---

name: HyperShift Agent Provider description: Use this skill when you need to deploy HyperShift clusters on bare metal, edge environments, or disconnected infrastructures using pre-provisioned agents

HyperShift Agent Provider

This skill provides implementation guidance for creating HyperShift clusters using the Agent provider, which is designed for bare metal and edge deployments where pre-provisioned agents are available.

When to Use This Skill

This skill is automatically invoked by the /hcp:generate agent command to guide the Agent provider cluster creation process.

Prerequisites

• HyperShift operator installed and configured
• Pre-provisioned agents available in a Kubernetes namespace
• Pull secret for accessing OpenShift images
• Understanding of the target deployment environment

Agent Provider Overview

What is the Agent Provider?

The Agent provider:

• Designed for bare metal and edge deployments
• Requires pre-provisioned agents (nodes registered in advance)
• No cloud provider automation - manual network configuration required
• Suitable for disconnected/airgapped environments
• Direct control over hardware and network configuration

Common Use Cases

• Edge Computing: Remote locations with limited connectivity
• Bare Metal: On-premises hardware without cloud automation
• Disconnected Environments: Airgapped environments with security requirements
• Custom Hardware: Specialized hardware configurations
• Regulatory Compliance: Environments requiring specific data locality

Implementation Steps

Step 1: Analyze Cluster Description

Parse the natural language description for Agent-specific requirements:

Environment Type Detection:

• Edge: "edge", "remote", "limited connectivity", "small footprint"
• Bare Metal: "bare metal", "on-premises", "physical hardware", "custom hardware"
• Disconnected: "airgapped", "disconnected", "offline", "secure environment"

Resource Indicators:

• Minimal: "small", "edge", "minimal resources", "single node"
• Standard: "production", "multi-node", "high availability"

Special Requirements:

• FIPS: "fips", "compliance", "security"
• Custom Storage: "local storage", "storage class", "persistent volumes"
• Network Isolation: "isolated", "private network", "custom networking"

Step 2: Apply Agent Provider Defaults

Required Parameters:

• --agent-namespace: Namespace where agents are located
• --pull-secret: Path to pull secret file
• --release-image: OpenShift release image
• --base-domain: Base domain for the cluster (prompt user if not in description)

Smart Defaults by Environment:

Edge Environment:

--control-plane-availability-policy SingleReplica
--node-pool-replicas 1
--arch amd64  # or arm64 if specified

Bare Metal Environment:

--control-plane-availability-policy HighlyAvailable
--node-pool-replicas 3
--auto-repair true

Disconnected Environment:

--render  # Always render for review
--image-content-sources /path/to/image-content-sources.yaml
--additional-trust-bundle /path/to/ca-bundle.pem

Step 3: Interactive Parameter Collection

Required Information Collection:

1. Cluster Name

   🔹 **Cluster Name**: What would you like to name your cluster?
      - Must be DNS-compatible (lowercase, hyphens allowed)
      - Will be used for resource naming
      - Example: edge-lab-01, production-cluster
   

2. Agent Namespace

   🔹 **Agent Namespace**: In which namespace are your agents located?
      - This should be the namespace where you registered your agents
      - Example: default, agent-system, cluster-agents
      - [Press Enter for default: default]
   

3. Pull Secret

   🔹 **Pull Secret**: Path to your OpenShift pull secret file?
      - Required for accessing OpenShift container images
      - Download from: https://console.redhat.com/openshift/install/pull-secret
      - Example: /home/user/pull-secret.json
   

4. Base Domain (if not specified in description)

   🔹 **Base Domain**: What base domain should be used for cluster DNS?
      - This will be used for cluster API and application routes
      - Example: example.com, lab.internal, edge.local
   

5. OpenShift Version

   🔹 **OpenShift Version**: Which OpenShift version do you want to use?
   
      📋 **Check supported versions**: https://amd64.ocp.releases.ci.openshift.org/
   
      - Enter release image URL: quay.io/openshift-release-dev/ocp-release:X.Y.Z-multi
      - [Press Enter for default: quay.io/openshift-release-dev/ocp-release:4.18.0-multi]
   

Optional Configuration (based on description analysis):

6. Architecture (if detected)

   🔹 **Architecture**: Detected ARM architecture mention. Confirm architecture:
      - amd64 (x86_64)
      - arm64 (aarch64)
      - [Press Enter for default: amd64]
   

