A cautionary tale from Knight Capital Group illustrates how misconfigured feature flags can lead to catastrophic financial losses.
Feature flags decouple deployment from release, enabling synchronized rollouts, A/B testing, targeted releases, and immediate kill switches.
Feature flags can be categorized into four types: short-lived release toggles, long-lived ops toggles, dynamic permission toggles, and targeted experiment toggles.
The feature flag maturity model helps you assess your current practices, from basic environment variables to fully operationalized systems with audit trails and observability.
OpenFeature is a CNCF incubating project that provides a standardized, community-driven API for feature flagging to avoid vendor lock-in.
The OpenFeature evaluation API uses a consistent singleton pattern across multiple languages and supports various data types beyond booleans, always providing a safe fallback value.
The provider model abstracts the feature flag management backend, allowing you to switch between homegrown solutions, vendors, or even multiple providers without changing application code.
Use the evaluation context to pass arbitrary data for dynamic flag rules and a targeting key to ensure deterministic rollouts for user cohorts.
OpenFeature hooks allow you to execute arbitrary code during the evaluation lifecycle, enabling observability through the OpenTelemetry hook to identify and clean up stale flags.
Advanced OpenFeature APIs like the tracking API help validate experiment outcomes, while the events API allows you to react to provider state changes in real-time.
Get involved with the OpenFeature project by exploring the documentation and SDKs, trying the reference implementation FlagD, or joining the community on Slack.
Jobs that call for the skills explored in this talk.