failscale — watch it fail at scale

▟ chaos engineering

inject failure

incident response (it won't be enough)

traffic — 3,500 req/s

Auto-heal dead nodes 🐒 Chaos Monkey (hands-off mayhem) 📖 Show failure lessons

▟ cluster topology ● healthy ● degraded ● dead

requests / sec

error rate

0.0%

p99 latency

0ms

availability

100%

nodes alive

0/0

database load

▟ live telemetry

requests / sec 0

error rate % 0.0

p99 latency ms 0

failure mode

Welcome to the control room

This is a fully functional toy distributed system, currently serving traffic without complaint. That will not last. Press a red button — or just hit 1 — and find out how production really feels.