Python Production Debugging — Middle¶

Make a service diagnosable before an incident.

• Log request or job IDs, operation, duration, and outcome.
• Emit metrics for rate, errors, latency, and saturation.
• Use traces to follow a request across HTTP, queues, and databases.
• Add health checks that reflect the ability to serve traffic.

When an alert fires, compare a failing request with a healthy one and change one variable at a time.