Implement Prometheus metrics endpoint

README.md

@@ -138,3 +138,19 @@ The MCP server provides the following tools for interacting with the OpenShift A
 * **Create a cluster**: "Create a new cluster named 'my-cluster' with OpenShift 4.14 and base domain 'example.com'"
 * **Check cluster events**: "What events happened on cluster abc123?"
 * **Install a cluster**: "Start the installation for cluster abc123"
+
+## Prometheus Metrics
+
+The MCP server exposes Prometheus metrics to monitor tool usage and performance. The metrics are available at `http://localhost:8000/metrics` when the server is running.
+
+### Available Metrics
+
+* **assisted_service_mcp_tool_request_count** - Number of tool requests.
+* **assisted_service_mcp_tool_request_duration_sum** - Total time to run the tool, in seconds.
+* **assisted_service_mcp_tool_request_duration_count** - Total number of tool requests measured.
+* **assisted_service_mcp_tool_request_duration_bucket** - Number of tool requests organized in buckets.
+
+### Metric Labels
+
+All metrics include the following label:
+* **tool** - The name of the tool, for example `cluster_info`, `list_clusters`, etc.

pyproject.toml

@@ -11,6 +11,7 @@ dependencies = [
     "requests>=2.32.3",
     "retry>=0.9.2",
     "types-requests>=2.32.4.20250611",
+    "prometheus_client>=0.22.1",
 ]
 
 [dependency-groups]

server.py

@@ -8,13 +8,17 @@
 import json
 import os
 
+
 import requests
-from mcp.server.fastmcp import FastMCP
+import uvicorn
 from assisted_service_client import models
+from mcp.server.fastmcp import FastMCP
 
-from service_client import InventoryClient
+
+from service_client import InventoryClient, metrics, track_tool_usage
 from service_client.logger import log
 
+
 mcp = FastMCP("AssistedService", host="0.0.0.0")
 
 
@@ -115,6 +119,7 @@ def get_access_token() -> str:
 
 
 @mcp.tool()
+@track_tool_usage()
 async def cluster_info(cluster_id: str) -> str:
     """
     Get comprehensive information about a specific assisted installer cluster.
@@ -141,6 +146,7 @@ async def cluster_info(cluster_id: str) -> str:
 
 
 @mcp.tool()
+@track_tool_usage()
 async def list_clusters() -> str:
     """
     List all assisted installer clusters for the current user.
@@ -174,6 +180,7 @@ async def list_clusters() -> str:
 
 
 @mcp.tool()
+@track_tool_usage()
 async def cluster_events(cluster_id: str) -> str:
     """
     Get the events related to a cluster with the given cluster id.
@@ -197,6 +204,7 @@ async def cluster_events(cluster_id: str) -> str:
 
 
 @mcp.tool()
+@track_tool_usage()
 async def host_events(cluster_id: str, host_id: str) -> str:
     """
     Get events specific to a particular host within a cluster.
@@ -222,6 +230,7 @@ async def host_events(cluster_id: str, host_id: str) -> str:
 
 
 @mcp.tool()
+@track_tool_usage()
 async def cluster_iso_download_url(cluster_id: str) -> str:
     """
     Get ISO download URL(s) for a cluster.
@@ -278,6 +287,7 @@ async def cluster_iso_download_url(cluster_id: str) -> str:
 
 
 @mcp.tool()
+@track_tool_usage()
 async def create_cluster(
     name: str, version: str, base_domain: str, single_node: bool
 ) -> str:
@@ -328,6 +338,7 @@ async def create_cluster(
 
 
 @mcp.tool()
+@track_tool_usage()
 async def set_cluster_vips(cluster_id: str, api_vip: str, ingress_vip: str) -> str:
     """
     Configure the virtual IP addresses (VIPs) for cluster API and ingress traffic.
@@ -362,6 +373,7 @@ async def set_cluster_vips(cluster_id: str, api_vip: str, ingress_vip: str) -> s
 
