Adaptive sampling with AWS X-Ray to capture critical spans

This article explains AWS X-Ray's adaptive sampling feature, which dynamically adjusts trace sampling based on runtime conditions to balance observability and cost.

• Adaptive sampling combines Sampling Boost and Anomaly Span Capture mechanisms
• Sampling Boost automatically increases sampling rates when anomalies detected
• Anomaly Span Capture records critical spans independently of sampling rules
• Root services make sampling decisions; downstream services cannot override
• Configure local SDK settings via YAML for anomaly conditions and capture limits
• Sampling rules define baseline rates, reservoir sizes, and maximum boost rates
• Use low baseline rates for maximum adaptive sampling benefit
• Cooldown windows prevent continuous elevated sampling during incidents
• Multi-account/region: boost triggers only in same account/region as root service
• Monitor SamplingRate metrics to track boost activation and cost impact

Adaptive sampling enables cost-effective capture of critical diagnostic data during failures and latency spikes without increasing steady-state trace volume.