Cloud Logging

Cloud Logging centralizes logs from Cloud Run, GKE, Cloud Functions, load balancers, and audit trails. Structured JSON logs from Node.js (`pino`, `winston`) map cleanly to filters, sinks, and log-based metrics without fragile regex parsing.

					import pino from 'pino';

const logger = pino({
  level: process.env.LOG_LEVEL ?? 'info',
  base: { service: 'checkout-api', revision: process.env.K_REVISION },
});

logger.info({ orderId, latencyMs }, 'checkout completed');
				

| Tool | Purpose | |------|---------| | Log sink | Export to BigQuery or Cloud Storage for analysis | | Log-based metric | Turn error patterns into alerts | | Trace integration | Tie logs to Cloud Trace spans via `trace` field |

Set retention per log bucket, redact PII at the source, and avoid high-cardinality labels in metrics derived from logs.

On interviews: structured logging versus plain text; correlation with trace IDs; cost of verbose debug in production; alerting on symptoms not raw log volume.

Common pitfalls: logging tokens or card data; printf-style strings that break search; no request ID across services; paying for 90-day retention on noisy debug logs.

The trade-off is rich observability versus ingestion cost, retention compliance, and alert noise.

Checklist:

• Emit JSON logs with request, trace, and service fields.
• Redact secrets and PII before write.
• Route long-term analytics through sinks, not infinite retention.
• Alert on SLO symptoms using log-based metrics sparingly.