tachyon-event-reporter v32.0.0

Tachyon Event Reporter

Provides a library for reporting events to Spade as well as other data sinks, like the console, or whatever adapter matches the interface. Pairs nicely with tachyon-event-tracker.

Spade reporting features include:

• Event batching
• Beacon use when possible to ensure event delivery even during unload

Installation

$ yarn add tachyon-event-reporter

How To Use

Function Interface

import {
  createConsoleReportEvent,
  createMultiReportEvent,
  createSpadeReportEvent,
} from 'tachyon-event-reporter';

const spadeURI = ... // can be pulled from dynamic settings or elsewhere

// in client/universal code somewhere
function createReportEvent() {
  return createMultiReportEvent([
    createConsoleReportEvent(),
    createSpadeReportEvent(spadeURI),
  ]);
}

// Silly example of reporting
const reportEvent = createReportEvent();
reportEvent({
  event: spadeReporter,
  properties: { spadeReporterStatus: 'initialized' },
});

Component Interface

Example that turns on spade debugger. Switch debugSpade to false and set a Spade URI to use production spade.

import { EventReporterRoot, SPADE_ROTATOR_URI, useEventReporter } from 'tachyon-event-reporter';

const SomeComponent = () => {
  const { reportEvent } = useEventReporter();
  return (
    <button click={() => reportEvent(/* event stuff here */)} />,
  );
};

const root = (
  <EventReporterRoot
    console
    spade={{ spadeUrl: '...' // use SPADE_ROTATOR_URI to fetch the current Spade reporting URL }
  >
    <SomeComponent />
  </EventReporterRoot>
);

Service Dependencies and Degradation

This section explains how tachyon-spade-reporter handles degraded availability or outages with upstream services. The Tachyon Spade Reporter package makes use of two services:

• The spade ingest URL provider (Tier 2?, unclear).
• Spade ingest service (Tier 2).

Spade Ingest URL Provider

The package will attempt to periodically fetch the URL to which the client should send spade events. If this attempt fails then it will fall back to the last known-good URL that was fetched, if none exists it will fall back to the default.

Spade Ingest Service

This package follows a fire-and-forget strategy for spade events. This is in accordance with advice from Data Science when the implementation was written. We have a pending ticket to update the strategy to follow the retry logic the core data science team now recommends.

Flume Reporting

Flume is a reporting flow parallel to Spade reporting that uses application services to populate all events to CloudWatch. Flume is a replacement for a Kinesis sub-stream of Spade data, avoiding extra infra while resulting in the same outcome: client event data available in CloudWatch streams.

Usage

To activate it, pass the required config (flumeUrl and allowedProperties) to the flume prop on EventReporterRoot (or create your own with createFlumeReportEvent). To prevent accidentally logging PII, the allowedProperties list is used to filter for specific event fields. Once activated, the reporting framework will mirror all events to the flume endpoint of your choosing.

+------------+  3. Process/Log  +------------+
| CloudWatch | <--------------- | App Server | <+
+------------+                  +------------+  |
                                |               |
                                | 1. SSR        | 2. POST /flume/handler/path
                                v               |
                                +------------+  |
                                | Web Client | -+
                                +------------+

Serverside Configuration

A flume reporting endpoint must accept POST requests with support for url-encoded form request bodies. The event payload will be under the data key, and can simply be sent to the logFlume function to have the server properly log the event in a CloudWatch-consumable way.

Once this route is configured and logging, events will be available in CloudWatch logs under the normal tachyon-logger format where message is equal to the event name (e.g. pageview) and context is equal to the filtered event properties (e.g. location). In dev mode, the server logs will show these messages as well, so that can be used as a reference to the event shapes.

In Tachyon

In a Tachyon app, it is recommended to use the flumeHandler function from tacyhon-page-utils as a drop-in route handler:

// .../pages/api/my/flume/path.ts
export { flumeHandler as default } from 'tachyon-page-utils';

Furthermore, it is recommended to use an absolute path (e.g. /api/flume) in the flume config for portability between environments.