miércoles, octubre 25, 2017

Adding Chaos on OpenShift cluster


Pumba is a chaos testing tool for Docker and Kubernetes (hence OpenShift too). It allows you to add some chaos on containers instances such as stopping, killing or removing containers randomly. But it also adding some network chaos such as delaying, packet loss or re-ordering.

You can see in next screen recording how to add Pumba on OpenShift cluster and add some chaos on it.



The security calls that you need to run before deploying Pumba are the next ones:

Also the Kubernetes script to deploy Pumba in OpenShift is:


I have sent a PR to upstream Pumba project with this file but meanwhile is not accepted you can use it this one.

I'd like to say thank you to Slava Semushin and Jorge Morales for helping me on understanding the OpenShift security model.

We keep learning,
Alex.
Ce joli rajolinet, que les oques tonifique, si le fique en une pique, mantindra le pompis net (El baró de Bidet - La Trinca)
Music: https://www.youtube.com/watch?v=4JWIbKGe4gA

Follow me at https://twitter.com/alexsotob


martes, octubre 24, 2017

Testing Code that requires a mail server

Almost all applications has one common requirement, they need to send an email notifying something to a registered user. It might be an invoice, a confirmation of an action or a password recovery. How to test this use case might be challenging, using mocks and stubs are ok for unit tests, but having a component test that tests the whole stack. 

In this post I am going to show you how Docker and MailHog, can help you on testing this part of code.

MailHog is  super simple SMTP server for email testing tool for developers:
  • Configure your application to use MailHog for SMTP delivery
  • View messages in the web UI, or retrieve them with the JSON API
  • Optionally release messages to real SMTP servers for delivery
  • Docker image with MailHog installed
Notice that since you can retrieve any message sent to mail server using JSON API, it makes really simple to validate if the message has been really delivered and of course assert on any of message fields.

Arquillian Cube is an Arquillian extension that can be used to manager Docker containers in your tests. To use it you need a Docker daemon running on a computer (it can be local or not), but probably it will be at local.

Arquillian Cube offers three different ways to define container(s):
  • Defining a docker-compose file.
  • Defining a Container Object.
  • Using Container Object DSL.
In this example I am going to show you Container Object DSL approach, but any of the others works as well.

To use Container Object DSL you simply need to instantiate the ContainerDslRule (in case you are using JUnit Rules) or use Arquillian runner in case of using JUnit, TestNG or Spock. You can read more about Container Object DSL at http://arquillian.org/arquillian-cube/#_arquillian_cube_and_container_object_dsl

As an example of definition of Redis container:


When running this test, Redis Docker image is started, tests are executed and finally the Docker instance is stopped.

So let's see how to do the same but instead of Redis using Mailhog Docker image.
It is important to notice that ContainerDslRule is a generic class that  can be extended to become more specific to a concrete use case. And this is what we are going to do for Mailhog.

First of all we need to create a class extending from ContainerDslRule, so everything is still a JUnit rule, but a customized one. Then we create a factory method which creates the MailhogContainer object, setting the image name and the binding ports. Finally an assertion method is used to connect to Rest API of Mailhog server to check if there is any message with given subject.

Then we can write a test using this new rule.


This test just configures MailService class with Docker container configuration, it sends an email and finally it delegates to container object to validate if the email has been received or not.

Notice that putting everything into an object makes this object reusable in other tests and even in other projects. You can create an independent project with all your custom developed container objects and just reuse them importing them as test jar in hosted project.

Code: https://github.com/lordofthejars/mailtest

We keep learning,
Alex.
'Cause I'm kind of like Han Solo always stroking my own wookie, I'm the root of all that's evil yeah but you can call me cookie (Fire Water Burn - Bloodhound Gang)




martes, septiembre 19, 2017

Testing code that uses Java System Properties

Sometimes your code uses any Java System Property to configure itself. Usually these classes are configuration classes that can get properties from different sources and one of valid one is using Java System Properties.

The problem is how to write tests for this code? Obviously to maintain your tests isolated you need to set and unset them for each test, restoring old value if for example you are dealing with a global property which is already set before executing tests, and of course in case of an error do not forget to unset/restore the value. So arrived at this point the test might look like:

Notice that this structure should be repeated for each test method that requires an specific system property, so this structure becomes a boilerplate code.

To avoid having to repeat this code over and over again, you can use System Rules project which is a collection of JUnit rules for testing code that uses java.lang.System 

The first thing you need to do is add System Rules dependency as test scope in your build tool, which in this case is com.github.stefanbirkner:system-rules:1.16.0.

Then you need to register the JUnit Rule into your test class.

In this case, you can freely set a System property in your test, the JUnit rule will take care of saving and restoring the System property.

That's it, really easy, no boilerplate code and help you maintaining your tests clean.

Just one notice, these kind of tests are not thread safe, so you need to be really cautious when you use for example surefire plugin with forks. One possible way to avoiding this is by using net.jcip.annotations.NotThreadSafe.class capabilities.

