The following package contains the integration test suite for Juju.

Tests are structured into test suites. Each suite contains a root task (akin to a package test) that will setup and run each individual test.

To help break tests down, each test can have a number of subtests. Subtests are meant for individual units of work, without having to bootstrap a controller for every test. Each subtest will just ensure that it does have one; failing to find a suitable controller, it will create one for you.

/suites/deploy/                    # test-suite
               task.sh             # root/package test (setup)
               deploy_bundles.sh   # tests

run_deploy_bundles() {             # Subtest

}

test_deploy_bundles() {            # Test
    run "run_deploy_bundles"       # Run subtest
}

All tests will run through until the end of a test/subtest, unless it encounters a non-zero exit code. In other words, if you want to assert something passes, ensure that the command returns exit 0. Failure can then be detected of the inverse.

echo "passes" | grep -q "passes"   # passes
echo "failed" | grep -q "passes"   # fails

Before running tests, you'll need to install jq, yq, shellcheck and expect:

sudo snap install jq
sudo snap install yq
sudo snap install shellcheck
sudo snap install expect

curl is also required, but this should be preinstalled on most systems.

The static analysis tests also require golangci-lint:

go install github.com/golangci/golangci-lint/cmd/[email protected]

To get started, it's best to quickly look at the help command from the runner.

cd tests && ./main.sh -h

Running a full sweep of the integration tests (which will take a long time), can be done by running:

cd tests && ./main.sh -A

To run the tests, they can be broken down into steps:

./main.sh deploy                     # Runs deploy test suite
./main.sh deploy test_deploy_bundles # Runs test (and all of the subtests)
./main.sh deploy run_deploy_bundle   # Runs subtest

Note: running subtests, will also invoke the parent test to ensure that it has everything setup correctly.

Running ./main.sh deploy run_deploy_bundle will also run test_deploy_bundles, but no other subtests, just run_deploy_bundle.

The use of local controllers whilst development is advantagous because you don't have to rebootstrap a controller, or you can test a particular setup that has been manually created.

To do so, just specify a controller name and pass it though.

./main.sh -l <local-controller-name> deploy

Note: because you're using a local controller, you don't get the same guarantees of setup and cleanup that you can with letting the test suite do that for you. So if you expect everything to be cleaned up and leave no trace, then don't use this method of bootstrapping.