Skip to main content

Experimental Features

Overview​

At any given time, there are experimental features within FluidFramework that are being actively developed/tested. In order to allow developers to adopt and test these features in a controlled environment, we introduced feature gates. Feature gates are user defined values that allow experimental features to be easily enabled, and disabled if they cause unexpected behavior. This document will provide details on how to use feature gates to enable experimental features.

warning

Experimental features are by definition not fully supported and should not be used in production applications. Moreover, experimental features may be added, modified, or removed without warning in any minor or major release.

API​

To enable experimental features, you will be required to provide an implementation of IConfigProviderBase. Please note the following about IConfigProviderBase:

  • IConfigProviderBase will accept input of a map-like object. The keys must be type string with the exact names of the features you want to enable. The values must be of type ConfigType, which is defined as a union of primitive types and arrays of primitive types.
  • All experimental features will default to a disabled/neutral state.

Observability​

When a container is loaded, there will be several events logged to the provided logger. One of these events is "fluid:telemetry:ContainerLoadStats". Within this event, there is a property "featureGates", which will contain each of the experimental features (and the corresponding ConfigType values) that the container was loaded with.

Usage​

The following is an example of how to enable experimental features with AzureClient.

  1. First, implement IConfigProviderBase. For example:

    const configProvider = (settings: Record<string, ConfigTypes>): IConfigProviderBase => ({
      getRawConfig: (name: string): ConfigTypes => settings[name]
    });

  2. Next, define the experimental features you want to enable. You will need to know the exact string name of the feature and any corresponding ConfigType values. As mentioned above, ConfigType only supports primitive types and arrays of primitive types.

    const featureGates = {
      "Fluid.ContainerRuntime.ExampleFeature1": true,
      "Fluid.ContainerRuntime.ExampleFeature2": ["exampleConfig1", "exampleConfig2"],
    };

  3. Finally, call the previously defined configProvider function with featureGates as the input, and pass the result in the AzureClient constructor.

    const azureClient = new AzureClient({
      connection: connectionProps,
      logger: myLogger,
      configProvider: configProvider(featureGates),
    });