Status Draft
INC-001: Degraded performance on Ripeness API
1. The promise
Before anything went wrong, we wrote down what “good” means. The Ripeness API should give correct readings for of minutes over any 90 days. The rest is the error budget: the unreliability we can afford.
2. Service status
Ripeness is only up when every component on the status page above is up, so its availability is the product of theirs. That page is still following you down this one.
Over the last 90 days, Ripeness is at against a target of , .
3. Incident communications
Impact began at 08:40 UTC. The alert that should have caught it did not exist, so a customer found it first. Below is the page that never came, then the updates we posted, in order.
4. Postmortem
5. What good looks like
- Write down what “good” means before you need it.
- Alert on customer pain, not on causes.
- Acknowledge fast, before you know the cause.
- Lead with customer impact, in plain language.
- Post on the schedule you promised, even with no news.
- Give the next update time, never a guessed fix time.
- Look for contributing factors, not a culprit.
- Test the obvious theory before you believe it.
Pineapple: “Pineapple Scan” by Franco Pizzani, CC BY 4.0, compressed for the web.