How to use feature flags to safely test fixes in production

Feature Flags as a Debugging Tool

Feature flags let you deploy code to production without activating it for all users. This means you can test a bug fix in production -- with real data and real infrastructure -- against a small percentage of traffic before rolling it out fully. If the fix causes new problems, you turn off the flag instantly.

Implementing a Simple Flag

// Simple flag check backed by a config service
const flags = {
  useNewDiscountEngine: process.env.NEW_DISCOUNT_ENGINE === 'true'
};

function calculateDiscount(order) {
  if (flags.useNewDiscountEngine) {
    return newDiscountEngine(order);  // new fix, active for test users
  }
  return legacyDiscountEngine(order); // current code, all other users
}

// With a flag service (LaunchDarkly, Unleash, etc.)
const enabled = await flagClient.isEnabled('new-discount-engine', {
  userId: user.id,
  rolloutPercentage: 5  // enable for 5% of users
});

The Kill Switch Pattern

Every feature flag for a bug fix should also function as a kill switch -- a way to instantly revert to old behavior without redeployment. Deploy the fix behind a flag. Ramp up slowly: 1%, 5%, 25%, 50%, 100%. Monitor error rates and key metrics at each step. If errors spike, set the flag to 0% in seconds.