* The test method names, which will be executed for each test class, will be written at the proper running order to the log at the beginning of each test class(done by KcArquillian class).
* All the triggered arquillian lifecycle events and executed observers listening to those events will be written to the log
* The bootstrap of WebDriver will be unlimited. By default there is just 1 minute timeout and test is cancelled when WebDriver is not bootstrapped within it.
By default, WebDriver has 10 seconds timeout to load every page and it timeouts with error after that. Use this to increase timeout to 1 hour instead:
-Dpageload.timeout=3600000
### Surefire debugging
For debugging, the best is to run the test from IDE and debug it directly. When you use embedded Undertow (which is by default), then JUnit test, Keycloak server
and adapter are all in the same JVM and you can debug them easily. If it is not an option and you are forced to test with Maven and Wildfly (or EAP), you can use this:
and you will be able to attach remote debugger to the test. Unfortunately server and adapter are running in different JVMs, so this won't help to debug those.
When tests are run on JBoss based container (WildFly/EAP) there is possibility to attach a debugger, by default on localhost:5005.
The server won't wait to attach the debugger. There are some properties what can change the default behaviour.
-Dauth.server.debug.port=$PORT
-Dauth.server.debug.suspend=y
More info: http://javahowto.blogspot.cz/2010/09/java-agentlibjdwp-for-attaching.html
### JBoss app server debugging
Analogically, there is the same behaviour for JBoss based app server as for auth server. The default port is set to 5006. There are app server properties.
It is configured in `testsuite/integration-arquillian/tests/base/src/test/resources/log4j.properties` . You can see that logging of testsuite itself (category `org.keycloak.testsuite`) is debug by default.
When you run tests with undertow (which is by default), there is logging for Keycloak server and adapter (category `org.keycloak` ) in `info` when you run tests from IDE, but `off` when
you run tests with maven. The reason is that, we don't want huge logs when running mvn build. However using system property `keycloak.logging.level` will override it. This can be used for both IDE or maven.
So for example using `-Dkeycloak.logging.level=debug` will enable debug logging for keycloak server and adapter.
For more fine-tuning of individual categories, you can look at log4j.properties file and temporarily enable/disable them here.
When using Keycloak on Wildfly/EAP, there is INFO logging level enabled by default for most of the java packages.
You can use those system properties to enable DEBUG logging for particular packages:
*`-Dinfinispan.logging.level=DEBUG` - for package `org.infinispan`
*`-Dorg.keycloak.services.scheduled=DEBUG` - for package `org.keycloak.services.scheduled`
You can use value `TRACE` if you want to enable even TRACE logging.
There is no support for more packages ATM, you need to edit the file `testsuite/integration-arquillian/servers/auth-server/jboss/common/jboss-cli/add-log-level.cli`
1) Download JBoss Fuse 6.3 to your filesystem. It can be downloaded from http://origin-repository.jboss.org/nexus/content/groups/m2-proxy/org/jboss/fuse/jboss-fuse-karaf
Assumed you downloaded `jboss-fuse-karaf-6.3.0.redhat-229.zip`
2) Install to your local maven repository and change the properties according to your env (This step can be likely avoided if you somehow configure your local maven settings to point directly to Fuse repo):
1) Download JBoss Fuse 7 to your filesystem. It can be downloaded from http://origin-repository.jboss.org/nexus/content/groups/m2-proxy/org/jboss/fuse/fuse-karaf (Fuse 7.3 or higher is required)
Assumed you downloaded `fuse-karaf-7.3.0.fuse-730065-redhat-00002.zip`
2) Install to your local maven repository and change the properties according to your env (This step can be likely avoided if you somehow configure your local maven settings to point directly to Fuse repo):
2) Install to your local maven repository and change the properties according to your env (This step can be likely avoided if you somehow configure your local maven settings to point directly to EAP repo):
3) Download Fuse EAP installer (for example from http://origin-repository.jboss.org/nexus/content/groups/m2-proxy/com/redhat/fuse/eap/fuse-eap-installer/6.3.0.redhat-220/ )
1) Prepare MySQL DB and ensure that MySQL DB is empty. See [../../docs/tests-db.md](../../docs/tests-db.md) for some hints for locally prepare Docker MySQL image.
Same test as above, but it uses manual migration mode. During startup of the new Keycloak server, Liquibase won't automatically perform DB update, but it
just exports the needed SQL into the script. This SQL script then needs to be manually executed against the DB.
For the available versions, take a look at the directory [tests/other/server-config-migration/src/test/resources/standalone](tests/other/server-config-migration/src/test/resources/standalone)
The UI tests are real-life, UI focused integration tests. Hence they do not support the default HtmlUnit browser. Only the following real-life browsers are supported: Mozilla Firefox, Google Chrome and Internet Explorer. For details on how to run the tests with these browsers, please refer to [Different Browsers](#different-browsers) chapter.
The tests also use some constants placed in [test-constants.properties](tests/base/src/test/resources/test-constants.properties). A different file can be specified by `-Dtestsuite.constants=path/to/different-test-constants.properties`
Currently we are testing Spring Boot with three different containers `Tomcat 8`, `Undertow` and `Jetty [9.2, 9.3, 9.4]`. We are testing two versions of Spring Boot 1.5.x and 2.1.x. All versions are specified in [root pom.xml](../../pom.xml) (see properties `spring-boot15.version` and `spring-boot21.version`).
