Question: What would be the downside of having two msw server instances running at the same time?

[brianzhou13]

Environment

n/a

Request handlers

So we have individual test files that'll spin up their own msw servers with setupServer:

// test-specific-file.test.ts

const server = setupServer(...handlers);

describe('a test', () => {
    beforeAll(() => server.listen());
    afterEach(() => server.resetHandlers());
    afterAll(() => server.close());
    test('api requests', () => {
      // ...
    })
})

Then in our jest setup, we spin up an additional server before each test run to mock out any accidentally non-mocked routes:

// setupFilesAfterEnv.js

beforeAll(() => globalServer.listen());
afterEach(() => globalServer.resetHandlers());
afterAll(() => globalServer.close());

Actual request

Mostly with Axios, but probably some sprinkles of fetch

Current behavior

The tests work correctly, but the downside of doing so isn't clear to me, and that's what I'm hoping to answer through this GH issue.

Expected behavior

That the tests run properly.

Screenshots

n/a

[kettanaito]

Hey, @brianzhou13.

There's little reason to have multiple instances of setupServer calls. If anything, that will patch the request-issuing modules multiple times (which is how request interception works in Node.js), which may lead to all sorts of unexpected behavior.

Instead, create a single setupServer integration point in your global test setup and modify it per test suite, if necessary. That way you get the same control over your API with a single setup.

import { setupServer } from 'msw'

export const server = setupServer(/* global handlers, if any */)

// setupFilesAfterEnv.js
import { server } from './server'

beforeAll(() => server.listen())
afterEach(() => server.resetHandlers())
afterAll(() => server.close())

// some.test.js
import { msw } from 'msw'
import { server } from './server'

beforeEach(() => {
  server.use(rest.get('/', resolver))
})

Note that if you decide to use .resetHandlers() in the global afterEach hook, then applying a test suite-wide handlers would be better via the beforeEach hook, since all runtime handlers you add via .use() get removed as the result of the afterEach global hook.