services-telemetry

@fluidframework/server-services-telemetry

Fluid server package containing telemetry utilities used across Fluid service code.

See GitHub for more details on the Fluid Framework and packages within.

Overview

The services-telemetry package is built around 3 main components:

1. A Lumberjack class: the main telemetry manager. Lumberjack is the telemetry interface to be used when instrumenting the code. It is a singleton responsible for managing a list of “engines” that implement the app-specific telemetry logic. It can be used to initialize a metric/data instance (which we call Lumber). It also supports the logging functionality.
2. An ILumberjackEngine interface: defining the requirements for an implementation of the engines employed by Lumberjack.
3. Lumber: the self-contained class representing a metric or event. Lumber should be created through Lumberjack, which will provide Lumber with the list of LumberjackEngines that define how to handle the data. After being created, Lumber can be used to add properties to a property bag and can be finally completed either as a successful or failed operation, while keeping track of duration of the event.

The idea behind Lumberjack is convenience, available through a globally accessible telemetry tool. It is customizable through the app-specific ILumberjackEngine implementations and is capable of handling 2 types of telemetry data. Metrics represent events or operations that can usually be associated with a result of success or failure, that can be measured in terms of duration and that can generally be used in the context of service monitoring/alerting. Logs, on the other hand, should be used in situations where there is no event being evaluated; in other words, where there is not an associated concept of success or failure, duration, etc.

At the start of each Node.js process, Lumberjack should be initialized so that it can be used in singleton style throughout the code. However, it is also possible to create individual instances of Lumberjack at any time. Lumberjack must be initialized with a mandatory ILumberjackEngine[] parameter (it uses each ILumberjackEngine in the list to emit the telemetry data according to the ILumberjackEngine logic). Lumberjack can also take an optional ILumberjackSchemaValidator, which defines the requirements for mandatory parameters and values associated with each Lumber instance.

How to use

In these examples, we will be using WinstonLumberjackEngine as our ILumberjackEngine sample implementation.

Initializing Lumberjack

// Singleton style, at the start of each process
const lumberjackEngine = new WinstonLumberjackEngine();
Lumberjack.setup([lumberjackEngine]);

// Instance style
const lumberjackEngine = new WinstonLumberjackEngine();
const customInstance = Lumberjack.createInstance([lumberjackEngine]);

In FluidFramework Server code, Lumberjack initialization is taken care of automatically thanks to configureLogging() in the server-services-utils package.

Metrics

Metrics are associated with types of events, and each event type should have a unique name. The list of Events is kept in the LumberEventName enum. Lumberjack also supports simple string event names. Example:

const lumberJackMetric = Lumberjack.newLumberMetric(LumberEventName.DeliHandler);

lumberJackMetric.setProperties({
    [BaseTelemetryProperties.tenantId]: this.tenantId,
    [BaseTelemetryProperties.documentId]: this.documentId,
});

try {
    ... // Event being tracked
    lumberJackMetric.success("Success message!");
}
catch (error) {
    lumberJackMetric.error("Error message :(", error);
}

Logs

Lumberjack provides a static method for logging. Example:

const props = {
    property1: "prop1",
    property2: "prop2",
};

Lumberjack.log("Sample message", LogLevel.Info, properties);