7. FIPS Mode (if compliance mentioned)

   🔹 **FIPS Mode**: Enable FIPS mode for compliance?
      - Required for certain regulatory environments
      - May impact performance
      - [yes/no] [Press Enter for default: no]
   

Step 4: Disconnected Environment Handling

If disconnected/airgapped environment is detected:

Additional Required Information:

1. Mirror Registry

   🔹 **Mirror Registry**: What is your mirror registry domain?
      - This registry should contain mirrored OpenShift images
      - Example: registry.example.com:5000
   

2. Image Content Sources

   🔹 **Image Content Sources**: Path to image content sources file?
      - Required for mapping official images to mirror registry
      - Example: /path/to/image-content-sources.yaml
   

3. Additional Trust Bundle (optional)

   🔹 **Additional Trust Bundle**: Path to custom CA bundle? (Optional)
      - Required if mirror registry uses custom certificates
      - Example: /path/to/ca-bundle.pem
      - [Press Enter to skip]
   

Always Use --render for Disconnected:

• Include --render flag to generate manifests for review
• Provide post-generation instructions for required manifest modifications

Step 5: Agent Availability Validation

Provide Agent Check Commands:

Before creating the cluster, verify your agents are available:

Check agents in namespace:
  kubectl get agents -n <agent-namespace>

Verify agent status:
  kubectl describe agents -n <agent-namespace>

Ensure agents are:
  - In "Available" state
  - Not bound to other clusters
  - Meet minimum resource requirements

Step 6: Generate Command

Basic Agent Cluster Command:

hypershift create cluster agent \
  --name <cluster-name> \
  --namespace <cluster-name>-ns \
  --agent-namespace <agent-namespace> \
  --pull-secret <pull-secret-path> \
  --release-image <release-image> \
  --base-domain <base-domain>

With Environment-Specific Flags:

Edge/Minimal Configuration:

hypershift create cluster agent \
  --name edge-cluster \
  --namespace edge-cluster-ns \
  --agent-namespace default \
  --pull-secret /path/to/pull-secret.json \
  --release-image quay.io/openshift-release-dev/ocp-release:4.18.0-multi \
  --base-domain edge.local \
  --control-plane-availability-policy SingleReplica \
  --node-pool-replicas 1

Production/HA Configuration:

hypershift create cluster agent \
  --name production-cluster \
  --namespace production-cluster-ns \
  --agent-namespace agent-system \
  --pull-secret /path/to/pull-secret.json \
  --release-image quay.io/openshift-release-dev/ocp-release:4.18.0-multi \
  --base-domain example.com \
  --control-plane-availability-policy HighlyAvailable \
  --node-pool-replicas 3 \
  --auto-repair

Disconnected Configuration:

hypershift create cluster agent \
  --name secure-cluster \
  --namespace secure-cluster-ns \
  --agent-namespace agent-system \
  --pull-secret /path/to/pull-secret.json \
  --release-image quay.io/openshift-release-dev/ocp-release:4.18.0-multi \
  --base-domain internal.example.com \
  --image-content-sources /path/to/image-content-sources.yaml \
  --additional-trust-bundle /path/to/ca-bundle.pem \
  --render

FIPS-Enabled Configuration:

hypershift create cluster agent \
  --name compliance-cluster \
  --namespace compliance-cluster-ns \
  --agent-namespace agent-system \
  --pull-secret /path/to/pull-secret.json \
  --release-image quay.io/openshift-release-dev/ocp-release:4.18.0-multi \
  --base-domain secure.example.com \
  --fips

Step 7: Post-Generation Instructions

For Disconnected Environments:

Provide additional configuration needed after cluster creation:

## Post-Creation Configuration (Disconnected)

After running the command above (with --render), you'll need to modify the generated manifests:

