This One Weird Trick to Trace Your Application at Scale for Free

Recently, detailed tracing was added to streaming API servers, but the memory required exceeded expectations. To manage this issue, a feature flag was implemented in the instrumentation code, allowing control over whether or not to allocate memory for storing traces. This approach enables on-demand activation of detailed tracing within 200ms and preserves memory when the flag is off. LaunchDarkly's feature flags can be used with custom attributes such as serviceName, operationName, account, project, environment, and key to target specific traces for storage or exclusion. This allows for efficient and customizable tracing in applications without compromising performance.