OpenShift extended test suite
This document describes how a developer can write a new extended test for
OpenShift and the structure of extended tests.
Running tests
From the top-level origin directory, run
$ test/extended/<some_script>.sh
Where <some_script>.sh is one of the bucket scripts such as "core.sh".
You can further narrow the set of tests being run by setting the environment
variable FOCUS='regex'
where 'regex' is a regular expression matching the
description of the test you want to run. For example one of the s2i tests
(s2i_incremental.go) defines:
var _ = g.Describe("[Feature:Builds][Slow] incremental s2i build", func() {
So you can write a focus regex that includes this test by setting
FOCUS='\[builds\]'
or FOCUS='incremental s2i'
.
Prerequisites
In order to execute the extended tests, you have to install
Ginkgo framework which is used in extended
tests. You can do it by running following command:
$ go get github.com/onsi/ginkgo/ginkgo
You also need to have the openshift
binary in the PATH
if you want to use
the shell script helpers to execute the extended tests.
Rapid local testing
If you already have a running OpenShift cluster, e.g. one created using oc cluster up
, you can skip having the extended test infrastructure spin up an
OpenShift cluster each time the tests are run by setting the TEST_ONLY
environment variable as follows:
$ oc cluster up
$ oc login -u system:admin
$ export KUBECONFIG=${KUBECONFIG-$HOME/.kube/config}
Then, for example:
$ make build-extended-test
$ FOCUS='\[builds\]' TEST_ONLY=1 test/extended/core.sh
By default the Kubernetes test framework will remove the project associated with
your test spec when it completes, regardless of whether it fails or not.
Origin's wrapper scripts may also do clean-up. Running tests in parallel can
also hinder debugging. To stop these behaviours, set the SKIP_TEARDOWN
environment variable, set DELETE_NAMESPACE=false
, and set PARALLEL_NODES=1
:
$ make build-extended-test
$ FOCUS='\[builds\]' TEST_ONLY=1 SKIP_TEARDOWN=1 DELETE_NAMESPACE=false PARALLEL_NODES=1 test/extended/core.sh
Test labels
See kinds of tests
for a full explanation of the labels used for each test spec. In brief:
-
If a test has no labels, it is expected to run fast (under five minutes), be
able to be run in parallel, and be consistent.
-
[Serial]: If a test cannot be run in parallel with other tests (e.g. it
takes too many resources or restarts nodes), it is labeled [Serial], and
should be run in serial as part of a separate suite.
-
[Slow]: If a test takes more than five minutes to run (by itself or in
parallel with many other tests), it is labeled [Slow]. This partition allows
us to run almost all of our tests quickly in parallel, without waiting for the
stragglers to finish.
OpenShift extended tests that run builds should be marked [Slow].
-
Tests should be marked [Conformance] when they provide test coverage for
functionality considered core and critical to a functional cluster (i.e. not
exotic features/configurations) and which is not overlapping with coverage
provided in other conformance tests. Example of a valid conformance test: "Do
builds work." Example of an invalid conformance test: "Do builds work when the
forcePull flag is set."
-
In general, accessing the local host (e.g. using the docker socket) in
extended tests should be avoided. If this is unavoidable, the test should be
marked [local].
Extended tests structure
Extended tests live under the ./test/extended
directory in the origin repository.
The structure of this directory is following:
test/extended/util
provides useful helpers and utilities to use in your extended test. It provides a easy-to-use interface to OpenShift CLI and also
access to the Kubernetes E2E framework helpers. It also contains OpenShift helpers that are shared across multiple test cases, to make the test cases more DRY.
test/extended/fixtures
contains the JSON and YAML fixtures that are meant to be used by the extended tests.
test/extended/[images,builds,...]
each of these Go packages contains extended tests that are related to each other. For example, the images
directory should contain test cases that are exercising usage of various Docker images in OpenShift.
hack/test-extended/[group]/run.sh
is the shell script that sets up any needed dependencies and then launches the extended tests whose top level ginkgo spec's Describe call reference the group
test/extended/extended_test.go
is a runner for all extended test packages. Look inside this file to see how you can add new extended test Go package to be compiled:
_ "github.com/openshift/origin/test/extended/builds"
_ "github.com/openshift/origin/test/extended/images"
Groups vs. packages
Each type of functional test should be in its own package. However, if your
package needs to specifically configure the server in a different way than
the standard path, you would create a new launcher script with the same name
as your package in the test/extended
dir.
