Implement metrics for mcp tool and operation counts and durations (#297)

* Implement metrics for mcp tool and operation counts and durations

* Changeset

* Unit test attribute setting in graphql.rs

* Add axum_otel_metrics for emitting basic http metrics about requests

* Lazy load singleton Meter for metrics

* Alphabetize

* Simplify result.is_error checking

Author: swcollard

.changesets/feat_otel_metrics.md

@@ -0,0 +1,11 @@
+### Implement metrics for mcp tool and operation counts and durations - @swcollard PR #297
+
+This PR adds metrics to count and measure request duration to events throughout the MCP server
+
+* apollo.mcp.operation.duration
+* apollo.mcp.operation.count
+* apollo.mcp.tool.duration
+* apollo.mcp.tool.count
+* apollo.mcp.initialize.count
+* apollo.mcp.list_tools.count
+* apollo.mcp.get_info.count

Cargo.lock

@@ -17,6 +17,7 @@ apollo-mcp-registry = { path = "../apollo-mcp-registry" }
 apollo-schema-index = { path = "../apollo-schema-index" }
 axum = "0.8.4"
 axum-extra = { version = "0.10.1", features = ["typed-header"] }
+axum-otel-metrics = "0.12.0"
 axum-tracing-opentelemetry = "0.29.0"
 bon = "3.6.3"
 clap = { version = "4.5.36", features = ["derive", "env"] }
@@ -65,6 +66,7 @@ chrono = { version = "0.4.41", default-features = false, features = ["now"] }
 figment = { version = "0.10.19", features = ["test"] }
 insta.workspace = true
 mockito = "1.7.0"
+opentelemetry_sdk = { version = "0.30.0", features = ["testing"] }
 rstest.workspace = true
 tokio.workspace = true
 tower = "0.5.2"

crates/apollo-mcp-server/Cargo.toml

@@ -1,6 +1,7 @@
 //! Execute GraphQL operations from an MCP tool
 
-use crate::errors::McpError;
+use crate::{errors::McpError, meter::get_meter};
+use opentelemetry::KeyValue;
 use reqwest::header::{HeaderMap, HeaderValue};
 use reqwest_middleware::{ClientBuilder, Extension};
 use reqwest_tracing::{OtelName, TracingMiddleware};
@@ -38,6 +39,9 @@ pub trait Executable {
     /// Execute as a GraphQL operation using the endpoint and headers
     #[tracing::instrument(skip(self))]
     async fn execute(&self, request: Request<'_>) -> Result<CallToolResult, McpError> {
+        let meter = get_meter();
+        let start = std::time::Instant::now();
+        let mut op_id: Option<String> = None;
         let client_metadata = serde_json::json!({
             "name": "mcp",
             "version": std::env!("CARGO_PKG_VERSION")
@@ -59,6 +63,7 @@ pub trait Executable {
                     "clientLibrary": client_metadata,
                 }),
             );
+            op_id = Some(id.to_string());
         } else {
             let OperationDetails {
                 query,
@@ -74,6 +79,7 @@ pub trait Executable {
             );
 
             if let Some(op_name) = operation_name {
+                op_id = Some(op_name.clone());
                 request_body.insert(String::from("operationName"), Value::String(op_name));
             }
         }
@@ -83,7 +89,7 @@ pub trait Executable {
             .with(TracingMiddleware::default())
             .build();
 
-        client
+        let result = client
             .post(request.endpoint.as_str())
             .headers(self.headers(&request.headers))
             .body(Value::Object(request_body).to_string())
@@ -118,7 +124,34 @@ pub trait Executable {
                 ),
                 meta: None,
                 structured_content: Some(json),
-            })
+            });
+
+        // Record response metrics
+        let attributes = vec![
+            KeyValue::new(
+                "success",
+                result.as_ref().is_ok_and(|r| r.is_error != Some(true)),
+            ),
+            KeyValue::new("operation.id", op_id.unwrap_or("unknown".to_string())),
+            KeyValue::new(
+                "operation.type",
+                if self.persisted_query_id().is_some() {
+                    "persisted_query"
+                } else {
+                    "operation"
+                },
+            ),
+        ];
+        meter
+            .f64_histogram("apollo.mcp.operation.duration")
+            .build()
+            .record(start.elapsed().as_millis() as f64, &attributes);
+        meter
+            .u64_counter("apollo.mcp.operation.count")
+            .build()
+            .add(1, &attributes);
+
+        result
     }
 }
 
