jueves, octubre 13, 2016

Build Docker Images with Maven and Gradle

One of the things that you might want to do if you are using Docker and Java is building the image from a Dockerfile in your build tool (Maven or Gradle).  In this post I am going to show you how to do it in both cases.

I am going to assume that you have the de-facto project layout, having the Dockerfile file at the root of the project.


There are several Maven plugins that can be used for building a Docker image in Maven, but one of the most used is fabric8-maven-plugin.

To start you need to register and configure the plugin in pom.xml:

In configuration section you set the image name and the directory where Dockerfile is located.

Any additional files located in the dockerFileDir directory will also be added to the build context. Since Dockerfile is on the root of the project, the target directory is added too. The problem arises because this plugin uses target/docker to generate the build and if you try to build it you'll get next exception:  tar file cannot include itself. To avoid this problem you need to create .maven-dockerignore file specifying which directory must be ignored at the same level as Dockerfile:

And that's all, after that you can do:

mvn package docker:build

Notice that this plugin honor Docker environment variables like DOCKER_HOST, DOCKER_CERT_PATH, ... so if your environment is correctly configured you don't need to do anything else.


There are several Gradle plugins that can be used for building a Docker image in Gradle, but one of the most used is gradle-docker-plugin.

To start you need to register and configure the plugin in build.gradle:

In case of Gradle, you need to configure Docker host properties since plugin does not honor Docker environment variables. You need to configure them in docker {} block.

Finally you create a task of type DockerBuildImage, where you set the Dockerfile root directory using inputDir attribute and image name using tag attribute.


So in this post you've seen different ways of doing the same in two different build tools, which is building a Docker image from a Dockerfile. Notice that these plugins also allows you to define the Dockerfile content as a configuration field, so you are not creating a Dockerfile file, but specifying its content inside the build tool. You can read more about this feature at https://dmp.fabric8.io/ in case of Maven plugin and  https://github.com/bmuschko/gradle-docker-plugin#creating-a-dockerfile-and-building-an-image in case of Gradle.

We keep learning,

Bees'll buzz, kids'll blow dandelion fuzz, And I'll be doing whatever snow does in summer., A drink in my hand, my snow up against the burning sand, Prob'ly getting gorgeously tanned in summer. (In Summer - Frozen)

jueves, septiembre 22, 2016

Authenticating with JGit

JGit is a lightweight, pure Java library implementing the Git version control system. You can do a lot of operations using Java language such as create or clone Git repos, create branches, make commits, rebase or tag, you can see this repo to learn how to use JGit and how to code the different commands.

But one thing that does not cover extensively is the authentication process. In this post I am going to show you how how to authenticate to a Git repository with JGit.

First thing to do is add JGit as dependency:

Then let's see a simple clone without authentication:

In this case no authentication method is set. Now let's see how to add a username and password in case of for example private repos:

In this case you only need to set as credential provider the UsernameAndPasswordCredentialsProvider and pass the required username and password.

The final scenario I am going to show here is how to authenticate against a git repository using your ssh keys, that is using (~/.ssh/id_rsa) and setting the passphrase to access it.

In this case you need to extend JSchConfigSessionFactory to be able to set passphrase to access to private key. To do it you set a custom UserInfo implementation where the getPassphrase method returns the passphrase to use and promptPassphrase method should return true.

After that you only need to set the transport configuration to the one created.

We keep learning,
Chan eil inneal-ciùil a ghleusar, 'Dhùisgeas smuain mo chléibh gu aoibh, Mar nì duan o bheul nan caileag, Oidhche mhath leibh, beannachd leibh (Oidche Mhath Leibh - Ossian)
Music: https://www.youtube.com/watch?v=mi4SCOYAdEk

lunes, septiembre 19, 2016

Arquillian Chameleon for the sake of simplicity

When using Arquillian, one of the things you need to do is defining under which container you want to execute all your tests.

And this is done by adding a dependency in the classpath for the adapter and depending on the mode used (embedded, managed or remote) having to download the application server manually. For example this happens when Wildfly is used in embedded or managed mode.

An example of a pom.xml using Wildfly could be:

Notice that in previous script, you need to define the Arquillian adapter, in this case the managed one, and use maven-dependency-plugin to download Wildfly distribution file used by Arquillian.

