volumes

That includes:

• bastion
• IdM
• mirror registry
• optional AD
• 3 control-plane nodes
• 3 infra nodes
• 3 worker nodes
• 3 ODF data disks

If you are not building the optional AD guest, subtract:

• 60 GiB

This storage figure does not include:

• host root disk sizing
• RHEL guest image cache
• mirror-content workspace overhead
• general operator headroom

Do not size the volume group to the raw sum and stop there.

Treat the AWS gp3 performance settings as workload hints. The current on-prem path preserves capacity and stable disk identity. It does not yet translate per-volume AWS gp3 IOPS and throughput settings into libvirt or host-level storage QoS controls.

Plan for:

• fast local SSD or NVMe storage, or a proven SAN-backed block tier
• enough aggregate I/O for concurrent guest boot, mirror-registry activity, and OpenShift node churn
• headroom beyond the raw LV capacity sum so the backend is not saturated under normal cluster and day-2 load

CPU And Memory Baseline

The current validated host baseline is still AWS m5.metal-like:

• 96 logical CPUs
• 384 GiB RAM
• symmetric socket / SMT layout

If the on-prem host drifts from that shape, read:

• HOST SIZING

Do not copy the current CPU-pool strings blindly onto a smaller or differently numbered host.

Red Hat And Content Inputs

You still need the same Red Hat content inputs as the AWS path:

• pull secret
• RHSM credentials
• RHEL 10.1 guest image source

The host image source is now your on-prem RHEL install, not an AWS AMI.

Keep the stock guidance nearby:

• AWS PREREQUISITES: RED HAT INPUTS

Optional Microsoft Inputs

If you plan to enable AD DS / AD CS, the optional Windows media requirements do not change.

Use the same source and placement guidance as the stock path:

• AWS PREREQUISITES: MICROSOFT INPUTS

Quick On-Prem Preflight

Before bootstrap, validate:

ssh -i ~/.ssh/id_ed25519 root@<hypervisor-management-ip> 'hostnamectl --static'ssh -i ~/.ssh/id_ed25519 root@<hypervisor-management-ip> 'vgs'ssh -i ~/.ssh/id_ed25519 root@<hypervisor-management-ip> 'virt-host-validate'test -f ~/pull-secret.txttest -f ~/.ssh/id_ed25519ssh-keygen -y -f ~/.ssh/id_ed25519 >/dev/null

From the repo:

cd <project-root>/on-prem-openshift-demoansible-playbook --syntax-check playbooks/site-bootstrap.ymlansible-playbook --syntax-check playbooks/site-precluster.ymlansible-playbook --syntax-check playbooks/site-lab.yml

For the smaller pre-cluster profile, copy the example override and edit it before the first run:

cd <project-root>/on-prem-openshift-democp inventory/overrides/precluster-64g.yml.example \  inventory/overrides/precluster-64g.yml

Where To Go Next

• for the on-prem build order: AUTOMATION FLOW
• for the on-prem run order: MANUAL PROCESS
• for resource drift and oversubscription guidance: HOST SIZING
• for the stock docs once the host contract is satisfied: AWS DOCS MAP