Steady State

Steady state is the measurable, normal operating behavior of a system, expressed as output metrics rather than internal details. It is the anchor of every chaos experiment: you confirm the steady state before a fault, then watch whether it holds during the fault.

Good Steady-State Signals

• p95 and p99 response latency (more revealing than averages)
• Error rate or success rate of a key transaction
• Throughput, such as orders processed per minute

Choosing the right signal matters: a gray failure can keep error rate at zero while latency quietly climbs, so a steady state defined only on error rate would miss it. This is why observability comes before chaos.