I have a REST API and I'm writing TDD for this project. My TDD is consisted of two parts: route and service. I chose to use Jest. I have a MongoDB database that I use for testing. When each test is completed, I reset my database using the afterAll() method. In this method, I run the mongoose.connection.dropDatabase function.
There is no error when I ran only one test file but when I run multiple test files, I get an error. The error message:
MongoError: Cannot create collection auth-db.users - database is in
the process of being dropped.
I share sample codes with you:
users.route.test.ts:
https://gist.github.com/mksglu/8c4c4a3ddcb0e56782725d6457d97a0e
users.service.test.ts:
https://gist.github.com/mksglu/837202c1048687ad33b4d1dee01bd29c
When all my tests run, "sometimes" gives errors. I wrote the above error message. The reason for this error is that the reset process still continues. I can't solve this problem. I'd appreciate it if you could help.
Thanks.
https://jestjs.io/docs/en/cli.html#runinband
What you are looking for is --runInBand command. Which makes jest to run serially instead of creating a worker pool of child processes that run tests
Related
thank you for looking into this!
We are running a quite comprehensive testsuite (some hunderds of test) with the goal to make sure that our tracking implemantation works as expected. We are executing this tests via CI 4x a day. Since a few weeks we have random test aborts, which are unfortunately extremely hard to track and reproduce.
What is the Current behavior?
Errors: Unhandled promise rejection:
Error [ERR_HTTP2_INVALID_SESSION]: The session has been destroyed
at new NodeError (node:internal/errors:371:5)
at ClientHttp2Session.request (node:internal/http2/core:1702:13)
at DestinationRequest._sendRealThroughHttp2 (/home/ec2-user/actions-runner/_work/ds_cerberus/ds_cerberus/node_modules/testcafe-hammerhead/lib/request-pipeline/destination-request/index.js:51:32)
at DestinationRequest._send (/home/ec2-user/actions-runner/_work/ds_cerberus/ds_cerberus/node_modules/testcafe-hammerhead/lib/request-pipeline/destination-request/index.js:110:18)
at runMicrotasks (:null:null)
at processTicksAndRejections (node:internal/process/task_queues:96:5)
For us it looks like a racing condition inside the hammerhead-lib which is used as a proxy for testcafe. But we have no idea how to fix this - or least make sure that not the whole suite of tests is aborted by it.
This is the stacktrace if the tests are aborted. Unfortunately this means that the execution of all tests is aborted and does not only affect a single test, which renders the whole suite for us pretty useless...
Steps to Reproduce
There seems to be more aborts if the tests are executed in the suite and not solely - but even there it is quite hard to identify a pattern.
TestCafe version
"testcafe": "^1.17.1",
Node.js version
node-version: '16.x'
Command-line arguments
testcafe --config-file .testcaferc-dev.json tests
The issue may be related to HTTP/2 requests. So, you can disable it using the following option: disableHttp2. Check if the issue is reproduced after that.
You can also try to increase timeouts, e.g. ajaxRequestTimeout and testExecutionTimeout.
If this does not help, please create a simple project where the issue is reproducible and share it here. We will research it on our side.
I am trying to speed up the test runtime by using pytest-xdist, previously I was using pytest, pytest-django to run the tests,
After installing pytest-xdist, One is an issue which I am facing is most of the tests are failing with messages that relation does not exist
pytest -n auto accounts/tests
for example
psycopg2.errors.UndefinedTable: relation "auth_user" does not exist
I guess there is error which says database does not exist:
django.db.utils.OperationalError: FATAL: database "test_dev" does not exist
But the interesting thing is some of tests are passing :(
The Setup
I have a NodeJS project that uses:
Jest for testing.
Sequelize as an ORM.
Sequelize is instantiated when loading the models module which gets imported by a few of the files that are being tested with Jest.
The Problem
Jest tests pass but then hang with the message:
Jest did not exit one second after the test run has completed.
This usually means that there are asynchronous operations that weren't
stopped in your tests. Consider running Jest with
--detectOpenHandles to troubleshoot this issue.
Note: Adding --detectOpenHandles to the test call does not affect the output.
I do not actually invoke the sequelize object from any test paths, however some of the tested files import the models module and therefore Sequelize is instantiated.
(I will also note that this only occurs on my TravisCI environment, but I suspect this is a red herring.)
The Context
Because Jest is running tests in parallel, the models module is loaded multiple times during the entire test process. I confirmed this with debug output saying SEQUELIZE LOADED which appears multiple times when I run the tests.
The Attempts
I did attempt to invoke sequelize.close() inside of a globalTeardown but this appears to simply open (and then close) a new sequelize connection.
Since none of the tests actually rely on a database connection, I attempted running sequelize.close() within the models module immediately before export. This fixed the issue (though obviously is not a solution).
I have attempted to configure the test connection pools to aggressively end connections.
const sequelizeConfig = {
...
pool: {
idle: 0,
evict: 0,
}
}
This did nothing.
The Requirements
I don't want to use a brute force solution such as running --forceExit via Jest when I run my tests. This feels like it is ignoring the root issue and might expose me to other kinds of mistakes down the line.
My tests are spread across dozens of files, which means needing to invoke something in an afterAllTests would require a lot of redundancy and would introduce a code smell.
The Question
How can I ensure that sequelize connections are closed after tests finish, so they don't cause Jest to hang?
Jest provides a way to create universal setup before every test suite. This means it is possible to leverage Jest's afterAll to close the sequelize connection pool without having to manually include it in every single test suite.
Example Jest config in package.json
"jest": {
...
"setupFilesAfterEnv": ["./src/test/suiteSetup.js"]
}
Example suiteSetup.js
import models from '../server/models'
afterAll(() => models.sequelize.close())
// Note: in my case sequelize is exposed as an attribute of my models module.
Since the setupFilesAfterEnv is loaded before every test suite, this ensures that every Jest thread that opened a connection ultimately has that connection closed.
This doesn't violate DRY and it doesn't rely on a clunky --forceExit.
It does mean that connections will be closed that may never have been opened (which is a bit brute force) but that might be the best option.
When I locally execute my mock test in c# then my test case gets passed but when I queue my build to azure then some of the random test cases fail which was getting passed locally? Any solution on this?
I got to know this was getting failed because my mock object was static. Check my another question Why my few test case get failed after "Running all" controller and pass individually?
When I run unit tests via Ospec for Mithril, I can see if tests fail locally in the console.
What I'm looking for is a solution that will not allow a following Node.js build script to execute if one or more of the tests fail.
I don't want code to be pushed up to another environment/lane if the unit tests aren't passing.
I don't see how to accomplish this in the docs.
In Node, I'm running ospec && someBuildProcess.
The answer might be a Node.js thing, but I'm at a loss for what to look for now.
ospec calls process.exit(1) if any tests fail, and the command string you posted should work. I just verified it locally with the following setup:
https://gist.github.com/tivac/d90c07592e70395639c63dd5100b50a6
ospec runs, fails, and the echo command never gets called.
Can you post some more details about your setup?