# Rhdh Ocp Pool

> List existing RHDH OCP Hive ClusterPool configurations and generate new pool YAML for a target OCP version, with imageSetRef aligned from other pools in the openshift/release repository. Covers OCP pools only, not K8s platforms

- Skill: `openshift/rhdh-ocp-pool` (Agent Skill, multi-file: 5 files)
- Install (CLI): `npx skillmds@latest add openshift/rhdh-ocp-pool`
- Raw SKILL.md: https://api.skillmd.com/api/skills/openshift/rhdh-ocp-pool/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: DevOps & Infra
- Author: openshift (https://skillmd.com/u/openshift)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/openshift/rhdh-ocp-pool

---

# RHDH OCP Cluster Pool Management

List and manage OCP Hive ClusterPool configurations for RHDH in the `openshift/release` repository. This skill covers OCP cluster pools only — K8s platform clusters (AKS, EKS, GKE) are provisioned via MAPT and OSD-GCP via a separate claim workflow.

## When to Use

Use this skill when you need to:
- List current RHDH cluster pools and their OCP versions
- Generate a new cluster pool YAML for a new OCP version
- Review cluster pool capacity (size, maxSize, runningCount)
- Remove a cluster pool for an end-of-life OCP version

## Prerequisites

- `yq` (v4+) must be available for YAML parsing
- Working directory must be the root of the `openshift/release` repository
- `CLAUDE_SKILL_DIR` is set automatically by the Claude Code skill loader. For manual invocation, set it to this skill's directory (e.g., `CLAUDE_SKILL_DIR=ci-operator/config/redhat-developer/rhdh/.claude/skills/rhdh-ocp-pool`)

## Listing Cluster Pools

Run the bundled script from the repository root:

```bash
bash "${CLAUDE_SKILL_DIR}/scripts/list-cluster-pools.sh"
```

### Output format

The script outputs a table with columns:

| Column | Description |
|--------|-------------|
| VERSION | OCP minor version (e.g., `4.18`) |
| POOL_NAME | Hive ClusterPool resource name |
| SIZE | Desired pool size |
| MAX | Maximum pool size |
| RUNNING | Number of clusters kept running (hibernation bypass) |
| IMAGE_SET | ClusterImageSet reference name |
| FILENAME | Pool YAML filename |

## Generating a New Cluster Pool

Use the bundled generation script:

```bash
bash "${CLAUDE_SKILL_DIR}/scripts/generate-cluster-pool.sh" --version 4.22
```

### With a specific reference pool

```bash
bash "${CLAUDE_SKILL_DIR}/scripts/generate-cluster-pool.sh" --version 4.22 --reference 4.21
```

### Preview without writing (dry-run)

```bash
bash "${CLAUDE_SKILL_DIR}/scripts/generate-cluster-pool.sh" --version 4.22 --dry-run
```

With `--dry-run`, the script outputs the generated YAML to stdout but does not write the file to disk. Use this to review the output before committing to the change.

### What the script does

1. **Looks up the `imageSetRef`** by scanning ALL cluster pools across the entire `clusters/hosted-mgmt/hive/pools/` directory (not just RHDH pools). This ensures alignment with other teams' pools for the same OCP version.
2. **Copies an existing RHDH pool** as a structural template (defaults to the latest, or use `--reference` to pick one).
3. **Updates version-specific fields**: `version`, `version_lower`, `version_upper`, pool `name`, and `imageSetRef`.
4. **Sets conservative sizing**: `size: 1`, `maxSize: 2`, no `runningCount`.
5. **Writes the file** directly to `clusters/hosted-mgmt/hive/pools/rhdh/`.
6. **Prints the generated YAML** to stdout for review.

### Error cases

- If no existing pool in the repo uses the target OCP version, the script **errors out** rather than guessing a patch version. This means the OCP version hasn't been released yet or image sets haven't been added to the repo.
- If an RHDH pool for the target version already exists, the script errors out.

### imageSetRef alignment

The `imageSetRef.name` follows the pattern:
```
ocp-release-<major>.<minor>.<patch>-multi-for-<major>.<minor>.0-0-to-<next_major>.<next_minor>.0-0
```

The patch version (e.g., `4.14.63`) is determined by what other pools in the repo already use, not by guessing. All pools for a given OCP version across all teams use the same imageSetRef.

## Removing a Cluster Pool

To remove a cluster pool for an end-of-life OCP version:

1. Delete the `*_clusterpool.yaml` file
2. Verify no CI jobs still reference this pool's version via `cluster_claim.version`
3. Use the `rhdh-ocp-tests` skill to check for and remove any remaining OCP test entries

## File Layout

All cluster pool files live in:
```
clusters/hosted-mgmt/hive/pools/rhdh/
├── OWNERS                                                    # Approvers/reviewers
├── admins_rhdh-cluster-pool_rbac.yaml                        # RBAC for pool admins
├── rhdh-aws-us-east-2.yaml                                   # Install config secret template
└── rhdh-ocp-<major>-<minor>-0-amd64-aws-us-east-2_clusterpool.yaml  # One per OCP version
```

## Key Details

- **Region**: All RHDH pools use `us-east-2`
- **Architecture**: `amd64`
- **Base domain**: `rhdh-qe.devcluster.openshift.com`
- **Worker nodes**: `m6i.2xlarge` (configured in `rhdh-aws-us-east-2` install config)
- **Credentials**: `rhdh-aws-credentials` secret
- **Namespace**: `rhdh-cluster-pool`