To run tests execute following command. Default version of Spring Boot is 1.5.x, to run tests with version 2.1.x add profile `-Pspringboot21`
Similarly to Admin Console tests, these tests are focused on UI, specifically on the parts of the server that are accessed by an end user (like Login page, or Account Console).
They are designed to work with mobile browsers (alongside the standard desktop browsers). For details on the supported browsers and their configuration please refer to [Different Browsers chapter](#different-browsers).
You can use many different real-world browsers to run the integration tests.
Although technically they can be run with almost every test in the testsuite, they can fail with some of them as the tests often require specific optimizations for given browser. Therefore, only some of the test modules have support to be run with specific browsers.
* **Driver download required:** [ChromeDriver](https://sites.google.com/a/chromium.org/chromedriver/) that corresponds with your version of the browser
You can rely on automatic driver downloads which is provided by [Arquillian Drone](http://arquillian.org/arquillian-extension-drone/#_automatic_download). To do so just omit the `-Dwebdriver.{browser}.driver` CLI argument when running the tests.
By default latest driver version is always downloaded. To download a specific version, add `-DfirefoxDriverVersion`, `-DchromeDriverVersion` or `-DieDriverVersion` CLI argument.
The support for testing with the mobile browsers is implemented using the [Appium](http://appium.io/) project.
This means the tests can be run with a real mobile browser in a real mobile OS. However, only emulators/simulators of mobile devices are supported at the moment (no physical devices) in our testsuite.
First, you need to install the Appium server. If you have Node.js and npm installed on your machine, you can do that with: `npm install -g appium`. For further details and requirements please refer to the [official Appium documentation](http://appium.io/docs/en/about-appium/intro/).
The tests will try to start the Appium server automatically but you can do it manually as well (just by executing `appium`).
To use a mobile browser you need to create a virtual device. The most convenient way to do so is to install the desired platform's IDE - either [Android Studio](https://developer.android.com/studio/) (for Android devices) or [Xcode](https://developer.apple.com/xcode/) (for iOS devices) - then you can create a device (smartphone/tablet) there. For details please refer to documentation of those IDEs.
#### Google Chrome on Android
* **Supported test modules:** `base-ui`
* **Supported host OS:** Windows, Linux, macOS
* **Supported browser version:** latest stable
* **Supported mobile OS version:** Android 7.x, 8.x
* **Run with:** `mvn clean test -Pandroid -Dappium.avd=name_of_the_AVD` where AVD is the name of your Android Virtual Device (e.g. `Nexus_5X_API_27`)
**Tips & tricks:**
* If the AVD name contains any spaces, you need to replace them with underscores when specifying the `-Dappium.avd=...`.
* It's probable that a freshly created device will contain an outdated Chrome version. To update to the latest version (without using the Play Store) you need to download an `.apk` for Chrome and install it with `adb install -r path/to/chrome.apk`.
* Chrome on Android uses ChromeDriver similarly to regular desktop Chrome. The ChromeDriver is bundled with the Appium server. To use a newer ChromeDriver please follow the [Appium documentation](http://appium.io/docs/en/writing-running-appium/web/chromedriver/).
#### Apple Safari on iOS
* **Supported test modules:** `base-ui`
* **Supported host OS:** macOS
* **Supported browser version:** _depends on the mobile OS version_
* **Supported mobile OS version:** iOS 11.x
* **Run with:** `mvn clean test -Pios -Dappium.deviceName=device_name` where the device name is your device identification (e.g. `iPhone X`)
Cluster tests use 2 backend servers (Keycloak on Wildfly/EAP or Keycloak on Undertow), 1 frontend loadbalancer server node and one shared DB. Invalidation tests don't use loadbalancer.
The browser usually communicates directly with the backend node1 and after doing some change here (eg. updating user), it verifies that the change is visible on node2 and user is updated here as well.
Failover tests use loadbalancer and they require the setup with the distributed infinispan caches switched to have 2 owners (default value is 1 owner). Otherwise failover won't reliably work.
The test uses Undertow loadbalancer on `http://localhost:8180` and two embedded backend Undertow servers with Keycloak on `http://localhost:8181` and `http://localhost:8182` .
You can use any cluster test (eg. AuthenticationSessionFailoverClusterTest) and run from IDE with those system properties (replace with your DB settings):
2) Run KeycloakServer server2 with same parameters but argument: `-p 8182`
3) Run loadbalancer (class `SimpleUndertowLoadBalancer`) without arguments and system properties. Loadbalancer runs on port 8180, so you can access Keycloak on `http://localhost:8180/auth`
The browser usually communicates directly with the frontent node and the test controls where the HTTP requests
land by adjusting load balancer configuration (e.g. to direct the traffic to only a single DC).
For an example of a test, see [org.keycloak.testsuite.crossdc.ActionTokenCrossDCTest](tests/base/src/test/java/org/keycloak/testsuite/crossdc/ActionTokenCrossDCTest.java).
b1) For **Undertow** Keycloak backend containers, you can run the tests using the following command (adjust the test specification according to your needs):
So skip steps 1,2 and add property `-Dmanual.mode=true` and change "cache.server.lifecycle.skip" to false `-Dcache.server.lifecycle.skip=false` or remove it.
5) If you want to debug or test manually, the servers are running on these ports (Note that not all backend servers are running by default and some might be also unused by loadbalancer):
