hubbasicsgammaregimeflowhealth

Health — is every feed and store actually working right now?

Three independent tests, because six real month-long outages in this suite passed all the easy ones. Freshness against each store’s own cadence — the schedule is read from the live crontab and replayed, so weekends, trading-hour windows and weekly jobs are exact rather than guessed. A scheduled writer exists — a store nothing is scheduled to write is red however fresh it looks. Content is not degenerate — a fresh file whose rows are all thin, null or zero-length is red, because HTTP 200 and a new timestamp are not health. Everything here is measured from the server that serves this page: a feed can be dead here and alive elsewhere, and where that is true by design the row says so. This page only ever observes — it never writes to, repairs or triggers a pipeline.

loading…

Right now

What is wrong

every non-green row, with what it breaks and why it is not a false alarm

Store health ladder

missed scheduled runs as a fraction of each store’s own death line — the one scale a 5-minute recorder and a weekly build can share

Every store on one comparable overdue scale

0 = wrote on the last run · 1.0 = at its death line
— drag to resize —

Stores

newest artifact, declared cadence and where that expectation came from

Upstream feeds

Probe latency against each feed’s own banked history

bar = now · tick = its own median · faint mark = its own p95
— drag to resize —

The schedule itself

read live, read-only — a wiped crontab is its own failure mode

Scheduled jobs over time

this page’s own observation log
— drag to resize —

What this page checks, and what it deliberately does not