Wide logging: Stripe's canonical log line pattern

5 hours ago

Stripe's canonical log line pattern emphasizes wide logging, where one structured record per unit of work includes all important fields.
Key fields in a canonical log line include route, method, status, duration, user/account ID, request/trace ID, build/deploy ID, feature flags, downstream timings, and error codes.
Stripe treats canonical log lines as critical infrastructure, ensuring they are emitted even during exceptions and used for long-term analysis.
High-value fields for diagnosis include route templates, identity info, release metadata, execution cost, decision inputs, and outcomes.
The real benefit of wide logging is correlation, allowing better questions about incidents, releases, and customer issues.
High-cardinality fields like user_id and request_id are acceptable in wide events if the system supports filtering and grouping.
Common mistakes include logging only happy paths, raw paths instead of templates, and letting schema drift.
Implementation typically involves middleware that collects request-scoped data and emits a structured log line at the end.

Hasty Briefsbeta