This approach is good and it works, but it has three drawbacks:

  1. You need to repeat all these lines in every build script you want to use Arquillian and Wildfly.
  2. In case you need to use another application server in another project, you need to know which adapter artifact is required and if it is necessary to download  the artifacts or not. For example in case of Jetty embedded it is not necessary to download any distribution, you only need set the embedded dependency.
  3. If you want to test your code against several application servers you have the problem number 2 plus start dealing with profiles.
But all these problems can be fixed using Arquillian Chameleon. Arquillian Chameleon is a generic container which reads from arquillian.xml which container, which version and which mode you want to use in your tests, and he will take care of adding required adapter into classpath, download any required distribution and configure the protocol (this is something that as a user you should not touch).

How to use Arquillian Chameleon is pretty easy. Do whatever you would do normally such as adding Arquillian bom and add Chameleon Container instead of any application-server specific artifact:

Then create in src/test/resources the Arquillian configuration file called arquillian.xml with next configuration:

Notice that now you only need to use a friendly property called chameleonTarget to define which container, version and mode you want to use. In previous example Wildfly 9.0.0.Final with managed adapter.

When running any test with this configuration, Chameleon will check if Wildfly 9.0.0.Final distribution is downloaded, and if not download it, then will add to classpath the managed adapter for Wildfly 9.0.0 and finally execute the test as any other Arquillian test.

What's happening if you want to use Payara instead of Wildfly? You only need to change chameleonTarget property to payara:, to for example run tests against Payara 4.1.1 in managed mode.

TIP: You can set this property using a Java system property (-Darq.container.chameleon.chameleonTarget = payara:

Currently next containers are supported by Chameleon:

  • JBoss EAP 6.x, 7.x
  • WildFly 10.x, 9.x, 8.x
  • JBoss AS 7.x
  • GlassFish 3.1.2, 4.x
  • Payara 4.x

We keep learning,
I can see you, Your brown skin shining in the sun, I see you walking real slow(The boys of summer - The Ataris)
Music: https://www.youtube.com/watch?v=Qt6Lkgs0kiU

lunes, agosto 29, 2016

Configuring Maven Release Plugin to Skip Tests

If you are using Maven and using Maven Release Plugin, you would like to skip the execution of tests during the release plugin execution. The reason might be very different but might depend on the nature of the project or how CI pipeline is implemented.

Notice that this might be a really improvement in releasing time since performing the release with Maven Release Plugin implies executing the same tests twice, one in prepare step and the other one in perform step.

To avoid executing tests in prepare phase you need to run as:

mvn -DpreparationGoals=clean release:prepare

If you want to avoid executing tests during perform phase you need to run as:

mvn -Darguments="-Dmaven.test.skip=true" release:perform

Please it is important to note that I am not saying you don't need to execute tests during release process, what I am saying is that something your release process doesn't fit the standard release process of the plugin and for example you are already running tests before executing the plugin.

We keep learning,
Say it ain't so, I will not go, Turn the lights off, carry me home, Keep your head still, I'll be your thrill, The night will go on, my little windmill (All The Small Things - Blink-182)

jueves, agosto 18, 2016

Making Web UI testing great again with Arquillian, Docker and Selenium (part 1)

Introduction to the Problem

Most of the time when you need to write functional tests/end-to-end tests for web UI, you end up by using Selenium, which it can consider the de-facto tool in Java world for web UI testing. I am sure you've already used it for these kind of tests.

But probably at the same time you've been faced on some the most common problems in functional testing, some related with Web UI testing and others not.

For example one of the major problems usually people find in functional tests are the preparation of the environment, to run the tests you need to boot up a server and deploy your application, then install/start the database, also maybe the cache system and so on with all the servers, leaving the user to install locally each of the service. Some errors could happen like installing incorrect version of the server used in production, reusing another local installation of the database which might not be the same version or for example running them in a different JDK version of the one used in production.

But also there are some other problems that are more specific to Web UI testing such as browser installation or configuration of WebDriver properties.

Fixing First Problem

To fix the first problem, the most easier solution you can think is using Docker containers and of course Docker compose since you can define and run multi-container Docker applications. So basically you define in docker-compose file, all the servers that you might need to run the tests so when you run tests you have all of them running and more important with a fixed version, so you can be sure that the tests are always run against a known/desired specific version of the servers, same JDK, ... and not depending on what is installed in developers/CI machine.

