仪表板

The operator dashboard is embedded directly in the admin server — no extra process, no Node toolchain. Open it at:

http://<admin-addr>/dashboard/

Default on localhost: http://localhost:9000/dashboard/

Logging in#

The dashboard uses a session cookie derived from the admin bearer token.

1. Navigate to /dashboard/login
2. Paste the admin token (same value as admin.token in config)
3. The session lasts 24 hours and slides on every request

Multi-pod requirement

All pods must share the same admin token. Each pod independently generates a random token if none is set, but session cookies are HMAC-signed with that token — a cookie issued by pod A is rejected by pod B if they have different tokens.

Set the token via the BOUINE_ADMIN_TOKEN environment variable (recommended for Kubernetes):

# Injected from Vault via chassis AppSecretSet — all pods get the same value

Or via the config file:

# Required for dashboard consistency across replicas
admin:
  token: "your-shared-secret"

On Kubernetes, set it via Helm:

helm upgrade bouine deploy/helm/bouine \
  --reuse-values \
  --set "config.admin.token=your-shared-secret"

Views#

Overview#

Live cluster-wide metrics polled every 5 seconds:

• Requests/s, Hit ratio, p99 latency, Error rate — with trend arrows vs the prior 60s window
• Throughput chart — 6h time series with 1H / 6H / 24H range selector
• Cache split donut — HIT / MISS / STALE / BYPASS / REVALIDATED breakdown
• Route performance table — per-route hit%, req/min, TTL, trend sparkline
• Hot & Warm store — tier fill bars, entry counts, evictions/min
• Consistent hash ring — proportional key-space ownership per pod
• Quick Purge — purge a URL directly from the overview

Performance#

Latency distribution and throughput analysis:

• p50 / p90 / p99 latency — per route and cluster-wide, updated every 5 s
• Stale-serving panel — STALE and REVALIDATED counts over time
• Throughput time series — requests/s with cache result breakdown

Routes#

Full route performance table with all cache policy columns (Pool, TTL, SWR, SIE, stayin_alive, jitter) plus:

• Hit ratio and req/min bar charts — per route, updated every 10 s
• URL drill-down — top URL paths (first 3 segments) sorted by request count

Cluster#

Live peer health table showing each pod’s Data addr, Admin addr, Weight, Joined time, and Status (online / stale). Includes:

• Ring stats — virtual nodes, load factor, hop limit, peer fetch timeout
• Circular ring SVG — key-space distribution with per-node labels and vnode counts
• Peer fetch stats — cluster-wide hit/miss counts and average peer latency

Invalidation#

Purge, Ban, and Refresh forms with real-time input validation:

The Recent invalidations list updates immediately after each successful operation.

Config#

• Running config viewer — structured read-only view of the live configuration (listen, storage, cluster, routes) with type-coloured values and inline hints
• Shows config file path and process uptime

Insights#

Rule-based operational insights that flag suboptimal configurations and anomalies across six categories:

• Cache — low hit rate, disabled caching on high-traffic routes, no TTL, no SWR, high eviction rate, warm tier near full
• Anomaly — bypass floods, p99 latency spikes, revalidation storms, vary explosion
• Upstream — unhealthy targets, high 5xx rate, no health checks, missing ETag or surrogate keys
• CDN — Cloudflare not configured, async latency, purge errors or skips
• Cluster — stale peers, hop limit ineffective, broadcast failures, degraded peer health
• Config — query params not stripped, allow-set-cookie enabled, zero jitter, TLS below 1.2, tracing sampling at zero

Each insight shows a severity badge, the affected route or component, and a recommended action. The page refreshes automatically every 30 seconds.

Time-range selector#

The 1H / 6H / 24H buttons in the tabs bar apply globally. The Overview throughput chart and route charts both respect the selected range. The default is 6H.

Theme#

Click ☀ light / 🌙 dark in the bottom-right corner. The preference is stored in localStorage and restored on page load.