index/status
Component status, bar by bar.
A delivery product owes you its own delivery record. One line per component, 90 days per line, every incident logged with a duration and a cause.
Sample The bars and incidents below are static fixtures showing the layout. Point this route at your monitor's output and delete this note.
all systems operational checked 2026-08-12 (at deploy)
No open incidents. The timestamp above is stamped by your browser at page load; the "operational" verdict is this demo's fixture, not a probe.
Last 90 days
Three components. One bar per day.
ingest api99.98% of 90 days sample
delivery workers99.99% of 90 days sample
dashboard99.95% of 90 days sample
- operational
- degraded
- outage
Incident log
What broke, for how long, and why.
| Date | Component | Impact | Duration | Cause and fix |
|---|---|---|---|---|
| ingest api | degraded | 21 min | p99 ingest latency rose past 400ms after a queue node lost its disk. Traffic drained to the standby; no events lost, deliveries delayed up to 90s. | |
| dashboard | outage | 12 min | Bad deploy served a blank shell. Rolled back at minute 9. Delivery pipeline unaffected: the dashboard reads the log, it never sits in the path. | |
| delivery workers | degraded | 34 min | A consumer returning 200 in 45s starved a worker pool. Added a per-endpoint concurrency cap and a 30s response deadline; backlog cleared in 6 min. |
Postmortems this specific are the point: an incident log that names causes reads as competence, not confession. Ship notes live in the changelog.