3 ms·Jepsen‑style chaos CI before clustering2 points by boulevard 1y agoboulevard 1y agoHow we validate crash, pause, and disk‑full on a single node in CI using Docker, tmpfs, SSE watches, and Prometheus—no server code changes, in our code changes