sentry-instrumentation

To prevent Sentry metric sprawl and tag cardinality explosions, enforce a governed contract using MetricDef schemas, bounded tag value constraints, and bucket functions. This ensures consistent naming, safe tagging, and prevents silent metric drift across production codebases.

The best way to instrument HTTP routes and external API calls for Sentry metrics is using standardized emission patterns like the InstrumentedHttpClient. This applies a standard triad of counter, latency, and failure metrics to ensure consistent observability across dependency surfaces.

You can enforce Sentry metric governance rules in CI pipelines using a 13-check CI gate. This gate applies AST-based checks alongside pytest gates to automatically block invalid, noncompliant metric emissions before code merges.

You should use a failure taxonomy instead of raw exception strings to prevent unbounded tag cardinality. The classify() function maps errors to closed sets of bounded failure classes, keeping Sentry billing predictable while maintaining accurate failure counters.

To safely add retry loop and fallback path instrumentation without spiking Sentry costs, apply enforced sampling, rate limiting, and loop aggregation policies. These cost and loop safety rules prevent metric emission floods during recursive failures or retry storms.

Yes, this Sentry instrumentation approach works with existing pytest test suites by integrating AST-based checks and pytest gates. These tests validate MetricDef schema compliance and reject noncompliant metric emissions directly within your current testing workflow.