Performance Monitoring

Dune's built-in performance monitor gives you operational visibility into your running server — request latency percentiles, memory usage, cache hit rates, and slow queries — without any external services or agents.

All data is in-process and resets on server restart. This is an operational tool, not long-term analytics.

Dashboard

Navigate to Admin → Performance to see the live dashboard (refreshes every 30 seconds).

The dashboard shows:

  • Uptime — how long the server has been running
  • Request stats — total requests, error rate, p50/p95/p99 latency
  • Memory usage — heap used/total and RSS in MB
  • Engine stats — indexed page count, rebuild count, last rebuild time
  • Page cache stats — entries, hit rate, eviction count (when page cache is enabled)
  • Top routes — the 10 most-requested URLs with per-route latency
  • Slow queries — last 20 collection or search queries that exceeded the threshold

REST API

GET /admin/api/metrics

Requires admin role. Returns a full metrics snapshot as JSON.

Example response

{
  "ts": "2026-03-28T15:04:11.000Z",
  "uptimeSeconds": 3820,
  "requests": {
    "total": 15420,
    "errors": 12,
    "errorRate": 0.00078,
    "latency": {
      "count": 15420,
      "p50": 4.2,
      "p95": 18.7,
      "p99": 42.1,
      "max": 310.5,
      "mean": 6.1
    }
  },
  "topRoutes": [
    {
      "route": "/blog",
      "requests": 4201,
      "errors": 0,
      "latency": { "count": 4201, "p50": 3.1, "p95": 11.2, "p99": 24.8, "max": 87.3, "mean": 4.4 }
    }
  ],
  "pageCache": {
    "entries": 87,
    "hits": 14203,
    "misses": 211,
    "hitRate": 0.985,
    "evictions": 0
  },
  "memory": {
    "heapUsed": 48234496,
    "heapTotal": 67108864,
    "rss": 112345088,
    "external": 2048000
  },
  "engine": {
    "pageCount": 124,
    "rebuildCount": 3,
    "lastRebuildMs": 0
  },
  "slowQueries": [
    {
      "ts": "2026-03-28T14:58:22.100Z",
      "type": "collection",
      "query": "all published pages in /blog",
      "durationMs": 143
    }
  ]
}

Configuration

# system.yaml
metrics:
  enabled: true                # boolean — Enable metrics collector (default: true)
  slowQueryThresholdMs: 100    # number  — Log queries taking longer than this (default: 100)

To disable metrics collection entirely:

metrics:
  enabled: false

Latency percentiles

Latency is tracked using a fixed-size circular buffer (1,000 samples per route). When the buffer fills, the oldest sample is overwritten. Percentiles are computed from the current window on each snapshot request.

Reading the numbers:

  • p50 — half of requests complete faster than this
  • p95 — 95% of requests complete faster than this (a good general health indicator)
  • p99 — 99% of requests complete faster than this (highlights outliers)

All latency values are in milliseconds.

Slow queries

Collection and search queries that exceed slowQueryThresholdMs are recorded. The last 20 are retained in memory and shown on the dashboard and in the API response.

If you see frequent slow queries, consider:

  • Enabling the in-process page cache to reduce re-rendering
  • Narrowing collection queries (add filter criteria, reduce limit)
  • Checking for large content directories with many files

Notes

  • Metrics are in-process only — no files are written, no external services called
  • Data resets on server restart
  • GET /admin/api/metrics and the dashboard require admin role
  • Metrics have no effect in dune build --static mode (no server to monitor)