But this approach has one problem. You need to specifically run docker-compose up, docker-compose down. Of course you can automate this in your build script, which will solve the problem on CI environment, but if a developer wants to execute a test from IDE, let's say for debugging, then he needs to be aware of that fact.

And this is what Arquillian Cube solves. Arquillian Cube is an Arquillian extensions that  uses docker-compose file to start and configure all the containers defined there, execute the tests and finally shutting down all of them. The good news is that since Arquillian works with JUnit (and TestNG and Spock), you can run the tests from the IDE without worrying about starting and stopping containers since Docker lifecycle is managed by Arquillian Cube.

So first part of the problem that is defining the test environment is fixed with Arquillian Cube. Let's see how to fix the second one.

Fixing Second Problem

Selenium project provides a Docker images with Selenium standalone or Selenium node with browser (Firefox or Chrome) and a VNC server installed.

So it seems a perfect fit to fix the problem of having to install browsers with a concrete version or concrete configurations locally since you can use a docker image with a browser configured for the tests.

New Problems When Using Docker for Testing

And that's cool, but it has some problems. The first one is that you need to create a docker-compose file specific for testing purposes, although this is not a bad thing per se, but it requires more management from dev part to maintain this file as well and of course repeat again and again in all the projects you want to use it, defining the browser to use and the VNC client image to get the recording for future inspection.

The second problem is the configuration of WebDriver. When running WebDriver against a remote browser, you need to set the location (IP) of the browser and configure the RemoteWebDriver accordantly with desired capabilities.

So again you have to write in all the tests the WebDriver configuration again and again. You can create a factory class that can be reused in all the projects, and it is good, but you still have one problem, some developers might use Docker machine so IP would not be static and might change every time, other might be using native Docker, and for example some phases of CI pipeline might run the tests against a remote fully environment like preproduction environment, so before executing tests you would need to specify manually the IP of container of Docker host.

And the third problem you'll get is that you need to instruct WebDriver to open a page:


The problem is that in this case the browser is inside the Docker infrastructre so you need to set the internal IP of the server container, so you don't only need to know the Docker host IP for connecting the remote web driver but also the internal IP of the server container to open the page in remote browser using the get method. And again this might be quite difficult to acquire in an automatic way.

But all these problems are solved when using the new integration between Arquillian Drone and Arquillian Cube.

Fixing New Problems

Arquillian Drone is an Arquillian extension that integrates Selenium WebDriver to Arquillian. This extension manages the configuration of the WebDriver so you don't need to repeat it in all your tests, and also the lifecycle of the browser.

So as you can see this pair of extensions seems a perfect fit for solving these problems. Drone takes care of configuration meanwhile Cube takes care of configuring correctly the Selenium/VNC containers and starting and stopping them.

As you might see, you don't need to worry about creating docker-compose file for testing purposes. You only need to create the one used for deploying, and Arquillian will take care of the rest.


The first thing to do is create a project with required dependencies. For this example we are using Maven, but you can achieve the same using other build tools.

Things important to notice is that you are using BOM definitions for setting versions of the components. Then we set Arquillian Standalone dependency because our test is not going to have @Deployment method since the deployment file is already created inside the Docker image used in the application. Finally Arquillian Cube and Arquillian Drone dependencies are added.

Next step is creating at src/test/resources a file called arquillian.xml which is used for configuring extensions.

You can see that:

  • You need to specify the docker machine name where to start containers in case of using docker machine. If you are using native Docker then you don't need to set this attribute.
  • You need to set a location relative to root folder of the project where docker-compose file is located. Note that you could use any other name.
You can customize WebDriver as well configuring Arquillian Drone (https://docs.jboss.org/author/display/ARQ/Drone),  but for this test the defaults are enough. Note that now the default browser is firefox.

IMPORTANT: if you are using native Linux Docker installation, comment the configuring line of machineName. If you are using docker machine and it is called different to dev, then adapt machineName in arquillian.xml too.

Next step is creating the docker-compose file at root directory.

Simple compose file which defines only one container. This containers exposes the 80 port but then it is bound to port 8080. This container start a Go program listening to root context and returning Hello World in HTML format.

