CodeGraph CLI MCP Server

# Monitoring Setup This document describes how to run the monitoring stack for the CodeGraph application. ## Prerequisites - Docker - Docker Compose ## Running the Monitoring Stack 1. Start the Docker daemon. 2. Run the following command from the root of the project: ```bash docker-compose up -d --build ``` This will start the following services: - `codegraph-api`: The main application. - `prometheus`: The Prometheus server. - `grafana`: The Grafana server. - `alertmanager`: The Alertmanager server. ## Accessing the Services - **Prometheus**: <http://localhost:9090> - **Grafana**: <http://localhost:3001> - **Alertmanager**: <http://localhost:9093> ## Metrics The application exposes the following metrics on the `/metrics` endpoint: - `sync_operations_total`: A counter for the total number of sync operations. - `sync_operation_duration_seconds`: A histogram for the duration of sync operations. - `memscope_active_memory_bytes` (feature: `leak-detect`): Active heap bytes tracked by memscope. - `memscope_active_allocations` (feature: `leak-detect`): Count of active allocations. - `memscope_leaked_memory_bytes` (feature: `leak-detect`): Bytes considered leaked. - `memscope_leaked_allocations` (feature: `leak-detect`): Leaked allocation count. ## Dashboards A pre-configured Grafana dashboard is available. To import it: 1. Go to the Grafana UI at <http://localhost:3001>. 2. Log in with the default credentials (admin/admin). 3. Go to Dashboards -> Import. 4. Upload the `grafana-dashboard.json` file. ## Alerts The following alerts are configured in `prometheus.rules.yml`: - `SyncJobFailed`: Fires when a sync job fails. - `SyncJobSlow`: Fires when the 95th percentile of sync job duration is over 1 hour. - `MemoryLeaksDetected`: Fires when leaked allocations are non-zero. - `RisingActiveMemory`: Warns on >100MB growth in 10m.