Docs · Concepts

Runs and history

What a run keeps, how long it keeps it, and the difference between history and a result snapshot.

#What a run stores

Counts, not rows. The question history exists to answer is were there more differences than yesterday, and four integers answer it: total, matched, differing, missing.

Alongside them a run keeps when it ran, what started it, whether it succeeded, how long it took, and the outcome of every Rule.

It also records the parameters it was run with as they were typed — never the resolved values. A webhook URL kept in an encrypted variable resolves during the run and has no business being written into a record a published reader can fetch.

#What started it

A run started through the API, or by a workflow's Run a check node, is recorded like any other run.

#Watching a run: the Live tab

A Check's Live tab, before History, shows the run under way, or the last one, step by step as the server begins each step:

The head of the tab says how the run started (from the Builder, a schedule, the published page or the API), when, and how long it took. Each step turns while it runs and ticks when it is done. A run whose process stopped reads as stopped after a minute, rather than as running for ever.

The steps are kept in the database, so a run on another process, such as a schedule on a worker, shows here too. Open in History goes to the run once it is saved.

#How long history is kept

History is pruned per Check, keeping the most recent runs. The ceiling comes from the licence:

The limit is read on every write rather than cached, so raising it takes effect without a restart.

#History and snapshots are different things

This distinction matters and the two are easy to confuse.

History is the counts. It is small, it is kept for many runs, and it is what the History tab draws.

A snapshot is the actual rows the run produced, compressed and stored so you can reopen a past run and look at it. Snapshots are kept for a much shorter window — days rather than runs — and they have a size ceiling: a run whose result is too large to store simply has no snapshot, and the run page says so.

When a snapshot has expired, the run is still in history and its counts are still readable. Only the rows are gone.

#The History tab

One of the tabs on a Check: "What each run of this check produced over time, from the oldest shown to the latest."

It shows:

• Clean runs — how many of the runs shown passed every Rule.
• Run time — average, median, P95, slowest and latest.
• Differences — rows that differ or are missing in each run. A failed run compared nothing and leaves a gap rather than a zero, which is the honest way to draw it.
• Runs — one bar per run, oldest to latest. Select one to see what it produced.

Selecting a run shows its duration, its four counts, how its Rules fared, and Since the run before — new, changed, unchanged and cleared.

#The run log

On the Result tab, the run log has its own tabs: Run, Rules, Patterns, Hosts and Notes. It heads itself either Healthy or with a count of things to look at.

The Hosts tab is the one to open when a source fans out across many machines and the totals look wrong — it is where a quietly unreachable host shows up.

#Clearing history

An admin can remove old run history from Storage → Maintenance, choosing an age in days: "Run history is the one thing here that grows without anybody deciding it should. Removing it loses the record of past runs, not the checks themselves."