Skip to main content

Dashboard Overview

The LoadFlux dashboard is a React application served at your configured path (default /loadflux). It has six pages accessible from the sidebar.

Dashboard (Home)​

Dashboard HomeDashboard Home

The main page shows a real-time overview:

  • Request rate (live) — RPS (requests per second) and RPM (requests per minute); rolling in-memory windows on a single instance, or summed from recent endpoint_metrics in cluster mode
  • Daily RPS — Average requests per second over the last 24 hours, computed from stored endpoint metrics (SUM(request_count) / 86,400). Refreshes with the snapshot DB cache (~8s), not every SSE tick’s live path
  • Total requests / error rate — O(1) lifetime counters (not reduced when retention deletes old rows). Subtitles show lifetime · since the first flush time; the footer shows the oldest stored metric (earliest row still in the DB) alongside the effective retention
  • Response time — Average, P95, and P99 latency (last 1 hour from the DB)
  • Error rate — Percentage from lifetime totals (see above)
  • Status breakdown — 2xx, 4xx, and 5xx counts (last 1 hour)
  • Top endpoints — Highest traffic routes (1h window)
  • Slowest endpoints — Routes with highest P95 latency (1h window)
  • Highest error rate — Routes with the most errors (1h window)
  • Server info — Node.js version, platform, PID, active SSE connections (per serving instance), cluster instance count when cluster mode is on

All data on this page updates in real-time via SSE (every 2 seconds).

System​

System MetricsSystem Metrics

System resource monitoring with time-series charts:

  • CPU usage — Percentage over time
  • Memory — Used vs total RAM
  • Disk — Used vs total space (where available)
  • Network I/O — Bytes received/transmitted (Linux only; macOS/Windows show N/A)

Use the time range control (custom from/to date and time, then Apply) to query any window within your retention. Last 1h resets to a rolling one-hour window (the default). Wide ranges are downsampled on the server (max_points query parameter) so charts stay responsive.

When cluster mode is enabled, an instance selector lets you view cluster-aggregated system metrics (default) or a single replica.

Endpoints​

EndpointsEndpoints

Detailed API endpoint analytics:

  • Top endpoints by requests — Most-hit routes
  • Top endpoints by latency — Slowest routes (by average duration)
  • Slow requests — Requests exceeding the slow threshold (default 500ms)
  • Endpoint metrics table — Full breakdown with request count, error count, and latency percentiles
  • Search — Filter top endpoints, the metrics table, and slow requests by path or method (debounced; server-side search query)

App Metrics​

App MetricsApp Metrics

Node.js process health:

  • Heap usage — V8 heap used vs allocated
  • Event loop delay — Average and max delay in milliseconds
  • GC pauses — Garbage collection pause duration
  • Uptime — How long the server has been running (real-time from SSE)

With cluster mode enabled, use the instance selector to filter process metrics to one container or view the cluster aggregate.

Errors​

ErrorsErrors

Error tracking and analysis:

  • Status distribution — Doughnut chart of 2xx/3xx/4xx/5xx responses
  • Error log — Table of individual error responses with timestamp, method, path, status code, error message, and duration
  • Search — Filter the log by path, method, message, or related text (debounced)
  • Status filter — Restrict the log to a specific HTTP status code seen in the current range

Settings​

SettingsSettings

Dashboard configuration:

  • Retention — View and change how many days of data to keep
  • Slow threshold — Adjust the millisecond threshold for slow requests
  • Export — Download metrics data as JSON for a selected time range (1 hour to the configured retention period)

Theme​

Toggle between dark and light mode using the theme switch in the navbar. The dashboard respects your system preference by default.

The navbar also shows a Live / Disconnected indicator for the SSE connection (and the serving instance_id when cluster mode is on).

Time range (historical pages)​

System, Endpoints, App metrics, and Errors share the same control:

  • From / To — Pick local date and time for the start and end of the window, then click Apply. The end must be after the start.
  • Last 1h — Clears the custom window and uses a rolling last hour (now - 1h … now), refreshed on each poll.

The home dashboard continues to use the live SSE snapshot (not this range).