This document covers how to run all of the tests that are present in the Linkerd2 repo. Most of these tests are run in CI, but you can use the instructions here to run the tests from source. For more information about working in this repo, see the BUILD.md guide.
Note that all shell commands in this guide are expected to be run from the root
of this repo, unless otherwise indicated by a cd
command.
To run tests:
go test -cover -race ./...
To investigate code coverage:
cov=`mktemp`
go test -coverprofile=$cov ./...
go tool cover -html=$cov
When running go test
, mismatched text is usually displayed as a compact diff.
If you prefer to see the full text of the mismatch with colorized output, you
can set the LINKERD_TEST_PRETTY_DIFF
environment variable or run go test ./cli/cmd/... --pretty-diff
.
When kubernetes templates change, several test fixtures usually need to be
updated (in cli/cmd/testdata/*.golden
). These golden files can be
automatically regenerated with the command:
go test ./cli/cmd/... --update
Javascript dependencies are managed via yarn and webpack. We use jest as our test runner.
To fetch dependencies and run tests, run:
bin/web setup
bin/web test
# or alternatively:
cd web/app
yarn && NODE_ENV=test yarn webpack
yarn jest "$*"
For faster testing, run a subset of the tests by passing flags to jest.
Run tests on files that have changed since the last commit:
bin/web test -o
Run tests that match a spec name (regex):
bin/web test -t name-of-spec
Run watch mode:
bin/web test --watch # runs -o by default (tests only files changed since last commit)
bin/web test --watchAll # runs all tests after a change to a file
bin/shellcheck -x bin/*
The test/integration
directory contains a test suite that can be run to
validate Linkerd functionality via a series of end-to-end tests.
The integration tests will configure their own k3s clusters by default (using the k3d helper). There are no prerequisites for this test path.
If integration tests should run on an existing Kubernetes cluster, then the
--skip-cluster-create
flag should be passed. This will disable the tests from
creating their own clusters and instead use the current Kubernetes context.
In this case, ensure the following:
- The Linkerd docker images you're trying to test have been built and are
accessible to the Kubernetes cluster to which you are deploying.
If you're testing locally through a KinD or k3d cluster and don't want to push
the images to a public registry, you can call
bin/image-load --kind|k3d
to load all the Linkerd images into those clusters. - The
kubectl
CLI has been configured to talk to that Kubernetes cluster
You can use the bin/tests
script to run one or all of the tests in the test
suite.
The bin/tests
script requires an absolute path to a linkerd
binary to test.
Optional flags can be passed that change the testing behavior:
--name
: Pass an argument with this flag to specify a specific test that should be run; all tests (except some special ones, see below) are run in the absence of this flag. Valid test names are included in thebin/tests --help
output--skip-cluster-create
: Skip KinD cluster creation for each test and use an existing Kubernetes cluster--images
: (Primarily for CI) Loads images from theimage-archive/
directory into the KinD clusters created for each test
View full help text:
bin/tests --help
Run individual test:
bin/tests --name upgrade /path/to/linkerd
You can run tests using your installed version of the linkerd
CLI. For
example, to run the full suite of tests using your installed CLI, run:
bin/tests `which linkerd`
If using an existing cluster to run tests, the resources can be cleaned up manually with:
bin/test-cleanup /path/to/linkerd
You can also test a locally-built version of the linkerd
CLI.
First build all of the Linkerd images by running:
bin/docker-build
That command also copies the corresponding linkerd
binaries into the
target/cli
directory, and you can use the bin/linkerd
script to load those
binaries when running tests. To run tests using your local binary, run:
bin/tests $PWD/bin/linkerd
Note: As stated above, if running tests in an existing KinD cluster by
passing --skip-cluster-create
, bin/kind-load
must be run so that the images are
available to the cluster
When running bin/tests
without specifying --name
all tests except for
cluster-domain
, cni-calico-deep
and multicluster
are run, because these require
creating the clusters with special configurations. To run any of these tests,
invoke them explicitly with --name
for the script to create the cluster (using
k3d) and trigger the test:
bin/tests --name cluster-domain
: This simply creates the cluster with a cluster domain setting different than the defaultcluster.local
, then installs Linkerd and triggers some smoke tests.bin/tests --name cni-calico-deep
: This installs a cluster replacing the default CNI plugin (which for k3s is Flannel) with the Calico CNI plugin, then installs the Linkerd CNI plugin and the Linkerd control plane, and finally triggers the full suite of deep tests.bin/tests --name multicluster
: Two k3d clusters are installed each one with separate instances of Linkerd sharing the same trust root. Then the multicluster component is installed, both clusters are linked together and a test ensures exported services can be reached between the two clusters.
