Dimensions
API Request Rate
- What: Requests per second/minute to exposed APIs.
- Why: Guards latency and stability.
- Symptoms: HTTP 429 (Too Many Requests); advise retry with backoff.
Endpoint Concurrency
- What: Max in‑flight requests per exposed top‑level Node.
- Why: Prevents hot‑spot overload on a single endpoint.
- Symptoms: Requests queue or fail fast.
External Trigger Frequency (Schedules)
- What: Minimum interval between scheduled invocations.
- Why: Limits fan‑out and repeated runs.
- Symptoms: Skipped ticks or delayed triggers.
Builds / Publish Actions
- What: Limit on number of build/publish operations per window.
- Why: Encourages batching of changes.
- Symptoms: Further publishes are deferred or rejected.
Resilience patterns
- Implement retry with jitter on 429s (e.g., exponential backoff).
- Use idempotency keys to avoid duplicate processing on retries.
- Add circuit breakers around hot paths that call expensive tools.
tip: Load shaping — Use lightweight validation Nodes at API edges to reject bad requests early and keep downstream capacity for real work.