Duplicate VITEST_POOL_ID

[flunderpero]

Describe the bug

We see duplicate VITEST_POOL_ID in our test suite.

The following output is created by logging in beforeAll and afterAll like this:
console.log("VITEST_POOL_ID", process.env.VITEST_POOL_ID, process.env.VITEST_WORKER_ID, performance.now())

The output:

VITEST_POOL_ID BEFORE 5 5 1879.1254999637604
VITEST_POOL_ID BEFORE 1 1 1865.7593338489532
VITEST_POOL_ID BEFORE 7 7 1861.8154590129852
VITEST_POOL_ID BEFORE 8 8 1861.237958908081
VITEST_POOL_ID AFTER 7 7 2069.352292060852
VITEST_POOL_ID AFTER 8 8 2106.785749912262
VITEST_POOL_ID BEFORE 1 9 2140.8526248931885
VITEST_POOL_ID BEFORE 1 10 2357.5155839920044
VITEST_POOL_ID BEFORE 1 11 2473.89066696167
VITEST_POOL_ID AFTER 1 1 2993.181624889374
VITEST_POOL_ID BEFORE 3 16 3702.6203339099884
VITEST_POOL_ID AFTER 3 16 3935.632458925247
VITEST_POOL_ID BEFORE 1 17 3931.5376670360565
VITEST_POOL_ID AFTER 1 17 3970.2642500400543
VITEST_POOL_ID BEFORE 1 18 4274.482708930969
VITEST_POOL_ID AFTER 1 18 4356.533874988556
VITEST_POOL_ID AFTER 1 10 10438.08066701889
VITEST_POOL_ID AFTER 1 9 13164.534374952316
VITEST_POOL_ID AFTER 1 11 13179.464833974838

You can clearly see in line 7 to 9 that VITEST_POOL_ID of 1 is used in parallel.

Our vite.config.ts:

export default defineConfig({
    test: {
        setupFiles: ["packages/lib.shared.test/test_setup.ts"],
        threads: true,
        testTimeout: 8000,
        hookTimeout: 10000,
        minThreads: 8,
        maxThreads: 8,
        coverage: false,
    },
    esbuild: {
        target: "node16",
    },
})

We ever only see VITEST_POOL_ID of 1 being used in parallel. I checked that VITEST_POOL_ID matches import {workerId} from "titnypool".

Reproduction

Currently I am trying to reproduce the error in tinypool, but did not manage to do so.

System Info

System:
    OS: macOS 12.4
    CPU: (10) arm64 Apple M1 Max
    Memory: 8.70 GB / 64.00 GB
    Shell: 5.8.1 - /bin/zsh
  Binaries:
    Node: 16.15.1 - ~/.nvm/versions/node/v16.15.1/bin/node
    Yarn: 3.1.0 - ~/.nvm/versions/node/v16.15.1/bin/yarn
    npm: 8.11.0 - ~/.nvm/versions/node/v16.15.1/bin/npm
  Browsers:
    Chrome: 103.0.5060.53
    Firefox: 100.0.2
    Safari: 15.5
  npmPackages:
    vitest: 0.16.0 => 0.16.0

Used Package Manager

yarn

Validations

• Follow our Code of Conduct
• Read the Contributing Guidelines.
• Read the docs.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.
• Check that this is a concrete bug. For Q&A open a GitHub Discussion or join our Discord Chat Server.
• The provided reproduction is a minimal reproducible example of the bug.

[Aslemammad]

Since there is no reproduction, What I recommend is trying to debug your app with ndb and see if you can find a solution!

[Lyokolux]

Hey ! I am currently running (integration) test suits against multiple sqlite databases at the same time. Thus the tests are concurrent depending of the core of the machine.

I duplicate the original one and assign them different ids. I used first VITEST_POOL_ID but the ids have clashed.

I concluded VITEST_POOL_ID is not unique. However I used VITEST_WORKER_ID and it worked like a charm.

I will provide a reproduction :) Give me some time.

My strategy for a reproduction:

• writing multiple test suits with 1-3 dummy tests in it
• creating a file named VITEST_POOL_ID for each test suits (in the setupFile)
• check if there is some duplicates

[Lyokolux]

Here the reproduction based on a Nuxt 3 template: https://stackblitz.com/edit/nuxt-starter-agwwgz?file=test%2FsetupIntegration.ts,vitest.config.ts

Workflow:

• Run yarn test → we see files created in a pool-ids directory
• Clean directory (to run it again): yarn test:clean

What we observe?

• pool-ids directory contains the generated file, 2 for each test suit / test file.
  • pool-x-yyyyyy... with x the VITEST_POOL_ID
  • worker-x-yyyyyy...with x the VITEST_WORKER_ID

Remarks:

• there is multiple threads and it runs in isolated "mode". Check vitest.config.ts to be sure of it.

Constat:

• VITEST_POOL_ID is repeated
• VITEST_WORKER_ID seems unique (work in this case and mine in my project, proof need to be made)

[sheremet-va]

VITEST_WORKER_ID seems unique (work in this case and mine in my project, proof need to be made)

VITEST_WORKER_ID is always unique, and it will increase with each watcher trigger.

[Lyokolux]

VITEST_WORKER_ID seems unique (work in this case and mine in my project, proof need to be made)

VITEST_WORKER_ID is always unique, and it will increase with each watcher trigger.

Ok also the point in the documentation is unclear about that in https://vitest.dev/config/#setupfiles:

You can use process.env.VITEST_POOL_ID (integer-like string) inside to distinguish between threads (will always be '1', if run with threads: false).

VITEST_POOL_ID is then not unique, on the contrary of VITEST_WORKER_ID. It could be useful to mention that uniqueness of VITEST_WORKER_ID.

Also "We see duplicate VITEST_POOL_ID in our test suite." is an expected behavior.
However, "We see duplicate VITEST_POOL_ID running in parallel in our test suite." is not.

To get a unique ID for each thread, VITEST_WORKER_ID should be used.
Right @flunderpero ?

[sheremet-va]

No, it is not expected behaviour. VITEST_WORKER_ID is generated by Vitest whenever we fire up a worker. VITEST_POOL_ID should be unique while running tests (it's a queue, when one worker is removed, we can reuse it's ID). See: #1469

[Aslemammad]

I created a test and I guess I fixed it, it's a challenging issue! But one thing, @Lyokolux tests were helpful but they don't test the behaviour, because there's a misunderstanding there, and it's that VITEST_POOL_ID can be repeated, no problem with that, but it cannot be repeated at the same two running workers or more!

I'll release a new tinypool version soon!

[Aslemammad]

I expect this to be resolved in tinypool@^0.2.3

cc @flunderpero

[Aslemammad]

feel free to re-open in case you faced the issue again!