hello-errors

Use a unified error format like { code, message, details? } with accurate HTTP status codes: 4xx for client errors and 5xx for server errors. Keep error codes enumerable so frontends can handle them programmatically, and never expose stack traces, SQL, or file paths to users.

Classify errors before writing code: expected errors like validation failures get friendly messages returned normally, while unexpected errors like system outages get logged with a generic user-facing response. Only wrap code in try-catch when you can actually handle the exception.

Combine a timeout with exponential-backoff retries, capped at three attempts. Design critical operations to be idempotent so retries are safe, and add a circuit breaker that fails fast after consecutive failures exceed a threshold.

Use structured JSON logs with timestamp, level, and context. Error-level logs should include the stack trace, request ID, and user context, but sensitive data like passwords, tokens, and ID numbers must be masked before logging.

Empty catch blocks create silent failures that hide bugs and make systems appear healthy while functionality is broken. Every caught exception should be handled, logged, or rethrown, and silent degradation or fallback paths should be avoided.