k6 is a performance testing tool. This post explains types of performance testing and dives into k6 usage, from configuration to running tests.

Load and stress testing

Load and stress testing are two types of performance testing used to evaluate how well a system performs under various conditions.

Load testing determines how the system performs under expected user loads. The purpose is to identify performance bottlenecks.

Stress testing assesses how the system performs when loads are heavier than usual. The purpose is to find the limit at which the system fails and to observe how it recovers from such failures.

Prerequisites

• k6 installed

• Script (Node.js) file with configuration and execution function

Configuration

Configuration is stored inside options variable, which allows you to set up different testing scenarios:

• constant user load, the number of virtual users (vus) remains constant throughout the test period

export const options = {
  vus: 30,
  duration: '10m'
};

• variable user load, the number of users increases and decreases over time

export const options = {
  stages: [
    {
      duration: '1m',
      target: 30
    },
    {
      duration: '10m',
      target: 30
    },
    {
      duration: '5m',
      target: 0
    }
  ]
};

Environment variables can be passed through the command line and are accessible within the script via the __ENV object.

k6 -e TOKEN=token run script.js

Execution function

This function defines what virtual users will do during the test. This function is called for each virtual user and typically includes steps that simulate user actions on the app.

export default function() {
  http.get(URL);
  // Add more actions as required
}

Test report

k6 generates a report that provides detailed insights into various benchmarks, such as the number of virtual users, requests per second, request durations and error rates.

Example

This example utilizes k6 to conduct a load test using a variable user load approach:

• User simulation: The script ramps up to 1,000 users, maintains that level to simulate sustained traffic, and gradually reduces to zero.

• Request handling: During the test, each virtual user sends a POST request to an API, with pauses between requests to mimic real user behavior.

• Performance insights: After the test, k6 provides a report that shows key information, such as how fast the app responds and how many requests fail.

Run it via k6 -e TOKEN=1234 run script.js command.

// script.js
import { check, sleep } from 'k6';
import { scenario } from 'k6/execution';
import http from 'k6/http';
export const options = {
  stages: [
    // Ramp up to 1000 users over 10 minutes
    {
      duration: '10m',
      target: 1000
    },
    // Hold 1000 users for 30 minutes
    {
      duration: '30m',
      target: 1000
    },
    // Ramp down to 0 users over 5 minutes
    {
      duration: '5m',
      target: 0
    }
  ]
};
export default () => {
  const response = http.post(
    URL,
    JSON.stringify({
      iteration: scenario.iterationInTest
    }),
    {
      headers: {
        Authorization: __ENV.TOKEN,
        'Content-Type': 'application/json'
      }
    }
  );
  check(response, {
    'response status was 200': (res) => res.status === 200
  });
  sleep(1);
};

The demo with the mentioned example is available here.

Node.js version 20 brings a stable test runner so you can run tests inside *.test.js files with node --test command. This post covers the primary usage of it regarding spies and mocking for the unit tests.

Spies are functions that let you spy on the behavior of functions called indirectly by some other code while mocking injects test values into the code during the tests.

mock.method can create spies and mock async, rejected async, sync, chained methods, and external and built-in modules.

• Async function

import assert from 'node:assert/strict';
import { describe, it, mock } from 'node:test';
const calculationService = {
  calculate: () => // implementation
};
describe('mocking resolved value', () => {
  it('should resolve mocked value', async () => {
    const value = 2;
    mock.method(calculationService, 'calculate', async () => value);
    const result = await calculationService.calculate();
    assert.equal(result, value);
  });
});

• Rejected async function

const error = new Error('some error message');
mock.method(calculationService, 'calculate', async () => Promise.reject(error));
await assert.rejects(async () => calculateSomething(calculationService), error);

• Sync function

mock.method(calculationService, 'calculate', () => value);

• Chained methods

mock.method(calculationService, 'get', () => calculationService);
mock.method(calculationService, 'calculate', async () => value);
const result = await calculationService.get().calculate();

• External modules

import axios from 'axios';
mock.method(axios, 'get', async () => ({ data: value }));

• Built-in modules

import fs from 'fs/promises';
mock.method(fs, 'readFile', async () => fileContent);

• Async and sync functions called multiple times can be mocked with different values using context.mock.fn and mockedFunction.mock.mockImplementationOnce.