And finally the test:

There are some interesting parts in this test.

  • It is a standard Arquillian test in the sense it uses Arquillian runner.
  • Uses @Drone injection mechanism provided by Arquillian Drone to enrich test with a WebDriver configured to connect to remote browser.
  • Uses @CubeIp annotation to enrich test with the internal IP of the container helloworld. Since browser is running inside Docker host, we can use the internal IP for this purpose. Also it is important that you need to use the exposed port and not the bind port.
  • Everything else is managed by Arquillian Cube like the start and stop of the Docker containers(helloworld in this case) but also the ones containing the browser and the VNC client. If you put a debug point inside test method, and then execute a docker ps on a terminal, you'll see that three containers are started, not just helloworld
  • If after running the test you inspect target/reports/videos directory you will find the video recording of the test.
You can also see an screencast of this in action:

So as you can see using Arquillian Cube with Arquillian Drone makes your test and docker-compose file looks really neat.  Test only contains things related of the test and not about WebDriver configuration. Also your docker-compose looks clear, it only contains things related to business, not about testing.

In this post you've seen how to use Arquillian Cube + Arquillian Drone. In next one you'll see the integration with Arquillian Graphene, which will simplify even more the test to just focusing testing and not on WebDriver calls.

We keep learning,

When I look 'round, I only see outta one eye
As the smoke surrounds my head, the sauna (Stickin' In My Eye - NOFX)

jueves, marzo 31, 2016

Continuous Stress Testing for your JAX-RS (and JavaEE) applications with Gatling + Gradle + Jenkins Pipeline

In this post I am going to explain how to use Gatling project to write stress tests for your JAX-RS Java EE endpoints, and how to integrate them with Gradle and Jenkins Pipeline, so instead of having a simple stress tests, what you have is a continuous stress testing, where each commit might fire these kind of tests automatically, providing automatic assertions and more important graphical feedback of each execution so you can monitorize how the performance is evolving in your application.

First thing to develop is the JAX-RS JavaEE service:

There is nothing special, this is an asynchronous JAX-RS endpoint that connects to swapi.co site, retrieves all the information of Star Wars planets, calculates the average of orbital period and finally it returns it in form of text. For sake of simplicity, I am not going to show you all the other classes but they are quite simple and at the end of the post I will provide you the github repository. 

The application is packaged inside a war file and deployed into an application server. In this case into an Apache TomEE 7 deployed inside the official Apache TomEE Docker image.

Next step is configuring Gradle build script with Gatling dependencies. Since Gatling is written in Scala you need to use Scala plugin.

After that, it is time to write our first stress test. It is important to notice that writing stress tests for Gatling is writing a Scala class using the provided DSL. Even for people who has never seen Scala is pretty intuitive how to use it.

So create a directory called src/test/scala and create a new class called AverageOrbitalPeriodSimulation.scala with next content:

Every simulation must extends Simulation object. This simulation takes base URL of the service from starwars_planets_url environment or system property, it creates the scenario pointing to the endpoint defined in JAX-RS,  and finally during 3 seconds it will gradually add users until 10 users are running at the same time. The test will pass only if all the requests succeed in less than 3 seconds.

Now we need to run this test. You will notice that this is not a JUnit test, so you cannot do a Run As JUnit test. What you need to do is use a runnable class provided by Gatling which requires you pass as argument the simulation class. This is really easy to do with Gradle.

We are defining a Gradle task of type JavaExec, since what we want is to run a runnable class. Then we make the life a bit easier for developer by automatically detect that if starwars_planets_url is not set, we are running this test into a machine that has Docker installed so probably this is the host to be used.
Finally we override the environment variable if it is required, we set the runnable class with required properties and we configure Gradle to execute this task every time the test task is executed (./gradlew test).

If you run it, you might see some output messages from Gatling, and after all a message like: please open the following file: /Users/..../stress-test/build/reports/gatling-results/averageorbitalperiodsimulation-1459413095563/index.html and this is where you can get the report. Notice that a random number is appended at the end of the directory and this is important as we are going to see later. The report might looks like:

At this time we have Gatling integrated with Gradle, but there is a missing piece here, and it is adding the continuous part on the equation. For adding continuous stress testing we are going to use Jenkins and Jenkins Pipeline as CI server so for each commit stress tests are executed among other tasks such as compile, run unit, integration tests, or code quality gate.

