Logging Level and Error Reporting Usage

Context

Currently, the volume and quality of logs—particularly Sentry errors—are reducing the effectiveness of the tool for debugging and operations. Much of the logging noise stems from messages that do not require action, creating false alarms and distracting from actionable issues.

Sentry

Sentry should be a high-signal tool. All logs sent to Sentry must be actionable.

• Do not log messages to Sentry unless they require immediate investigation or fixing. Use the regular logger module for informational or ignorable issues.
• If a Sentry issue is raised:
  • Immediately assign it to yourself.
  • Fix it promptly. If unsure how to proceed, ask for help immediately.
• The long-term goal is to have a silent Sentry.

Logger

Use the logger module for all standard logging purposes.

• error logs are treated as 500 errors and must be addressed.
  • Prefer error over critical, alert, or emergency unless there is a strong justification, in which case document the reasoning clearly.
• critical, alert, and emergency should be reserved for extreme situations, such as data corruption or unrecoverable system failures.
• warning logs are treated like 400 errors.
  • They signal that attention might be needed, but action is not always required.
  • If a warning proves meaningless or too frequent, it should either be removed, lowered, or replaced with metrics.
• Info/debug logs are always acceptable for internal understanding and observability.

Any log with the error critical alert or emergency must be acted upon. Ideally, we should not have any of those logs. Treat them as an 500 error. Being said, use error level instead of critical alert or emergency unless there is a strong reason not to; please document it in that case.

Use warning to call the attention, but it may or may not require action. Ideally, we move the warning to a lower or higher level or add metrics instead. Treat warning the log level as an 400 error.

Resolution

• You MUST NOT log to Sentry unless the issue is actionable and needs fixing.
• You MUST treat error logs as production issues to be fixed. You SHOULD validate log level usage as part of code reviews. PRs introducing error or higher logs must justify them.
• You SHOULD avoid using critical, alert, or emergency unless strictly necessary.
• You SHOULD avoid noise from warning logs; triage them regularly.
• You MUST ensure all log levels reflect the severity and actionability of the issue.
• You SHOULD consider replacing repetitive logs with metrics when appropriate. You SHOULD prefer metrics for tracking recurring behavior or performance patterns instead of relying on log statements.
• You MUST use info or debug levels for logs that are purely observational and don’t require intervention.

Terminology

Error Reporter

An Error Reporter would be a service like Sentry, Datadog Error Tracking.