Root-cause analysis when the system is under load

Build a shared timeline

Align load events, deployments, logs, traces, and infrastructure metrics. Mark ramps, steady-state windows, failures, and recovery so every team investigates the same period.

Follow the constraint

Look for saturation, queueing, contention, and errors—not just high utilisation. A busy resource may be healthy; the constraint is the component that limits useful throughput or increases wait time.

Test the hypothesis

Change one meaningful variable and rerun a focused experiment. A credible root cause predicts behaviour: removing or moving the constraint should create a measurable change.

• State the hypothesis before changing the system
• Keep workload and data comparable
• Record configuration with every result

Close the loop

Document the causal chain, remediation, evidence, residual risk, and follow-up guardrail. The best RCA also makes the next occurrence easier to detect.