gen3 integration tests - run by https://jenkins.planx-pla.net/ via a Jenkinsfile pipeline in each github repo
- docker-compose (see https://docs.docker.com/compose/install/)
- gnupg2 pass
sudo apt install gnupg2 pass
For using linux/amd64 selenium images:
# start test infra
docker compose -f docker-compose-test-infra.yaml up -d
For using arm64 selenium images:
docker compose -f docker-compose-test-infra-arm64.yaml up -d
# create database
curl -i -XPOST http://localhost:8086/query --data-urlencode "q=CREATE DATABASE ci_metrics"
Configure the grafana dashboard by importing grafana_test_observability.json. This is a pre-configured dashboard, though the data source needs to be selected while importing.
Edit your /etc/hosts file to point the selenium-hub and influxdb host to your Docker container running on localhost.
/etc/hosts
127.0.0.1 selenium-hub
127.0.0.1 influxdb
More information about Selenium Hub in cloud-automation's documentation
Run a test locally against a dev environment that you can ssh to and run gen3 commands like this:
RUNNING_LOCAL=true NAMESPACE=yourDevNamespace PORTAL_SUFFIX="" TEST_DATA_PATH=./testData npm test -- --verbose --grep '@dataClientCLI|@reqGoogle' --invert --reporter mocha-multi suites/.../myTest.js
Some of our test suites require access to the commons backend, but most of our tests can run against an arbitrary commons given an access token that can interact with the gen3 api.
The getAccessToken.sh script is a little helper that can retrieve an access token given an API key.
- First - generate some test data (if necessary for the test) with the appropriate dictionary (see below)
- Next - generate an access token (it will expire in 20 minutes) using an API key from the target commons
mkdir -p output
export GEN3_TOKEN_MAIN=$(bash ./getAccessToken.sh ~/Secrets/gen3/qa-brain.planx-pla.net/credentials.json)
- Finally - run some tests
GEN3_SKIP_PROJ_SETUP=true GEN3_COMMONS_HOSTNAME=qa-brain.planx-pla.net npm test -- --verbose suites/portal/roleBasedUITest.js | tee output/roleBasedResults.txt
GEN3_SKIP_PROJ_SETUP=true GEN3_COMMONS_HOSTNAME=qa-brain.planx-pla.net TEST_DATA_PATH=./testData npm test -- --verbose suites/sheepdogAndPeregrine/submitAndQueryNodesTest.js | tee output/sheepdogPeregrineResults.txt
Interactive test plans are our way of managing manual test plans. An interactive test presents a manual tester with instructions, collects the results of each test, and generates a report.
mkdir -p output
GEN3_SKIP_PROJ_SETUP=true GEN3_COMMONS_HOSTNAME=qa-brain.planx-pla.net npm test -- --verbose suites/portal/roleBasedUITest.js | tee output/roleBasedResults.txt
- the OAuth flow tests require
fence-config.yamlto be configured withMOCK_GOOGLE_AUTH: true, since Google doesn't trust robots - set the GEN3_INTERACTIVE environment variable to
falseto disable interactive tests that require user feedback. See the executable test plans overview for more details.
export TEST_DATA_PATH="$(pwd)/testData"
mkdir -p "$TEST_DATA_PATH"
export TEST_DICTIONARY=https://s3.amazonaws.com/dictionary-artifacts/datadictionary/develop/schema.json
docker run -it -v "${TEST_DATA_PATH}:/mnt/data" --rm --name=dsim --entrypoint=data-simulator quay.io/cdis/data-simulator:master simulate --url "${TEST_DICTIONARY}" --path /mnt/data --program jnkns --project jenkins --max_samples 10
docker run -it -v "${TEST_DATA_PATH}:/mnt/data" --rm --name=dsim --entrypoint=data-simulator quay.io/cdis/data-simulator:master submission_order --url "${TEST_DICTIONARY}" --path /mnt/data --node_name submitted_unaligned_reads
RUNNING_LOCAL=true NAMESPACE=reuben npm test -- --verbose --grep '@dataClientCLI|@reqGoogle' --invert suites/sheepdogAndPeregrine/submitAndQueryNodesTest.js
Note that the --grep ... --invert ... is necessary to disable
setup for "google" tests that require a custom test environment.
