Hi HN! I built Khaos to solve a problem I kept hitting: testing Kafka monitoring and alerting with realistic traffic patterns.

  It lets you:
  - Simulate scenarios like consumer lag, hot partitions, broker failures, rebalance storms
  - Generate realistic message schemas with Faker (names, emails, addresses, etc.)
  - Run correlated event flows across multiple topics
  - Test against local Docker cluster or external Kafka (Confluent Cloud, MSK, etc.)

  Example:
    khaos run consumer-lag -d 60

  This spins up a 3-broker cluster, creates topics, produces/consumes at configured rates, and injects the incident (slow consumers) at the scheduled time.

  Built with Python, confluent-kafka, and Typer. Single `uv tool install khaos-cli` to get started.

  Would love feedback on the scenario DSL and what other failure modes would be useful to simulate.