Historically Jenkins jobs were configured using web UI, requiring users to manually create jobs, fill the details of the job and create the pipeline through web browser. Also this makes keeping configuration of the job separated from the actual code being built.

With the introduction of Jenkins Pipeline plugin. This plugin is a Groovy DSL that let's implement you the entire build process in a file and store that alongside its code. Jenkins 2.0 comes by default with this plugin, but if you are using Jenkins 1.X you can install it as any other plugin (https://wiki.jenkins-ci.org/display/JENKINS/Pipeline+Plugin)

So now we can start coding our release plugin but for the purpose of this post only stress part is going to be covered. You need to create a file called Jenkinsfile (the name is not mandatory but it is the de-facto name) on the root of your project, and in this case with next content:

In this case we are defining a new stage which is called Stress Test. Stage step is only used as informative and it will be used for logging purposes. Next a node is defined. A node is a Jenkins executor where to execute the code. Inside this node, the source code is checked out from the same location where Jenkinsfile is placed, sets a new environment variable pointing out to the location where the application is deployed, and finally a shell step which executes the Gradle test task.

Last step in Jenkins is to create a new job of type Pipeline and set the location of the Jenkinsfile. So go to Jenkins > New Item > Pipeline and give a name to the job.

Then you only need to go to Pipeline section and configure the SCM repository where the project is stored.

And then if you have correctly configured the hooks from Jenkins and your SCM server, this job is going to be executed for every commit, so your stress tests are going to run continuously.

Of course probably you have noticed that stress tests are executed but no reports are published in Jenkins, so you have no way to see or compare results from different executions. For this reason you can use publishHtml plugin to store the generated reports in Jenkins. If you don't have the plugin installed yet, you need to install it as any other Jenkins plugin.

PublishHtml plugin allows us to publish some html files generated by our build tool to Jenkins so they are available to users and also categorised by build number. You need to configure the location of the directory of files to publish, and here we find the first problem, do you remember that Gatling generates a directory with a random number? So we need to fix this first. You can follow different strategies, but the easiest one is simply rename the directory to a known static name after the tests.

Open Gradle build file and add next content.

We are creating a new task executed at the end of test task that renames the last created directory to averageorbitalperiodsimulation.

Final step is add after shell call in Jenkinsfile next call:

publishHTML(target: [reportDir:'stress-test/build/reports/gatling-results/averageorbitalperiodsimulation', reportFiles: 'index.html', reportName: 'Gatling report', keepAll: true])

After that you might see a link in the job page that points to the report.

And that's all, thanks of Gradle and Jenkins you can implement a continuous stress testing strategy in an easy way and just using code the language all developers speak.

We keep learning,
I can live whatever way I please, I move around among the seven seas, No one will miss me when the sun goes down, and in the morning I'be out of town (Movin' Cruisin' - The Fantastic Oceans)

Music: https://www.youtube.com/watch?v=Byg5Xq_pb74
Source Code: https://github.com/lordofthejars/starwars

lunes, marzo 07, 2016

Docker and Jenkins - Orchestrating Continuous Delivery

Past week I had the honour of speaking in Docker Barcelona Meetup about how to use Jenkins for doing typical Docker tasks like creating images, publishing them or having a trace of what has occurred on them. Finally I introduced the new (or not so new) Jenkins Pipeline plugin which allows you to create your continuous delivery pipeline by coding it using a Groovy DSL instead of relaying on static steps like happens when you use FreeStyle jobs. At the end I showed how to use it with Docker.

You can see the slides in slideshare or as html.

We keep learning,

Hello, it's me, I was wondering if after all these years you'd like to meet, To go over everything, They say that time's supposed to heal ya (Hello - Adele)

Music: https://www.youtube.com/watch?v=YQHsXMglC9A

viernes, enero 08, 2016

Container Object pattern. A new pattern for your tests.

If you search for a description of what Page Object is, you’ll find that The Page Object Pattern gives us a common sense way to model content in a reusable and maintainable way.

And also points that: Within your web app’s UI there are areas that your tests interact with. A Page Object simply models these as objects within the test code.
This reduces the amount of duplicated code and means that if the UI changes, the fix need only be applied in one place.