After running the test suite several times (or once) you may need to clear out the
databases in your dev environment, so that old data does not interfere with new test runs. The gen3 reset command automates the process to reset the fence, indexd, and sheepdog databases, but should only be run against a personal copy of the database (most of our dev environments point at a personal database schema). gen3 reset also deletes and recreates all deployments and secrets, so you will need to run usersync again too (gen3 job run usersync).
# change linux to osx on mac
ssh -t [email protected] bash -c 'source ~/.bashrc; source cloud-automation/gen3/gen3setup.sh; gen3 aws s3 cp s3://cdis-dc-builds/master/dataclient_linux.zip dataclient.zip'
scp [email protected]:dataclient.zip ~/dataclient.zip
(cd ~/ && unzip ~/dataclient.zip)
chmod a+rx ~/gen3-client
- a test suite goes in
suites/../test.js test.jsimports a tasks and a questions module- basic test structure:
run some task
ask questions about whether the task result matches expectations
This is the repository for managing integration tests used for Gen3 data commons.
Frameworks/Tools:
- CodeceptJS: primary testing framework (docs)
- Chai: additional assertion writing (docs)
- Jenkins: CI, automated builds and tests
The testsuite is run either in Jenkins, when testing a build, or on a local machine, when developing tests. The setup is slightly different for each and is explained below.
Tests require access tokens, usernames, passwords, etc., which we store in environment variables.
The environment variable NAMESPACE must be defined in your shell running the tests. This variable defines the data commons to run the test on. Note that this variable must be a variable in your shell, not in your auto-qa-config file (explained below).
The process of fetching most of the commons variables is automated by the script set for bootstrap in the CodeceptJS configuration file.
If there are things which cannot/should not be fetched in the commons by the setup script, you can setup a JSON config file named auto-qa-config.json in a .gen3 folder in your home directory, ~/.gen3/auto-qa-config.json. For example:
# homeDirectory/.gen3/auto-qa-config.json
{
SPECIAL_SECRET: '*******',
MY_VAR: {
myString: 'abcd123'
myNum: 12
}
}
In setup, these values will be read and exported to environment variables before running the test. So in the example above MY_VAR is actually going to be a string that will need to be parsed when it's used. Note that the setup gives environment variables higher priority over the config file. So if you did export MY_VAL="rock" in the shell you're running the test and in your config had { MY_VAL: "scissors" }, when the test is run, process.env.MY_VAL === "rock".
After cloning this repo, get the required packages by running npm install.
Once you have your environment variables configured and the Selenium server is running, you should be able to successfully run the tests.
If you want to run gen3-qa against a dev environment, you just need to set the environment variable NAMESPACE={dev env namespace} and the GOOGLE_APP_CREDS_JSON to the credentials getting from Google
(export GOOGLE_APP_CREDS_JSON="$(cat creds.json)"), then run ./run-tests-local.sh.
But as mentioned above, some tests have special requirements so you may not want to run them. Instead, you can run a selection of tests that have certain tag by altering the line in run-tests-local.sh file npm test -- --grep "@MyTag" (see info about tags in the Writing Tests section).
The special dev environment setup that is required by some tests is described here.
Each API or web page feature is contained in a singe .js file. They are stored in suites/** filenames matching the pattern *Test.js.
Refer to the CodeceptJS docs existing tests for info and writing patterns. TLDR for the basics:
- A
*Test.jsfile tests a singleFeature - A
Featureis a collection ofScenarios - A
Scenariois a single use case/flow of a feature (e.g. a flow with good input, a flow with bad input)
Here's what it looks like in code:
// in myFeatureTest.js
Feature('MyFeature');
Scenario('use feature good input', () => {
// do something with good input and expect success!
});
Scenario('use feature bad input', () => {
// do something with bad input and expect some error!
});Scenarios are where we actually do things and ask questions about the results. So what does the 'doing' and the 'asking'? That's the job of a service!
A service directly translates to a single gen3 API or web page (e.g. sheepdog API or dictionary web page). They allow you to perform actions and ask about the state of it's subject. Rather than making direct API calls or clicking around on a page within a test, we create a service to abstract away the details (e.g. specific endpoints, css locators for HTML elements, etc) for doing a task or making an assertion about state.
services are located in the directories service/api and service/portal. To include a service in a Scenario, see Using services in Scenarios below.
