kibana/packages/kbn-apm-synthtrace

@kbn/apm-synthtrace

@kbn/apm-synthtrace is a tool in technical preview to generate synthetic APM data. It is intended to be used for development and testing of the Elastic APM app in Kibana.

At a high-level, the module works by modeling APM events/metricsets with a fluent API. The models can then be serialized and converted to Elasticsearch documents. In the future we might support APM Server as an output as well.

Usage

This section assumes that you've installed Kibana's dependencies by running yarn kbn bootstrap in the repository's root folder.

This library can currently be used in two ways:

• Imported as a Node.js module, for instance to be used in Kibana's functional test suite.
• With a command line interface, to index data based on a specified scenario.

Using the Node.js module

Concepts

• Service: a logical grouping for a monitored service. A Service object contains fields like service.name, service.environment and agent.name.
• Instance: a single instance of a monitored service. E.g., the workload for a monitored service might be spread across multiple containers. An Instance object contains fields like service.node.name and container.id.
• Timerange: an object that will return an array of timestamps based on an interval and a rate. These timestamps can be used to generate events/metricsets.
• Transaction, Span, APMError and Metricset: events/metricsets that occur on an instance. For more background, see the explanation of the APM data model

Example

import { service, timerange, toElasticsearchOutput } from '@kbn/apm-synthtrace';

const instance = service({ name: 'synth-go', environment: 'production', agentName: 'go' }).instance(
  'instance-a'
);

const from = new Date('2021-01-01T12:00:00.000Z').getTime();
const to = new Date('2021-01-01T12:00:00.000Z').getTime();

const traceEvents = timerange(from, to)
  .interval('1m')
  .rate(10)
  .flatMap((timestamp) =>
    instance
      .transaction({ transactionName: 'GET /api/product/list' })
      .timestamp(timestamp)
      .duration(1000)
      .success()
      .children(
        instance
          .span('GET apm-*/_search', 'db', 'elasticsearch')
          .timestamp(timestamp + 50)
          .duration(900)
          .destination('elasticsearch')
          .success()
      )
      .serialize()
  );

const metricsets = timerange(from, to)
  .interval('30s')
  .rate(1)
  .flatMap((timestamp) =>
    instance
      .appMetrics({
        'system.memory.actual.free': 800,
        'system.memory.total': 1000,
        'system.cpu.total.norm.pct': 0.6,
        'system.process.cpu.total.norm.pct': 0.7,
      })
      .timestamp(timestamp)
      .serialize()
  );

const esEvents = toElasticsearchOutput(traceEvents.concat(metricsets));

Generating metricsets

@kbn/apm-synthtrace can also automatically generate transaction metrics, span destination metrics and transaction breakdown metrics based on the generated trace events. If we expand on the previous example:

import {
  getTransactionMetrics,
  getSpanDestinationMetrics,
  getBreakdownMetrics,
} from '@kbn/apm-synthtrace';

const esEvents = toElasticsearchOutput([
  ...traceEvents,
  ...getTransactionMetrics(traceEvents),
  ...getSpanDestinationMetrics(traceEvents),
  ...getBreakdownMetrics(traceEvents),
]);

CLI

Via the CLI, you can run scenarios, either using a fixed time range or continuously generating data. Scenarios are available in packages/kbn-apm-synthtrace/src/scenarios/.

For live data ingestion:

node scripts/synthtrace simple_trace.ts --target=http://admin:changeme@localhost:9200 --live

For a fixed time window:

node scripts/synthtrace simple_trace.ts --target=http://admin:changeme@localhost:9200 --from=now-24h --to=now

The script will try to automatically find bootstrapped APM indices. If these indices do not exist, the script will exit with an error. It will not bootstrap the indices itself.

The following options are supported:

Connection options

Option	Type	Default	Description
--target	[string]		Elasticsearch target
--kibana	[string]		Kibana target, used to bootstrap datastreams/mappings/templates/settings
--versionOverride	[string]		String to be used for observer.version. Defauls to the version of the installed package.

Note:

• If --target is not set, Synthtrace will try to detect a locally running Elasticsearch and Kibana.
• For Elastic Cloud urls, --target will be used to infer the location of the Cloud instance of Kibana.
• The latest version of the APM integration will automatically be installed and used for observer.version when ingesting APM data. In some cases, you'll want to use --versionOverride to set observer.version explicitly.

Scenario options

Option	Type	Default	Description
--from	[date]	now()	The start of the time window
--to	[date]		The end of the time window
--live	[boolean]		Generate and index data continuously
--scenarioOpts			Raw options specific to the scenario

Note:

• The default --to is 15m.
• You can combine --from and --to with --live to back-fill some data.
• To specify --scenarioOpts you need to use yargs Objects syntax. (e.g. --scenarioOpts.myOption=myValue)

Setup options

Option	Type	Default	Description
--clean	[boolean]	false	Clean APM data before indexing new data
--workers	[number]		Amount of Node.js worker threads
--logLevel	[enum]	info	Log level

Testing

Run the Jest tests:

node scripts/jest --config ./packages/kbn-apm-synthtrace/jest.config.js

Typescript

Run the type checker:

node scripts/type_check.js --project packages/kbn-apm-synthtrace/tsconfig.json