Break it on purpose.

Pick a failure, add mitigations if you want them, and run the test. The score is how much traffic still got through.

1. Pick a failure

2. Apply mitigations

Add a read replicaA second copy of the database serves reads if the primary goes down.Add a circuit breakerTrips after repeated failures, so a dying part stops being hammered.Add monitoringA failed part is detected and restarted automatically instead of staying down.Let app servers autoscaleMore instances are added automatically while the surviving ones run hot.

Metrics ingestion

A firehose of writes, buffered and spread over shards. This is the design the failure runs against.

Pick a failure on the left, add any mitigations you want to test, then run it. The score below is how many requests still succeeded.

Want to design the system yourself instead of testing a ready-made one? The System Builder has the same chaos monkey, plus a free canvas to build whatever you like.

Cookie preferences

Break it on purpose.

1. Pick a failure

2. Apply mitigations

Metrics ingestion