1. **Add ImageContentSources to HostedCluster:**
```yaml
spec:
  imageContentSources:
  - source: quay.io/openshift-release-dev/ocp-v4.0-art-dev
    mirrors:
    - registry.example.com:5000/openshift/release
  - source: quay.io/openshift-release-dev/ocp-release
    mirrors:
    - registry.example.com:5000/openshift/release-images
  - source: registry.redhat.io/multicluster-engine
    mirrors:
    - registry.example.com:5000/openshift/multicluster-engine

2. Disable Default OperatorHub Sources:

spec:
  configuration:
    operatorhub:
      disableAllDefaultSources: true

3. Apply the manifests:

kubectl apply -f <rendered-manifest-files>

**For All Environments:**

Next Steps

1. Verify agents are available: kubectl get agents -n <agent-namespace>

2. Monitor cluster creation: kubectl get hostedcluster -n <cluster-namespace> kubectl get nodepool -n <cluster-namespace>

3. Check agent assignment: kubectl get agents -n <agent-namespace> -o wide

4. Access cluster when ready: hypershift create kubeconfig --name <cluster-name> --namespace <cluster-namespace>

## Error Handling

### No Agents Available

**Scenario:** No agents found in specified namespace.

**Action:**

No agents found in namespace "default".

Please ensure:

1. Agents are properly registered
2. Agents are in "Available" state
3. Correct namespace specified

Check agents: kubectl get agents -A

Register new agents using the agent-based installer or manual process.

### Insufficient Agents

**Scenario:** Not enough agents for requested replicas.

**Action:**

Only 2 agents available, but you requested 3 node replicas.

Options:

1. Reduce node-pool-replicas to 2
2. Register additional agents
3. Use SingleReplica control plane (reduces total requirement)

Would you like me to adjust the configuration?

### Agent Already Bound

**Scenario:** Agents are already assigned to another cluster.

**Action:**

Some agents are already bound to other clusters.

Available agents: 1 Bound agents: 2

Check agent status: kubectl describe agents -n <namespace>

Free up agents by deleting unused clusters or register new agents.

### Disconnected Image Access

**Scenario:** Can't access release images in disconnected environment.

**Action:**

Release image access failed. In disconnected environments:

1. Ensure mirror registry contains required images
2. Verify image-content-sources file mapping
3. Check custom CA bundle configuration
4. Validate network connectivity to mirror registry

Mirror required images using: oc adm release mirror --to <mirror-registry>

## Best Practices

### Agent Management

1. **Pre-provision agents:** Register agents before cluster creation
2. **Agent naming:** Use descriptive names for easier management
3. **Resource allocation:** Ensure agents meet minimum requirements
4. **Health monitoring:** Monitor agent status regularly

### Network Configuration

1. **DNS planning:** Ensure proper DNS resolution for base domain
2. **Load balancing:** Configure external load balancer for API access
3. **Ingress:** Plan for application ingress and routing
4. **Firewall:** Configure appropriate firewall rules

### Security

1. **Pull secret security:** Protect pull secret files
2. **FIPS compliance:** Enable FIPS for regulated environments
3. **Certificate management:** Plan for custom certificate authorities
4. **Network isolation:** Implement appropriate network segmentation

### Disconnected Deployments

1. **Mirror planning:** Pre-mirror all required images
2. **Catalog management:** Prepare custom operator catalogs
3. **Certificate management:** Plan for custom CA bundles
4. **Registry security:** Secure mirror registry access

## Anti-Patterns to Avoid

❌ **Assuming cloud automation**

"Create cluster with autoscaling"

✅ Agent provider requires manual agent provisioning

❌ **Insufficient agent planning**

Not checking agent availability before cluster creation

✅ Always verify sufficient agents are available

❌ **Ignoring disconnected requirements**

Using public image references in airgapped environment

✅ Properly configure image content sources and mirror registry

❌ **Inadequate resource planning**

Using minimal agents for production workloads

✅ Size agents appropriately for expected workloads

## Example Workflows

### Edge Deployment

Input: "small edge cluster for remote office monitoring"

Analysis:

• Environment: Edge
• Scale: Minimal
• Use case: Monitoring

Generated Command: hypershift create cluster agent
--name edge-monitoring
--namespace edge-monitoring-ns
--agent-namespace default
--pull-secret /path/to/pull-secret.json
--release-image quay.io/openshift-release-dev/ocp-release:4.18.0-multi
--base-domain edge.local
--control-plane-availability-policy SingleReplica
--node-pool-replicas 1

### Secure Data Center

Input: "airgapped production cluster for financial data processing"

Analysis:

• Environment: Disconnected/Secure
• Scale: Production
• Use case: Financial (compliance required)

Generated Command: hypershift create cluster agent
--name financial-prod
--namespace financial-prod-ns
--agent-namespace secure-agents
--pull-secret /path/to/pull-secret.json
--release-image quay.io/openshift-release-dev/ocp-release:4.18.0-multi
--base-domain secure.financial.internal
--image-content-sources /path/to/image-content-sources.yaml
--additional-trust-bundle /path/to/ca-bundle.pem
--fips
--render

## See Also

- Agent-based OpenShift Installation Documentation
- HyperShift Agent Provider Documentation
- OpenShift Disconnected Installation Guide
- Agent-based Installer Documentation