OpenShift MCP Server Multi Cluster with ACM Demo

Key Concepts

• OpenShift MCP Server: A server that exposes tools and functionalities through the Model Context Protocol (MCP).
• MCP Inspector: A tool used to inspect and interact with MCP servers.
• Advanced Cluster Management (ACM): A Red Hat product for managing multiple Kubernetes clusters.
• Cluster Provider Strategy: Configuration within the MCP server that defines how it discovers and connects to different clusters.
• ACM Cube Config: A cluster provider strategy that allows the MCP server to access clusters via ACM using a local kubeconfig file.
• ACM Cluster Proxy Add-on: A component that enables the MCP server to communicate with managed clusters through the ACM hub cluster.
• Hub Cluster: The central management cluster in an ACM setup.
• Managed Clusters: Clusters that are managed by the ACM hub cluster.
• pods list tool: An example tool within the MCP server that retrieves a list of pods from a specified cluster.
• Cluster Parameter: An optional parameter added to MCP tools when using ACM, allowing users to select the target cluster.

Introduction to OpenShift MCP Server and ACM Integration

The video demonstrates how to leverage the OpenShift MCP server with Advanced Cluster Management (ACM) to achieve multi-cluster tool access. The presenter, Callum, begins by showcasing the MCP inspector, a tool for interacting with MCP servers. Initially, the MCP server is configured to connect to a local Kind cluster, and the available tools, such as pods list, are shown to operate only on this local environment. The absence of multi-cluster capabilities is noted, prompting the transition to ACM integration.

Configuring the MCP Server for ACM

The core of the demonstration involves configuring the OpenShift MCP server to utilize ACM. This is achieved by modifying the MCP server's configuration file.

Key Configuration Steps:

1. Setting the Cluster Provider Strategy: The cluster_provider_strategy is set to acm_kubeconfig. This directive instructs the MCP server to discover and connect to clusters through ACM, using a local kubeconfig file.
2. Specifying the ACM Hub Cluster Context: The configuration requires identifying the specific context within the kubeconfig file that points to the ACM hub cluster. In the presented example, this context is named ACM.
3. Configuring the ACM Cluster Proxy Add-on: To communicate with managed clusters, the acm_cluster_proxy add-on is employed. This requires specifying the host (the route to the ACM cluster proxy) and a ca_file for secure communication. The presenter notes that the actual route is redacted for the demo but would be pasted into this field.

Alternative Configuration:

The presenter also mentions an alternative configuration where acm could be used instead of acm_kubeconfig. In this scenario, the context_name would not be needed, and the proxy_host would be set to the service URL of the proxy. This alternative allows for deploying the MCP server directly within the hub cluster.

Multi-Cluster Tool Access with ACM

After applying the ACM configuration, the MCP server is restarted. The impact of this change is immediately visible when relisting the available tools.

Key Observations:

• Introduction of the Cluster Parameter: Tools like pods list now include an optional cluster parameter. This parameter allows users to specify which cluster they want to target for the tool's execution.
• Default Behavior: By default, the cluster parameter is set to the hub cluster, meaning that running a tool without specifying a cluster will execute it on the hub cluster itself.
• Targeting Managed Clusters: By providing the name of a managed cluster (e.g., JBMC1) to the cluster parameter, the MCP server authenticates against that specific cluster and retrieves the requested information (e.g., pods) from it.

This mechanism enables all MCP tools to be dynamically configured to run against different clusters within the ACM-managed environment.

Conclusion and Key Takeaways

The demonstration successfully illustrates how to integrate the OpenShift MCP server with ACM to enable multi-cluster tool access. By configuring the cluster_provider_strategy to acm_kubeconfig and utilizing the acm_cluster_proxy add-on, the MCP server gains the ability to discover and interact with managed clusters. The introduction of the cluster parameter in MCP tools is the key enabler for selecting target clusters, allowing for unified tool execution across a distributed Kubernetes landscape managed by ACM. This provides a powerful way to manage and inspect resources across multiple clusters from a single point of access.