Reading the dashboard

The overview reports freshness, uptime, memory use, root storage use and power conditions. Detailed sections cover compute, memory, storage, network, hardware and collection. The catalog provides definitions and availability requirements. History ranges are 1 hour, 6 hours, 24 hours, 7 days and 14 days, capped at approximately 600 points per series.

Counters appear as per-second rates over a five-minute window in charts. Gauge charts show measured values. CPU time is split by core and mode; idle, I/O wait and busy modes are distinct. Load is the number of runnable or uninterruptible tasks, not a CPU percentage. Available memory includes reclaimable capacity. Negotiated Wi-Fi bitrate is not throughput. Core voltage is not supply input voltage or wall power.

The portal polls every 15 seconds and pauses polling when hidden. A healthy API process does not imply healthy collectors. Freshness is calculated from the collector’s report timestamp: live through 60 seconds, stale above 60 seconds, offline above five minutes. No received report is missing. A failed API is failed, not zero usage. Optional unsupported sources remain unavailable.

Details beneath each chart contain its definition and an accessible table of the latest 20 points. Keyboard focus is visible; selectors and navigation have touch-sized controls. Mobile navigation opens from the menu button. Reduced-motion preferences disable chart animation.

The oldest received sample defines the start of coverage. A retention setting alone does not mean 14 days of observations exist. No availability percentage should be interpreted as an SLO or inferred from an incomplete history.

Availability and coverage

The overview evaluates one check every five minutes across the previous 24 hours (288 possible checks). A check is fresh when the collector report is at most 60 seconds old. Freshness availability is the percentage of observed checks that were fresh; coverage is the percentage of the 288 checks that had an observation. Uncovered time is unknown, not healthy or failed. This is sampled collection availability, not an application SLO or continuous uptime calculation. Normal metric charts suppress stale-collector intervals and show gaps.