22 / 95 · 05 Performance & Chaos Engineering · Chaos Engineering Principles and Cycle← prev⊞ allnext →☰ Read as one page
4.4Common Mistakes in Chaos Engineering
| Mistake | Why It Happens | How to Avoid |
|---|---|---|
| No steady state definition | Teams skip straight to breaking things | Require a written hypothesis before every experiment |
| Testing only in staging | Fear of production impact | Graduate experiments through environments with increasing blast radius |
| One-time experiments | "We tested it once, it passed" | Automate experiments on a schedule |
| No kill switch | Overconfidence in the experiment design | Every experiment must have an automatic abort condition |
| Blaming chaos for outages | Misunderstanding the purpose | Chaos experiments should be boring -- they confirm resilience, not create incidents |
| Starting too big | Ambition outpaces maturity | Kill one pod first. Network partition can wait |