 
 @mcp.tool()
+@track_tool_usage()
 async def install_cluster(cluster_id: str) -> str:
     """
     Trigger the installation process for a prepared cluster.
@@ -391,6 +403,7 @@ async def install_cluster(cluster_id: str) -> str:
 
 
 @mcp.tool()
+@track_tool_usage()
 async def list_versions() -> str:
     """
     List all available OpenShift versions for installation.
@@ -411,6 +424,7 @@ async def list_versions() -> str:
 
 
 @mcp.tool()
+@track_tool_usage()
 async def list_operator_bundles() -> str:
     """
     List available operator bundles for cluster installation.
@@ -431,6 +445,7 @@ async def list_operator_bundles() -> str:
 
 
 @mcp.tool()
+@track_tool_usage()
 async def add_operator_bundle_to_cluster(cluster_id: str, bundle_name: str) -> str:
     """
     Add an operator bundle to be installed with the cluster.
@@ -458,6 +473,7 @@ async def add_operator_bundle_to_cluster(cluster_id: str, bundle_name: str) -> s
 
 
 @mcp.tool()
+@track_tool_usage()
 async def cluster_credentials_download_url(cluster_id: str, file_name: str) -> str:
     """
     Get presigned download URL for cluster credential files.
@@ -501,6 +517,7 @@ async def cluster_credentials_download_url(cluster_id: str, file_name: str) -> s
 
 
 @mcp.tool()
+@track_tool_usage()
 async def set_host_role(host_id: str, infraenv_id: str, role: str) -> str:
     """
     Assign a specific role to a discovered host in the cluster.
@@ -528,4 +545,6 @@ async def set_host_role(host_id: str, infraenv_id: str, role: str) -> str:
 
 
 if __name__ == "__main__":
-    mcp.run(transport="sse")
+    app = mcp.sse_app()
+    app.add_route("/metrics", metrics)
+    uvicorn.run(app, host="0.0.0.0")

service_client/__init__.py

@@ -7,5 +7,6 @@
 
 from .assisted_service_api import InventoryClient
 from .logger import log
+from .metrics import metrics, track_tool_usage
 
-__all__ = ["InventoryClient", "log"]
+__all__ = ["InventoryClient", "log", "metrics", "track_tool_usage"]

service_client/metrics.py

@@ -0,0 +1,56 @@
+"""
+Metrics for the MCP server.
+
+This module provides metrics for the MCP server.
+"""
+
+from typing import Callable, Any
+from functools import wraps
+
+from prometheus_client import (
+    CONTENT_TYPE_LATEST,
+    Counter,
+    Histogram,
+    generate_latest,
+)
+from starlette.requests import Request
+from starlette.responses import PlainTextResponse
+
+
+# Define counter for request count
+REQUEST_COUNT = Counter(
+    "assisted_service_mcp_tool_request_count",
+    "Request count",
+    ["tool"],
+)
+
+# Define histogram for request latency
+REQUEST_LATENCY = Histogram(
+    "assisted_service_mcp_tool_request_duration",
+    "Request latency",
+    ["tool"],
+    buckets=(0.1, 1.0, 10.0, 30.0, float("inf")),
+)
+
+
+def track_tool_usage() -> Callable:
+    """Decorate MCP tools with this decorator to track tool usage metrics."""
+
+    def decorator(func: Callable) -> Callable:
+        @wraps(func)
+        async def wrapper(*args: Any, **kwargs: Any) -> Any:
+            tool_name = func.__name__
+            REQUEST_COUNT.labels(tool=tool_name).inc()
+            with REQUEST_LATENCY.labels(tool=tool_name).time():
+                response = await func(*args, **kwargs)
+            return response
+
+        return wrapper
+
+    return decorator
+
+
+# Metrics route
+async def metrics(_request: Request) -> PlainTextResponse:
+    """Metrics endpoint."""
+    return PlainTextResponse(generate_latest(), media_type=CONTENT_TYPE_LATEST)