Status › Run history

Run history, as data

The same fortnight the front page chart shows, published as a file you can read. The timeline below is drawn in your browser from /api/runs.json; the table underneath is the identical set of records and needs no scripting. Seventy runs of five nightly pipelines, 3–16 October 2026, on a staging install.

Dataset generated 2026-10-16 · demonstration data, not a live feed.

What the four flagged nights were

• 09 Oct · nightly-ledger · empty. Exit code zero, 11 rows written instead of about 2.4 million. Upstream handed us an empty extract and the job was none the wiser; this is the exact failure Slatepipe exists to catch and no exit-code alarm ever will.
• 11 Oct · crm-export · missed. The scheduler never fired. There is no run to inspect, so the missed_schedule rule reported it after the 45-minute grace window instead.
• 14 Oct · inventory-delta · slow. 30 minutes against a 9 minute median, 3.4× its own trailing figure, from a staging replica missing an index on the snapshot step.
• 15 Oct · billing-rollup · failed. Exit 1 inside the publish step after four minutes, rows out zero, downstream table left untouched.

Every run in the window

Newest first. The same rows the timeline uses; filtering above hides non-matching rows here too. Durations are wall-clock; a dash in the exit column is a run that never started.

Machine-readable copies: /api/runs.json and /data/runs.csv. Their fields are documented on the data page, which also lists the two database views the real interface reads.