Services have the following attributes
| attribute | description | verb/accessor | example |
|---|---|---|---|
| props | collection of properties about the service/page (e.g. endpoints, or page element locators) | props | homepage.props.chart |
| tasks | performs a single action of the service/page | do | sheepdog.do.addNode(aliquotNode) |
| questions | makes assertions about the service/page | ask | homepage.ask.seeProp(homepage.props.chart) |
| sequences | sequence of multiple tasks and questions | complete | sheepdog.complete.addNodes(allNodesList) |
A task is usually composed of calls to CodeceptJS's actor(), which is referenced as I. CodeceptJS's actor gives us easy access to things like navigating webpages and making REST API calls (e.g. I.click('.my-button') and I.sendGetRequest('/user/user')). Get familiar with the CodeceptJS helpers and their functions:
- WebdriverIO (web page stuff)
- REST (API stuff)
A question is a method that makes an assertion. For web pages, they are usually written using CodeceptJS's actor() (e.g. I.seeNumberOfElements('.header1', 2)). For everything that we can't use CodeceptJS's actor for (API assertions) we use Chai. Checkout the Chai expect documentation here.
A service's props is just a collection of values that are used when performing tasks and asking questions. This would include:
- API endpoints
- Expected API results
- Web page URL
- Locators for HTML elements
A sequence is just a collection of tasks and questions. For example, if you had a series of tasks and questions you were repeating in your tests (e.g. submitting something then validating it was submitted successfully) you can just create a sequence to wrap this up into one call.
It's recommended that you only make assertions about state in questions, keep 'em out of any tasks. If you're tempted to write an assertion in a task, think about whether you are really doing a fundamental action of the service/page. Just write it into a sequence if you really need to do both a task and question.
To use a service in a Scenario, insert it in the Scenario callback:
// in file myTest.js
Feature('MyFeatureA')
// See codeceptjs docs for info on Before/BeforeSuite/After/AfterSuite
BeforeSuite((myService) => {
myService.complete.someSetup();
});
Scenario('use feature with good input', (myService) => {
const result = myService.do.someTask(goodInput);
myService.ask.resultSuccess(result);
});
Scenario('use feature with bad input', (myService) => {
const result = myService.do.someTask(badInput);
myService.ask.resultHasError(result, 'invalid input!');
});To create a new service, use the command npm create-service.
A couple of files will be generated with the given service name in either services/apis or services/portal depending on the service type you select.
After running the command, you must add the service to the codecept.conf.js file for it to be accessible by Scenarios:
// in codecept.conf.js
...
include: {
myService: "./services/apis/myService.js"
}
...Most tasks are asynchronous, so asking questions about their state require a little async/await work. For most funcitons, you can return another funciton's promise and await for the promise to resolve in the Scenario. For example
// myServiceTasks.js
module.exports = {
async getSomething() {
return I.sendGetRequest('/endpoint')
}
}
// myServiceQuestions.js
module.exports = {
hasNoErrors(response) {
expect(response).to.not.have.property('error')
}
}
// myTest.js
Scenario('my test', (actor) => {
// we have to await here before continuing!
const response = await actor.do.getSomething();
actor.ask.hasNoErrors(response);
});When creating new Promises, avoid using reject unless you are going to catch it yourself. If we end up calling the reject() it will throw an error into the "global promise chain" of CodeceptJS, and importantly the test will fail without any meaningful stacktrace about what went wrong. Instead, you could do resolve({ error: err_obj }), and in a question assert there is no error.
Adding a tag (@tag) to a Scenario name allows filtering which tests are run. If your test is special for some reason (e.g. only applies to one commons) add a tag like so:
Scenario('test something @reqGoogle', () => {...});
Then using CodeceptJS's --grep flag we can filter the tags for example, to run only tests that have been flagged with @DCFOnly, we could do npm test -- --grep '@DCFOnly'. We can also invert the flags with --grep '@DCFOnly' --invert (see CodeceptJS's command docs).
Edit run-tests.sh.
Example situation: Fence only supports centralized auth after version 3.0.0. In other words, we don't want to run the @centralizedAuth tag in envs that are not running at least fence 3.0.0. You can make that explicit in the run-tests.sh file:
runTestsIfServiceVersion "@centralizedAuth" "fence" "3.0.0"
This will disable the @centralizedAuth tests on any pinned version of fence less than 3.0.0.
NOTE: Fence
masterand any branch will run the tests.