describe('mocking same method multiple times with different values', () => {
  it('should resolve mocked values', async (context) => {
    const firstValue = 2;
    const secondValue = 3;
    const calculateMock = context.mock.fn(calculationService.calculate);
    calculateMock.mock.mockImplementationOnce(async () => firstValue, 0);
    calculateMock.mock.mockImplementationOnce(async () => secondValue, 1);
    const firstResult = await calculateMock();
    const secondResult = await calculateMock();
    assert.equal(firstResult, firstValue);
    assert.equal(secondResult, secondValue);
  });
});

• To assert called arguments for a spy, use mockedFunction.mock.calls[0] value.

mock.method(calculationService, 'calculate');
await calculateSomething(calculationService, firstValue, secondValue);
const call = calculationService.calculate.mock.calls[0];
assert.deepEqual(call.arguments, [firstValue, secondValue]);

• To assert skipped call for a spy, use mockedFunction.mock.calls.length value.

mock.method(calculationService, 'calculate');
assert.equal(calculationService.calculate.mock.calls.length, 0);

• To assert how many times mocked function is called, use mockedFunction.mock.calls.length value.

mock.method(calculationService, 'calculate');
calculationService.calculate(3);
calculationService.calculate(2);
assert.equal(calculationService.calculate.mock.calls.length, 2);

• To assert called arguments for the exact call when a mocked function is called multiple times, an assertion can be done using mockedFunction.mock.calls[index] and call.arguments values.

const calculateMock = context.mock.fn(calculationService.calculate);
calculateMock.mock.mockImplementationOnce((a) => a + 2, 0);
calculateMock.mock.mockImplementationOnce((a) => a + 3, 1);
calculateMock(firstValue);
calculateMock(secondValue);
[firstValue, secondValue].forEach((argument, index) => {
  const call = calculateMock.mock.calls[index];
  assert.deepEqual(call.arguments, [argument]);
});

Running TypeScript tests

Add a new test script

{
  "type": "module",
  "scripts": {
    "test": "node --test",
    "test:ts": "glob -c \"node --loader tsx --no-warnings --test\" \"./src/**/*.{spec,test}.ts\""
  },
  "devDependencies": {
    // ...
    "glob": "^10.3.1",
    "tsx": "^3.12.7"
  }
}

Demo

The demo with the mentioned examples is available here.

Integration testing means testing a component with multiple sub-components and how they interact. Some sub-components can be external services, databases, and message queues.

External services are running, but their business logic is mocked based on received parameters (request headers, query parameters, etc.). Databases and message queues are spun up using test containers.

This post covers testing service as a component and its API endpoints. This approach can be used with any framework and language. NestJS and Express are used in the examples below.

API endpoints

Below is the controller for two endpoints. First communicates with an external service and retrieves some data based on the sent parameter. The second one retrieves the data from the database.

// users.controller.ts
@Controller('users')
export class UsersController {
  constructor(private userService: UsersService) {}
  @Get()
  async getAll(@Query('type') type: string) {
    return this.userService.findAll(type);
  }
  @Get(':id')
  async getById(@Param('id', new ParseUUIDPipe()) id: string) {
    return this.userService.findById(id);
  }
}

External dependencies

External service is mocked to send data based on the received parameter.

export const createDummyUserServiceServer = async (): Promise<DummyServer> => {
  return createDummyServer((app) => {
    app.get('/users', (req, res) => {
      if (req.query.type !== 'user') {
        return res.status(403).send('User type is not valid');
      }
      res.json(usersResponse);
    });
  });
};

Tests setup

Tests for endpoints can be split into two parts. The first is related to the external dependencies setup.

The example below creates a mocked service and spins up the database using test containers. The environment variables are set for before mentioned dependencies, and the leading service starts running.

The database is cleaned before every test run. External dependencies (mocked service and database) are closed after tests finish.

