Feature Flags

  4 minute read  

Feature flags allow you to enable, disable, or configure features in Pepr. They provide a way to manage feature rollouts, toggle experimental features, and maintain configuration across different environments. Feature flags work across both CLI and in-cluster environments, using the same underlying implementation.

Pepr’s feature flag system supports:

  • Boolean flags
  • String values
  • Numeric values
  • Default values for flags
  • Type safety through TypeScript
  • Configuration via environment variables
  • Configuration via initialization strings
  • Consistent behavior in both CLI and in-cluster environments

Best Practices

  1. Temporary Usage: Use feature flags for temporary changes, gradual rollouts, or experimental features. Do NOT use feature flags for permanent configuration.
  2. Descriptive Names: Choose descriptive flag names and provide detailed descriptions in the metadata.
  3. Default Values: Always provide sensible default values for feature flags.
  4. Type Safety: Leverage TypeScript’s type system when accessing feature flags.
  5. Cleanup: Remove feature flags that are no longer needed to maintain a clean codebase and stay within the feature flag limit.

Getting Started

Using Feature Flags

  1. Enable a CLI feature flag via environment variable:

    PEPR_FEATURE_REFERENCE_FLAG=true npx pepr@latest

  2. Enable a CLI feature flag via argument:

    npx pepr@latest --features="reference_flag=true"

  3. Configure feature flags for a Pepr module: Add to your package.json:

    "pepr": {
  "env": {
    "PEPR_FEATURE_REFERENCE_FLAG": true
  }
}

Defining Feature Flags

Feature flags in Pepr are defined in the FeatureFlags object. The FeatureFlags object is the source of truth for all available feature flags.

Important: To reduce complexity, Pepr enforces a limit of 4 active feature flags. Exceeding this limit results in an error (e.g., Error: Too many feature flags active: 5 (maximum: 4).)

Feature Flags have:

  • A unique key used for programmatic access
  • Metadata containing name, description, and default value

Feature flag values are converted to appropriate types:

  • "true" and "false" strings are converted to boolean values
  • Numeric strings are converted to numbers
  • All other values remain as strings

Example definition:

export const FeatureFlags: Record<string, FeatureInfo> = {
  REFERENCE_FLAG: {
    key: "reference_flag",
    metadata: {
      name: "Reference Flag",
      description: "A feature flag to show intended usage.",
      defaultValue: false,
    },
  },
};

Configuration and Workflows

Feature flags are configured in two ways for either CLI or Module use. Only feature flags defined in the FeatureFlags object can be configured. Attempting to use an undefined flag will result in an error.

Pepr CLI Workflow

  1. Using environment variables with the format PEPR_FEATURE_<FLAG_NAME>:

    # Run Pepr with debug logging to verify feature flags
PEPR_FEATURE_REFERENCE_FLAG=true LOG_LEVEL=debug npx pepr@latest

  2. Using the --features CLI argument:

    LOG_LEVEL=debug npx pepr@latest --features="reference_flag=true"

Pepr Module Deployment Workflow

For deployed modules, configure feature flags in the module’s package.json:

  "pepr": {
    "env": {
      "PEPR_FEATURE_REFERENCE_FLAG": true
    }
  },

Build and deploy the module once environment variables are set with a known $APP_NAME:

# Edit your package.json to include feature flags
jq '.pepr.env.PEPR_FEATURE_REFERENCE_FLAG = true' package.json

npx pepr@latest build
npx pepr@latest deploy -i pepr:dev

APP_NAME=pepr-module-name
kubectl logs -n pepr-system --selector app=$APP_NAME | grep "Feature flags store initialized"

Troubleshooting

Feature Flag Not Recognized

If your feature flag isn’t recognized:

  1. Verify the flag is defined in the FeatureFlags object
  2. Check the case of the environment variable (PEPR_FEATURE_REFERENCE_FLAG vs pepr_feature_reference_flag)
  3. Ensure correct format for CLI flags (reference_flag=true not REFERENCE_FLAG=true)

Too Many Feature Flags

If you receive the error: Error: Too many feature flags active: 5 (maximum: 4)

  1. Review active feature flags using featureFlagStore.getAll()
  2. Remove or consolidate unnecessary feature flags
  3. Consider if some flags could be regular configuration instead

Flag Value Incorrect Type

If your feature flag has an unexpected type:

  1. Remember that strings "true" and "false" are automatically converted to booleans
  2. Numeric strings are automatically converted to numbers
  3. Use explicit type checking in your code: typeof featureFlagStore.get(...)

For Maintainers

Development Guidelines

Accessing and Implementing Feature Flags

To add a new feature flag, update the FeatureFlags object in the source code. Any feature flag not defined in this object cannot be used in the application.

Access feature flags the same way in both CLI and in-cluster environments using the featureFlagStore.

This snippet shows how to execute different code paths based upon a feature flag value. To execute within the Pepr project, save the snippet to file.ts and run npx tsx file.ts.

import { featureFlagStore } from "./lib/features/store";
import { FeatureFlags } from "./lib/features/FeatureFlags";

// Get a boolean flag
const isReferenceEnabled = featureFlagStore.get(FeatureFlags.REFERENCE_FLAG.key);

const allFlags = featureFlagStore.getAll();

// Use feature flags for conditional logic
if (featureFlagStore.get(FeatureFlags.REFERENCE_FLAG.key)) {
  Log.info(`Flag value: ${isReferenceEnabled} of type ${typeof featureFlagStore.get(FeatureFlags.REFERENCE_FLAG.key)}`)
} else {
  Log.info(`All flags: ${JSON.stringify(allFlags)}`)
}