Resilience Pattern Tuner

Overview

Adjust resilience controls to reduce cascade risk without causing retry storms or unnecessary rejections. Produce configuration recommendations tied to measurable outcomes.

Workflow

1. Establish baseline behavior

• Record current latency, error, and queue metrics under normal and stressed loads.
• Identify whether failures are timeout-dominated, saturation-dominated, or dependency-dominated.

2. Tune timeout and retry chain

• Set bounded end-to-end deadlines.
• Tune retries with exponential backoff and jitter.
• Prevent retries from exceeding user-visible deadlines.

3. Tune isolation controls

• Configure circuit breaker thresholds and recovery windows.
• Configure bulkhead limits per critical path.
• Configure rate limits and load shedding policy.

4. Evaluate tradeoffs

• Measure changes in p99, error rate, and throughput.
• Check for side effects (false opens, early shedding, starvation).

5. Deliver tuning profile

• Provide settings, rationale, and expected impact.
• Include rollback conditions if regression appears.

Output format

1. Baseline issues
2. Proposed settings
3. Expected impact
4. Validation plan
5. Rollback criteria

References

• Read references/resilience-tuning-matrix.md for parameter heuristics.
• Use these repo sources:
• planning/IMPLEMENTATION_PLAN.md (Phase 5)
• docs/04-distributed-systems-and-failures.md (Chapter 24)
• canonical-catalogue/DSDS Canonical Catalogue - Patterns for scenario generation.csv