การตรวจสอบ การบันทึก และการสังเกตการณ์ระบบ

Deploying is only half the job. Once your backend serves real traffic you need to see what it is doing. Observability is the practice of understanding system state from its outputs.

• Logs tell you what happened
• Metrics tell you how much and how fast
• Traces tell you where time was spent

Supabase exposes logs for every service in the dashboard under Logs: API gateway, Postgres, Auth, Storage, and Edge Functions. Each is queryable with SQL-like syntax.

Filtering by status code and path quickly surfaces failing requests.

The Logs Explorer accepts SQL over structured log events. Here we count API errors grouped by path over the last hour.

Inside Edge Functions, prefer structured JSON logs over plain strings. They are far easier to filter and aggregate later.

Use consistent levels so you can filter noise from signal.

• debug for development detail
• info for normal events
• warn for recoverable problems
• error for failures needing attention

In production, avoid logging at debug to control cost and volume.

Watch a handful of Postgres metrics closely:

• Active connections vs the pool limit
• Cache hit ratio (aim above 99%)
• Slow query count
• Disk and CPU usage

Supabase surfaces these on the Reports page.

Enable pg_stat_statements to see which queries consume the most total time. This is the single most useful tool for performance triage.

Logs are persisted and often shared. Never write passwords, tokens, full card numbers, or service-role keys into logs.

• Redact sensitive fields before logging
• Log identifiers, not raw payloads
• Treat log access as a security boundary

Attach a request ID to every log line in a function so you can reconstruct a single request's full journey across logs.

Logs you never look at help no one. Set up alerts for the conditions that matter: error-rate spikes, connection exhaustion, disk nearing capacity.

Forward logs to an external sink (e.g. via webhooks or a log drain) when you need richer alerting than the dashboard provides.

Logs cost money and storage. Decide a retention policy deliberately.

• Keep high-value audit logs longer
• Sample or drop chatty debug logs
• Archive to cheap storage before deletion if needed for compliance

You now have a foundation for production observability.

• Use the Logs Explorer and structured JSON logs
• Apply consistent log levels and request IDs
• Track connections, cache hit ratio, and slow queries
• Use pg_stat_statements for performance triage
• Never log secrets; set alerts and a retention policy