@@ -127,6 +160,11 @@ mod test {
     use crate::errors::McpError;
     use crate::graphql::{Executable, OperationDetails, Request};
     use http::{HeaderMap, HeaderValue};
+    use opentelemetry::global;
+    use opentelemetry_sdk::metrics::data::{AggregatedMetrics, MetricData};
+    use opentelemetry_sdk::metrics::{
+        InMemoryMetricExporter, MeterProviderBuilder, PeriodicReader,
+    };
     use serde_json::{Map, Value, json};
     use url::Url;
 
@@ -366,4 +404,76 @@ mod test {
         assert!(result.is_error.is_some());
         assert!(result.is_error.unwrap());
     }
+
+    #[tokio::test]
+    async fn validate_metric_attributes_success_false() {
+        // given
+        let exporter = InMemoryMetricExporter::default();
+        let meter_provider = MeterProviderBuilder::default()
+            .with_reader(PeriodicReader::builder(exporter.clone()).build())
+            .build();
+        global::set_meter_provider(meter_provider.clone());
+
+        let mut server = mockito::Server::new_async().await;
+        let url = Url::parse(server.url().as_str()).unwrap();
+        let mock_request = Request {
+            input: json!({}),
+            endpoint: &url,
+            headers: HeaderMap::new(),
+        };
+
+        server
+            .mock("POST", "/")
+            .with_status(200)
+            .with_header("content-type", "application/json")
+            .with_body(json!({ "data": null, "errors": ["an error"] }).to_string())
+            .expect(1)
+            .create_async()
+            .await;
+
+        // when
+        let test_executable = TestExecutableWithPersistedQueryId {};
+        let result = test_executable.execute(mock_request).await.unwrap();
+
+        // then
+        assert!(result.is_error.is_some());
+        assert!(result.is_error.unwrap());
+
+        // Retrieve the finished metrics from the exporter
+        let finished_metrics = exporter.get_finished_metrics().unwrap();
+
+        // validate the attributes of the apollo.mcp.operation.count counter
+        for resource_metrics in finished_metrics {
+            if let Some(scope_metrics) = resource_metrics
+                .scope_metrics()
+                .find(|scope_metrics| scope_metrics.scope().name() == "apollo.mcp")
+            {
+                for metric in scope_metrics.metrics() {
+                    if metric.name() == "apollo.mcp.operation.count" {
+                        if let AggregatedMetrics::U64(MetricData::Sum(data)) = metric.data() {
+                            for point in data.data_points() {
+                                let attributes = point.attributes();
+                                let mut attr_map = std::collections::HashMap::new();
+                                for kv in attributes {
+                                    attr_map.insert(kv.key.as_str(), kv.value.as_str());
+                                }
+                                assert_eq!(
+                                    attr_map.get("operation.id").map(|s| s.as_ref()),
+                                    Some("mock_operation")
+                                );
+                                assert_eq!(
+                                    attr_map.get("operation.type").map(|s| s.as_ref()),
+                                    Some("persisted_query")
+                                );
+                                assert_eq!(
+                                    attr_map.get("success"),
+                                    Some(&std::borrow::Cow::Borrowed("false"))
+                                );
+                            }
+                        }
+                    }
+                }
+            }
+        }
+    }
 }

crates/apollo-mcp-server/src/graphql.rs

@@ -8,6 +8,7 @@ mod graphql;
 pub mod health;
 mod introspection;
 pub mod json_schema;
+pub(crate) mod meter;
 pub mod operations;
 pub mod sanitize;
 pub(crate) mod schema_tree_shake;

crates/apollo-mcp-server/src/lib.rs

@@ -0,0 +1,8 @@
+use opentelemetry::{global, metrics::Meter};
+use std::sync::OnceLock;
+
+static METER: OnceLock<Meter> = OnceLock::new();
+
+pub fn get_meter() -> &'static Meter {
+    METER.get_or_init(|| global::meter(env!("CARGO_PKG_NAME")))
+}

crates/apollo-mcp-server/src/meter.rs

@@ -4,6 +4,7 @@ use opentelemetry_sdk::{
     metrics::{MeterProviderBuilder, PeriodicReader, SdkMeterProvider},
     trace::{RandomIdGenerator, SdkTracerProvider},
 };
+
 use opentelemetry_semantic_conventions::{
     SCHEMA_URL,
     attribute::{DEPLOYMENT_ENVIRONMENT_NAME, SERVICE_VERSION},