Skip to content
- Block implemented with explicit contract (inputs/outputs, errors, timeouts).
- Config for guardrails (network allowlist, fs scope, secrets scope).
- Observability sink (logs, traces, metrics) and dashboard access.
- Package the block with lockfile and runtime manifest (name, version, limits).
- Define environment config: secrets, allowlists, timeouts, retries, idempotency keys.
- Add healthcheck endpoint or command; test locally and in staging.
- Provision runtime (container/job) with CPU/mem limits and network policy.
- Wire observability: structured logs, trace context propagation, key metrics.
- Create rollout plan: canary cohort, success criteria, and rollback strategy.
- Deploy to staging → run fixtures → promote to canary → full rollout on green.
- Healthcheck green in staging and production.
- Canary success rate and latency within thresholds for agreed window.
- No policy violations (network, fs, secrets) in audit logs.
- Failing healthcheck: verify env vars, secrets scope, and network policy.
- Timeouts: increase budget or optimize external calls; add retries with backoff.
- Missing logs/traces: confirm instrumentation and sink credentials.
