Logging & Observability Standards

loggingobservabilitymonitoringdevops

Standards for structured logging, tracing and operational observability

Целевые файлы

Файл	Формат
.cursorrules	plaintext
CLAUDE.md	markdown

Содержимое

Use structured JSON logging in all production services; never log plain concatenated strings. Include timestamp, level, service, traceId, spanId, and userId in every log entry. Use log levels correctly: DEBUG (dev details), INFO (business events), WARN (recoverable issues), ERROR (action needed). Never log sensitive data: passwords, tokens, PII, credit card numbers, or SSNs. Log at the boundary: incoming requests, outgoing calls, and background job lifecycle events. Include the original error object (with stack) when logging exceptions, not just the message. Emit metrics alongside logs: request duration, queue depth, cache hit rate, error rate. Use correlation IDs propagated from the entry point through all downstream calls. Set log sampling for high-volume DEBUG logs to avoid overwhelming storage. Ship logs to a centralized system (${log_destination}); never rely only on stdout in production. Define log retention policy: ${log_retention_days} days for production, 7 days for dev/staging. Create alerts on ERROR-level log rate increases, not just on specific error messages. Add trace IDs to HTTP response headers (X-Trace-Id) so clients can report them in bug reports. Review log coverage in code review: every error path must emit at least a WARN log.