Runs, charts and how long we keep them
The Runs tab
Section titled “The Runs tab”A Monitor’s Runs tab lists every run from every one of its locations, newest first, 50 runs at a time. Load older runs reads the next ones.
Narrow the list with the filters above it: Location, Check, and Failures only. Timeline shows each run’s time as the parts it was made of, and Numbers shows the same parts as figures.
Open a run that met an error to see what went wrong, in words. When the location tried again and that worked, the run says so, and it counts as up. When your server answered, the run also shows the first bytes of its answer, for as long as we keep them.
A Heartbeat Monitor has a Signals tab instead: each call your job made, and each time one did not come in time.
Time to answer by location
Section titled “Time to answer by location”The Monitor’s Overview has a chart of the time to answer from each location, one line per location, over 24 h, 7 d or 30 d. Each point is the median Answer time of the runs in it. Over 24 h there is one point every 10 minutes. The legend gives each location’s median for the whole window.
A line breaks where nothing answered, such as during a timeout. It is never drawn as zero. A Regional outage shows as a shaded band on its location’s line. A Heartbeat Monitor has no chart, because nothing is timed.
The Uptime card on the Overview has a bar with a cell for each of the last 90 days, cut at midnight in your Project’s timezone.
How long we keep it
Section titled “How long we keep it”Retention is the same on every Plan.
| What | How long we keep it |
|---|---|
| Runs, with their timings and what went wrong | 30 days |
| What your server answered to a failed run | 7 days |
| Hourly summaries of Answer time | 396 days |
| Daily summaries of Answer time | For good |
| Incidents, and the hours not measured | For good |
| What was sent to a Webhook, and what your endpoint answered | 30 days |
The Runs tab says how long runs are kept. Once a run is gone, the summaries still draw the 7 d and 30 d charts. Your uptime percentage is worked out from Incidents and not-measured hours, which we keep for good, so it never changes when old runs go.