Five Causes Behind Most Outages: An Opinionated Taxonomy
Incidents feel infinitely varied and mostly are not. Five recurring shapes, the tell for each, and why the taxonomy is worth more during an incident than after.
6 min read
Posts tagged Reliability
4 posts, newest first.
The Retry That Took Down the Service
A dependency got slow, every client retried, and the retries became the outage. The incident in order, including the twelve minutes it outlived its cause.
5 min read
Your p99 Is Lying to You: Percentiles Don't Compose
Ten services at p99 don't make a p99 page load: why per-service percentiles mislead, why averaging them is meaningless, and what the client sees that you don't.
5 min read
Backpressure and Rate Limiting: The Two Answers Candidates Skip
Adding a queue doesn't absorb a spike, it defers one. What backpressure and rate limiting actually do, why they're different, and the failure they each prevent.
4 min read
