DigiCloneMCP — Server Tools & Readiness

DigiCloneMCP — Server Tools & Readiness

Prometheus-powered dashboard to monitor the DigiCloneMCP server: readiness, tool throughput (QPS), and latency percentiles.

Overview

• Grafana: 12.2.0+
• Datasource: Prometheus
• Default range: 1h
• Refresh: 30m
• Title: DigiCloneMCP — Tools & Readiness

Panels

Health

• Target Up → up
• App Ready → app_ready

Throughput

• Tool QPS (1m) → rate(mcp_tool_calls_total[1m])
• Total Calls (since start) → sum(mcp_tool_calls_total)
• Top Tools (QPS 1m) → topk(10, rate(mcp_tool_calls_total[1m]))

Latency

• p50/p95/p99 (5m) → histogram_quantile(<q>, sum by (le) (rate(mcp_tool_duration_seconds_bucket[5m])))
• p95 by Tool (5m) → histogram_quantile(0.95, sum by (le, tool) (rate(mcp_tool_duration_seconds_bucket[5m])))

Readiness Timeline

• App Ready over time → app_ready

Variables

• datasource: Prometheus source
• namespace: label_values(app_ready, namespace)
• job: label_values(app_ready{namespace="$namespace"}, job)
• tool (multi): label_values(mcp_tool_calls_total{namespace="$namespace", job=~"$job"}, tool)

Key Queries

• Target Up: max(up{job=~"$job", namespace="$namespace"})
• App Ready: max(app_ready{job=~"$job", namespace="$namespace"})
• Tool QPS (1m): sum(rate(mcp_tool_calls_total{job=~"$job", namespace="$namespace", tool=~"$tool"}[1m]))
• Latency p95 (5m): histogram_quantile(0.95, sum by (le) (rate(mcp_tool_duration_seconds_bucket{job=~"$job", namespace="$namespace"}[5m])))
• p95 by Tool (5m): histogram_quantile(0.95, sum by (le, tool) (rate(mcp_tool_duration_seconds_bucket{job=~"$job", namespace="$namespace", tool=~"$tool"}[5m])))

Import

Upload JSON

1. Dashboards → New → Import
2. Upload the JSON and choose your Prometheus datasource for DS_PROMETHEUS

Grafana Labs ID

• After publishing, you can import by ID directly.

Requirements

• DigiCloneMCP server must expose:
  • up
  • app_ready
  • mcp_tool_calls_total
  • mcp_tool_duration_seconds_bucket
• Metrics should include labels: namespace, job, tool

Usage

1. Select namespace and job for the DigiCloneMCP server.
2. Keep tool = All for aggregate view or pick tools to drill down.
3. Watch:
   • Up/Ready for quick health
   • QPS for throughput
   • Latency percentiles for slowdowns
   • Readiness timeline for flaps
4. For live ops, set refresh to 30s–2m.