// test/users.spec.ts
describe('UsersController (integration)', () => {
  let app: INestApplication;
  let dummyUserServiceServerClose: () => void;
  let postgresContainer: StartedTestContainer;
  let usersRepository: Repository<UsersEntity>;
  const databaseConfig = {
    databaseName: 'nestjs-starter-db',
    databaseUsername: 'user',
    databasePassword: 'some-r4ndom-pasS',
    databasePort: 5432,
  }
  beforeAll(async () => {
    const dummyUserServiceServer = await createDummyUserServiceServer();
    dummyUserServiceServerClose = dummyUserServiceServer.close;
    postgresContainer = await new GenericContainer('postgres:15-alpine')
      .withEnvironment({
        POSTGRES_USER: databaseConfig.databaseUsername,
        POSTGRES_PASSWORD: databaseConfig.databasePassword,
        POSTGRES_DB: databaseConfig.databaseName,
      })
      .withExposedPorts(databaseConfig.databasePort)
      .start();
    const moduleFixture: TestingModule = await Test.createTestingModule({
      imports: [AppModule],
    })
      .overrideProvider(ConfigService)
      .useValue({
        get: (key: string): string => {
          const map: Record<string, string | undefined> = process.env;
          map.USER_SERVICE_URL = dummyUserServiceServer.url;
          map.DATABASE_HOSTNAME = postgresContainer.getHost();
          map.DATABASE_PORT = `${postgresContainer.getMappedPort(databaseConfig.databasePort)}`;
          map.DATABASE_NAME = databaseConfig.databaseName;
          map.DATABASE_USERNAME = databaseConfig.databaseUsername;
          map.DATABASE_PASSWORD = databaseConfig.databasePassword;
          return map[key] || '';
        },
      })
      .compile();
    app = moduleFixture.createNestApplication();
    usersRepository = app.get(getRepositoryToken(UsersEntity));
    await app.init();
  });
  beforeEach(async () => {
    await usersRepository.delete({});
  });
  afterAll(async () => {
    await app.close();
    dummyUserServiceServerClose();
    await postgresContainer.stop();
  });
  // ...
});

Tests

The second part covers tests for the implemented endpoints. The first test suite asserts retrieving data from the external service based on the sent type as a query parameter.

// test/users.spec.ts
describe('/users (GET)', () => {
  it('should return list of users', async () => {
    return request(app.getHttpServer())
      .get('/users?type=user')
      .expect(HttpStatus.OK)
      .then((response) => {
        expect(response.body).toEqual(usersResponse);
      });
  });
  it('should throw an error when type is forbidden', async () => {
    return request(app.getHttpServer())
      .get('/users?type=admin')
      .expect(HttpStatus.FORBIDDEN);
  });
});

The second test suite asserts retrieving the data from the database.

// test/users.spec.ts
describe('/users/:id (GET)', () => {
  it('should return found user', async () => {
    const userId = 'b618445a-0089-43d5-b9ca-e6f2fc29a11d';
    const userDetails = {
      id: userId,
      firstName: 'tester',
    };
    const newUser = await usersRepository.create(userDetails);
    await usersRepository.save(newUser);
    return request(app.getHttpServer())
      .get(`/users/${userId}`)
      .expect(HttpStatus.OK)
      .then((response) => {
        expect(response.body).toEqual(userDetails);
      });
  });
  it('should return 404 error when user is not found', async () => {
    const userId = 'b618445a-0089-43d5-b9ca-e6f2fc29a11d';
    return request(app.getHttpServer())
      .get(`/users/${userId}`)
      .expect(HttpStatus.NOT_FOUND);
  });
});

Boilerplate

Here is the link to the boilerplate I use for the development. It contains the examples mentioned above with more details.

Custom repositories extend the base repository class and enrich it with several additional methods. This post covers the unit and integration testing.

// user.repository.ts
@Injectable()
export class UserRepository extends Repository<UserEntity> {
  constructor(private dataSource: DataSource) {
    super(UserEntity, dataSource.createEntityManager());
  }
  async getById(id: string) {
    return this.findOne({ where: { id } });
  }
  // ...
}

Setup

Inject a custom repository into the service.

// user.service.ts
export class UserService {
  constructor(private readonly userRepository: UserRepository) {}
  async getById(id: string): Promise<User> {
    return this.userRepository.getById(id);
  }
  // ...
}

Pass entity class to the forFeature method.

// user.module.ts
@Module({
  imports: [
    TypeOrmModule.forFeature([UserEntity])],
    // ...
  ],
  providers: [UserService, UserRepository],
  // ...
})
export class UserModule {}

Unit testing

To properly unit-test the custom repository, mock some methods.