As you can see, Page Object applies to UI elements. We (the Arquillian community) has coined a new pattern following Page Object pattern logic called Container Object pattern.
You can think about Container Object as areas of a container (for now Docker container) that your test might interact with. For example some of these areas could be:
  • To get the host IP where container is running.
  • The bounded port for a given exposed port.
  • Any parameter configured inside the configuration file (Dockerfile) like a user or password to access to the service which the container exposes.
  • Definition of the containers.
A Container Object might contain an aggregation of more than one Container Object inside it. This effectively builds a relation ship (link) between containers.

An example of configuration parameters might be for example, in case of running a MySQL database in a container, it could be the user and password to access to database. 
Notice that nothing prevents you to generate the correct URL for accessing to the service from the test, or execute commands against container like retrieving an internal file.

And of course as Page Object does, Container Object gives you a way to build a model content that can be reused for several projects.

Before looking at how this pattern is implemented in Arquillian Cube, let’s go thorough an example:

Suppose all of your applications need to send a file to an FTP server. To write an integration/component test you might need a FTP server to send the file and check that the file was correctly sent.
One way to do this is using Docker to start a FTP server just before executing the test, then execute the test using this Docker container for FTP server, before stopping the container check that the file is there, and finally stop the container.

So all these operations that involves the FTP server and container could be joined inside a Container Object. This container object might contain information of:
  • Which image is used
  • IP and bounded port of host where this FTP server is running
  • Username and password to access to the FTP server
  • Methods for asserting the existence of a file
Then from the point of view of test, it only communicate with this object instead of directly hard coding all information inside the test.
Again as in Page Object, any change on the container only affects the Container Object and not the test itself.

Now let’s see how Arquillian Cube implements Container Object pattern with a very simple example:

Arquillian Cube and Container Object

Let’s see a simple example on how you can implement a Container Object in Cube. Suppose you want to create a container object that encapsulates a ping pong server running inside Docker.
The Container Object will be like a simple POJO with special annotations:

In previous example you must pay attention at next lines:
  1. @Cube annotation configures Container Object.
  2. A Container Object can be enriched with Arquillian enrichers.
  3. Bounded port is injected for given exposed port.
  4. Container Object hides how to connect to PingPong server.
@Cube annotation is used to configure this Container Object. Initially you set that the started container will be named pingpong and the port binding information for the container instance, in this case 5000→8080/tcp.
Notice that this can be an array to set more than one port binding definition.

Next annotation is @CubeDockerFile which configure how Container is created. In this case using a Dockerfile located at default classpath location. The default location is the package+classname, so for example in previous case, Dockerfile should be placed at org/superbiz/containerobject/PingPongContainer directory.
Of course you can set any other class path location by passing as value of the annotation. CubeDockerFile annotation sets the location where the Dockerfile is found and not the file itself.
Also this location should be reachable from ClassLoader, so it means it should be loaded from classpath in order to find it.

Any Cube can be enriched with any client side enricher, in this case with @HostIp enricher, but it could be enriched with DockerClient using @ArquillianResource as well.

Finally the @HostPort is used to translate the exposed port to bound port.
So in this example port value will be 5000. You are going to learn briefly why this annotation is important.

And then you can start using this container object in your test:

The most important thing here is that you need to set Container Object as a field of the class and annotate with @Cube.

It is very important to annotate the field with Cube, so before Arquillian runs the test, it can detect that it needs to start a new Cube (Docker container), create the Container Object and inject it in the test.

Notice that this annotation is exactly the same as used when you defined the Container Object.
And it is in this way because you can override any property of the Container Object from the test side. This is why @HostPort annotation is important, since port can be changed from the test definition, you need to find a way to inject the correct port inside the container object.

In this post I have introduced Container Object pattern and how can be used in Arquillian Cube. But this is only an small taste, you can read more about Arquillian Cube and Container Object integration at https://github.com/arquillian/arquillian-cube#arquillian-cube-and-container-object

Also a running examples can be found at https://github.com/arquillian/arquillian-cube/tree/master/docker/ftest-docker-containerobject

We keep learning,

It's time to see what I can do, To test the limits and break through, No right, no wrong, no rules for me, I'm free! (Let It Go - Idina Menzel) 

Music: https://www.youtube.com/watch?v=moSFlvxnbgk