We use WebdriverIO to test how the web dashboard looks and operates locally in Chrome. For cross-browser testing, we use SauceLabs, which runs simultaneous tests on different browsers in the cloud.
If you're new to the repo, make sure you've installed web dependencies via Yarn:
brew install yarn # if you don't already have yarn
bin/web setup
Then start up the dashboard at localhost:7777
. You can do that in one of two
ways:
# standalone
bin/web run
OR
# with webpack-dev-server
bin/web dev
To run a local WebdriverIO instance that will run the tests on a local instance of Chrome, run:
bin/web integration local
To run cross-browser tests via SauceLabs, you need to do a few things first:
-
Sign up for a (free) SauceLabs sub-account for the account 'buoyant'. If you are not a Buoyant staffer, the best way to get an account invite is to ask in the Linkerd Slack channel.
-
Once you have your username and key, set them as permanent environment variables. This keeps your credentials private, and means that everyone on the team can run the tests via their unique login without modifying the test files. Open your
~/.bash_profile
file and add:export SAUCE_USERNAME="your Sauce username" export SAUCE_ACCESS_KEY="your Sauce access key"
-
Now you'll download Sauce Connect, the proxy server that will open a secure tunnel between a SauceLabs VM and the Linkerd dashboard instance you're running on
localhost:7777
. You'll want to save it in a separate directory from the rest of your development files. After downloading it, navigate to that directory and start it up:SC=sc-4.5.3-osx # OSX example wget -O - https://saucelabs.com/downloads/$SC.zip | tar xfz - -C ~/ cd ~/$SC bin/sc -u $SAUCE_USERNAME -k $SAUCE_ACCESS_KEY
Wait until you see
Sauce Connect is up, you may start your tests
in your terminal. Open a separate terminal window and run:bin/web integration cloud
SauceLabs will start running the tests in the cloud. If any tests fail, you'll immediately get the URL in your terminal window with a video of the test and information about what happened. The test(s) will also appear in your SauceLabs archives a minute or so after they end. (Depending on time of day and server load, it may take longer for the tests to appear in the archives.)
-
When you're finished, close the tunnel by pressing
CTRL-C
in the Sauce Connect window. If you forget to do this, it will close on its own after a few minutes.
To add a new test, create a new subdirectory inside the test/
directory.
Configuration files, such as Kubernetes configs, should be placed inside a
testdata/
directory inside the test subdirectory that you created. Then create
a test file in the subdirectory that's suffixed with _test.go
. This test file
will be run automatically by the test runner script.
The tests rely heavily on the test helpers that are defined in the testutil/
directory. For a complete description of how to use the test helpers to write
your own tests, view the testutil
package's godoc, with:
godoc github.com/linkerd/linkerd2/testutil | less
The scale tests deploy a single Linkerd control-plane, and then scale up multiple sample apps across multiple replicas across multiple namespaces.
Prerequisites:
- a
linkerd
CLI binary - Linkerd Docker images associated with the
linkerd
CLI binary - a Kubernetes cluster with sufficient resources to run 100s of pods
bin/test-scale
usage: test-scale /path/to/linkerd [namespace]
For example, to test a newly built Linkerd CLI:
bin/test-scale `pwd`/bin/linkerd
bin/test-cleanup /path/to/linkerd
The bin/test-clouds
script runs the integration tests
against 4 cloud providers:
- Amazon (EKS)
- DigitalOcean (DO)
- Google (GKE)
- Microsoft (AKS)
This script assumes you have a working Kubernetes cluster set up on each Cloud provider, and that Kubernetes contexts are configured via environment variables.
For example:
export AKS=my-aks-cluster
export DO=do-nyc1-my-cluster
export EKS=arn:aws:eks:us-east-1:123456789012:cluster/my-cluster
export GKE=gke_my-project_us-east1-b_my-cluster
For more information on configuring access to multiple clusters, see: https://kubernetes.io/docs/tasks/access-application-cluster/configure-access-multiple-clusters/#define-clusters-users-and-contexts
bin/test-clouds `pwd`/bin/linkerd
To cleanup all integration tests:
bin/test-clouds-cleanup