Chaos Monkey is great, but it may be more pragmatic to start by injecting failures manually, rather than through an automated process. We do this with Failure Fridays at PagerDuty - <a href="https://blog.pagerduty.com/2013/11/failure-friday-at-pagerduty/" rel="nofollow">https://blog.pagerduty.com/2013/11/failure-friday-at-pagerdu...</a>.