Skip to content

Reading monitor charts

Open any monitor and you get its full picture on one page: a set of summary tiles, an uptime timeline, and a latency chart — all driven by one time-window selector. Here's how to read each piece.

The window selector

Windows range from 3 hours to 12 months, with 24 hours as the default. Every number and chart on the page recalculates for the selected window, so "99.2% uptime" always means "in the window you're looking at". Short windows are for diagnosing something happening now; long windows are for trends, capacity conversations, and preparing client reviews.

The tiles

Five tiles summarize the window at a glance:

  • Uptime — the uptime percentage for the selected window.
  • Degraded — checks that succeeded but ran slower than the degraded threshold.
  • Failing — checks that failed.
  • Checks — how many checks ran in the window.
  • Last check — when the most recent result arrived.

Together they answer the first three questions of any investigation: is it healthy, is it slow, and is data actually arriving.

The uptime timeline

The timeline lays status across time: when the monitor was up, degraded, or down within the window. It's the fastest way to answer "when did this happen and how long did it last" — an incident shows as a red span you can line up against the latency chart below it, and repeated short red spans point at flapping rather than one clean outage.

The latency chart

Response times across the window (what is latency). Watch for three shapes: a step (latency jumped and stayed — something changed), a ramp (growing load or a leak — the outage is scheduled for later), and spikes (transient contention). The degraded threshold gives this chart its teeth: checks beyond it are the slow ones the amber status reflects (the degraded latency threshold).

Resolution and retention

UpAll keeps raw, per-check results for 100 days, and rolls results up into daily summaries kept forever. This shapes what the charts can draw:

  • Windows within recent raw data show per-check detail — individual response times, check by check.
  • Long windows — and anything reaching past the raw retention — render at daily resolution: one aggregated value per day.

So the 12-month view genuinely covers 12 months — nothing is lost — but it shows daily values rather than per-minute ones. Narrow the window when you need fine detail of a recent event; widen it when the question is a trend.

Reading tips

  • Amber is not downtime — degraded checks are successful checks, so they don't reduce uptime.
  • A gap with a paused or grey period isn't failure — no checks ran; see monitor statuses explained.
  • To scan all monitors at once, worst first, use the live status wall.