// user.repository.spec.ts
describe('UserRepository', () => {
  let userRepository: UserRepository;
  const dataSource = {
    createEntityManager: jest.fn()
  };
  beforeEach(async () => {
    const module: TestingModule = await Test.createTestingModule({
      providers: [
        UserRepository,
        {
          provide: DataSource,
          useValue: dataSource
        }
      ]
    }).compile();
    userRepository = module.get<UserRepository>(UserRepository);
  });
  describe('getById', () => {
    it('should return found user', async () => {
      const id = 'id';
      const user = {
        id
      };
      const findOneSpy = jest
        .spyOn(userRepository, 'findOne')
        .mockResolvedValue(user as UserEntity);
      const foundUser = await userRepository.getById(id);
      expect(foundUser).toEqual(user);
      expect(findOneSpy).toHaveBeenCalledWith({ where: user });
    });
  });
});

Integration testing

Integration testing is more suitable when working with databases. Read more about it on Integration testing Node.js apps post

Boilerplate

Here is the link to the boilerplate I use for the development. It contains the examples mentioned above with more details.

Besides asserting the output of the function call, unit testing includes the usage of spies and mocking. Spies are functions that let you spy on the behavior of functions called indirectly by some other code. Spy can be created by using jest.fn(). Mocking injects test values into the code during the tests. Some of the use cases will be presented below.

• Async function and its resolved value can be mocked using mockResolvedValue. Another way to mock it is by using mockImplementation and providing a function as an argument.

const calculationService = {
  calculate: jest.fn()
};
jest.spyOn(calculationService, 'calculate').mockResolvedValue(value);
jest
  .spyOn(calculationService, 'calculate')
  .mockImplementation(async (a) => Promise.resolve(a));

• Rejected async function can be mocked using mockRejectedValue and mockImplementation.

jest
  .spyOn(calculationService, 'calculate')
  .mockRejectedValue(new Error(errorMessage));
jest
  .spyOn(calculationService, 'calculate')
  .mockImplementation(async () => Promise.reject(new Error(errorMessage)));
await expect(calculateSomething(calculationService)).rejects.toThrowError(
  Error
);

• Sync function and its return value can be mocked using mockReturnValue and mockImplementation.

jest.spyOn(calculationService, 'calculate').mockReturnValue(value);
jest.spyOn(calculationService, 'calculate').mockImplementation((a) => a);

• Chained methods can be mocked using mockReturnThis.

// calculationService.get().calculate();
jest.spyOn(calculationService, 'get').mockReturnThis();

• Async and sync functions called multiple times can be mocked with different values using mockResolvedValueOnce and mockReturnValueOnce, respectively, and mockImplementationOnce.

jest
  .spyOn(calculationService, 'calculate')
  .mockResolvedValueOnce(value)
  .mockResolvedValueOnce(otherValue);
jest
  .spyOn(calculationService, 'calculate')
  .mockReturnValueOnce(value)
  .mockReturnValueOnce(otherValue);
jest
  .spyOn(calculationService, 'calculate')
  .mockImplementationOnce((a) => a + 3)
  .mockImplementationOnce((a) => a + 5);

• External modules can be mocked similarly to spies. For the following example, let's suppose axios package is already used in one function. The following example represents a test file where axios is mocked using jest.mock().

import axios from 'axios';
jest.mock('axios');
// within test case
axios.get.mockResolvedValue(data);

• Manual mocks are resolved by writing corresponding modules in __mocks__ directory, e.g., fs/promises mock will be stored in __mocks__/fs/promises.js file. fs/promises mock will be resolved using jest.mock() in the test file.

jest.mock('fs/promises');

• To assert called arguments for a mocked function, an assertion can be done using toHaveBeenCalledWith matcher.

const spy = jest.spyOn(calculationService, 'calculate');
expect(spy).toHaveBeenCalledWith(firstArgument, secondArgument);

• To assert skipped call for a mocked function, an assertion can be done using not.toHaveBeenCalled matcher.

const spy = jest.spyOn(calculationService, 'calculate');
expect(spy).not.toHaveBeenCalled();

• To assert how many times mocked function is called, an assertion can be done using toHaveBeenCalledTimes matcher.

const spy = jest.spyOn(calculationService, 'calculate');
calculationService.calculate(3);
calculationService.calculate(2);
expect(spy).toHaveBeenCalledTimes(2);

• To assert called arguments for the exact call when a mocked function is called multiple times, an assertion can be done using toHaveBeenNthCalledWith matcher.

const argumentsList = [0, 1];
argumentsList.forEach((argument, index) => {
  expect(calculationService.calculate).toHaveBeenNthCalledWith(
    index + 1,
    argument
  );
});

• Methods should be restored to their initial implementation before each test case.

// package.json
"jest": {
  // ...
  "restoreMocks": true
}
// ...

Demo

The demo with the mentioned examples is available here.