For example, you might want to write an extended test for the LDAP
authentication which means that you have to configure the OpenShift server to
enable this authentication method. You can create a new test group ldap
and
provide a shell launcher ./test/extended/ldap.sh
to start OpenShift with the
required configuration.
Then you place the source code for the extended test into the extended test Go
package that corresponds to functionality you are going to test. In the case of
LDAP, it can be ./test/extended/ldap
. You should include a prefix for your
test cases at your root suite level.
Example:
var _ = g.Describe("[ldap] Authenticate using LDAP", func() {
# ...
})
Creating new test group runner
If your test requires different configuration than the rest of the extended
test cases, you should create a new execution script in test/extended
. Be
sure to set your focus to ginkgo appropriately to select only your test.
If your tests cannot run as part of the default group, be sure to ensure your
package is not included by test/extended
.
Bash helpers for creating new test group runner
Common functions for extended tests are located in ./hack/util.sh
. Environment setup scripts are located in hack/lib/util/evironment.sh
.
ginkgo_check_extended()
verify if the Ginkgo binary is installed.
compile_extended()
perform the compilation of the Go tests into a test binary.
test_privileges()
verify if you have permissions to start OpenShift server.
os::util::environment::setup_all_server_vars()
setup all required environment variables related to OpenShift server.
os::start::configure_server()
generates all configuration files for OpenShift server.
os::start::server()
starts the OpenShift master and node.
os::start::router()
installs the OpenShift router service.
os::start::registry()
installs the OpenShift Docker registry service.
create_image_streams_extended()
creates ImageStream(s) for all OpenShift images.
CLI interface
In order to be able to call the OpenShift CLI and Kubernetes and OpenShift clients and simulate the OpenShift oc
command in the test suite, first we need to create an instance of the CLI, in the top-level Ginkgo describe container.
The top-level describe container should also specify the bucket into which the test belongs and a short test description. Other globally accessible variables (eg. fixtures) can be declared as well.
package extended
import (
g "github.com/onsi/ginkgo"
o "github.com/onsi/gomega"
)
var _ = g.Describe("[<test bucket>] <Testing scenario>", func() {
defer g.GinkgoRecover()
var (
oc = exutil.NewCLI("test-name", exutil.KubeConfigPath())
testFixture = filepath.Join("testdata", "test.json")
)
})
The test suite should be organized into lower-level Ginkgo describe(s) container, together with a message which elaborates on the goal of the test. Inside each lower-level describe container specify a single spec with the It
container , which shares the context in which the spec runs. The It
container also takes a message which explains how the test goal will be achieved.
var _ = g.Describe("[default] STI build", func() {
defer GinkgoRecover()
var (
stiBuildFixture = filepath.Join("testdata", "test-build.json")
oc = exutil.NewCLI("build-sti", kubeConfigPath())
)
g.Describe("Building from a template", func() {
g.It(fmt.Sprintf("should create a image from %q template", stiBuildFixture), func() {
...
}
}
}
After that you are free to simulate any oc
command by calling the CLI methods from the extended package.
As first, the command verb (get, create, start-build, ...) has to be specified upon the created CLI instance with the Run()
method.
oc = oc.Run("create")
Then the command parameters have to be specified by using the Args()
command. You may also notice the methods can be easily chained.
oc = oc.Run("create").Args("-f", testFixture)
A Go template can be set as a parameter for the OpenShift CLI command, by using the Template()
method. Keep in mind that in order to use this method, the get
verb has to be specified by the Run()
command.
oc = oc.Run("get").Args("foo").Template("{{ .spec }}")
is an equivalent to
$ oc get foo -o template --template='{{ .spec }}'
To execute the command you will need to call either Execute()
, which will execute the command and return any error that occurs, or Output()
which returns any error that occurs as well as the output.
err := oc.Run("create").Args("-f", testFixture).Execute()
buildName, err := oc.Run("start-build").Args("test").Output()
To print out the purpose of the next command, or set of commands, use the Ginkgo’s By
function.
g.By("starting a test build")
buildName, err := oc.Run("start-build").Args("test").Output()
To evaluate if the the command was successfully executed without any errors retrieved, use the Gomega’s Expect
syntax to make expectations on the error.
err = oc.Run("create").Args("-f", stiEnvBuildFixture).Execute()
o.Expect(err).NotTo(o.HaveOccurred())