We keep learning,
Alex

Polly wants a cracker, I think I should get off her first, I think she wants some water, To put out the blow torch (Polly - Nirvana)



martes, junio 27, 2017

Lifecycle of JUnit 5 Extension Model


JUnit5 final release is around the corner (currently it is M4), and I have started playing with it a bit on how to write extensions.

In JUnit5, instead of dealing with Runners, Rules, ClassRules and so on, you've got a single Extension API to implement your own extensions.

JUnit5 provides several interfaces to hook in its lifecycle. For example you can hook to Test Instance Post-processing to invoke custom initialization methods on the test instance, or Parameter Resolution for dynamically resolving test method parameters at runtime. And of course the typical ones like hooking before all tests are executed, before a test is executed, after a test is executed and so on so far, a complete list can be found at http://junit.org/junit5/docs/current/user-guide/#extensions-lifecycle-callbacks

But in which point of the process it is executed each of them? To test it I have just created an extension that implements all interfaces and each method prints who is it.


Then I have created a JUnit5 Test suite containing two tests:

So after executing this suite, what it is the output? Let's see it. Notice that for sake of readability I have added some callouts on terminal output.


<1> First test that it is run is AnotherLoggerExtensionTest. In this case there is only one simple test, so the lifecycle of extension is BeforeAll, Test Instance-Post-Processing, Before Each, Before Test Execution, then the test itself is executed, and then all After callbacks.

<2> Then the LoggerExtensionTest is executed. First test is not a parametrized test, so events related to parameter resolution are not called. Before the test method is executed, test instance post-processing is called, and after that all before events are thrown. Finally the test is executed with all after events.

<3> Second test contains requires a parameter resolution. Parameter resolvers are run after Before events and before executing the test itself.

<4> Last test throws an exception. Test Execution Exception is called after test is executed but before After events.

Last thing to notice is that BeforeAll and AfterAll events are executed per test class and not suite.

The JUnit version used in this example is org.junit.jupiter:junit-jupiter-api:5.0.0-M4

We keep learning,
Alex
That's why we won't back down, We won't run and hide, 'Cause these are the things we can't deny, I'm passing over you like a satellite (Satellite - Rise Against)
Music: https://www.youtube.com/watch?v=6nQCxwneUwA

Follow me at

viernes, junio 23, 2017

Test AWS cloud stack offline with Arquillian and LocalStack


When you are building your applications on AWS cloud stack (such as DynamoDB, S3, ...), you need to write tests against these components. The first idea you might have is to have one environment for production and another one for testing, and run tests against it.

This is fine for integration tests, deployment tests, end to end tests or performance tests, but for component tests it will be faster if you could run AWS cloud stack locally and offline.

Localstack provides this feature. It  provides a fully functional local AWS cloud stack so you can develop and test your cloud applications offline.

Localstack comes with different ways to start all stack, but the easiest one is by using Docker image. So if you run atlassianlabs/localstack then you get the stack up and running with next configuration:
  • API Gateway at http://localhost:4567
  • Kinesis at http://localhost:4568
  • DynamoDB at http://localhost:4569
  • DynamoDB Streams at http://localhost:4570
  • Elasticsearch at http://localhost:4571
  • S3 at http://localhost:4572
  • Firehose at http://localhost:4573
  • Lambda at http://localhost:4574
  • SNS at http://localhost:4575
  • SQS at http://localhost:4576
  • Redshift at http://localhost:4577
  • ES (Elasticsearch Service) at http://localhost:4578
  • SES at http://localhost:4579
  • Route53 at http://localhost:4580
  • CloudFormation at http://localhost:4581
  • CloudWatch at http://localhost:4582
So the next question is how do you automate all the process of starting the container, run the tests and finally stop everything and make it portable, so you don't need to worry if you are using Docker in Linux or MacOS? The answer is using Arquillian Cube.

Arquillian Cube is an Arquillian extension that can be used to manager Docker containers in your tests. To use it you need a Docker daemon running on a computer (it can be local or not), but probably it will be at local.

Arquillian Cube offers three different ways to define container(s):
  • Defining a docker-compose file.
  • Defining a Container Object.
  • Using Container Object DSL.
In this example I am going to show you Container Object DSL approach, but any of the others works as well.

The first thing you need to do is add Arquillian and Arquillian Cube dependencies on your build tool.


Then you can write the test which in this case tests that you can create a bucket and add some content using the S3 instance started in Docker host:

Important things to take into consideration:
  1. You annotate your test with Arquillian runner.
  2. Use @DockerContainer annotation to attribute used to define the container.
  3. Container Object DSL is just a DSL that allows you to configure the container you want to use. In this case the localstack container with required port binding information.
  4. The test just connects to Amazon S3 and creates a bucket and stores some content.
Nothing else is required. When you run this test, Arquillian Cube will connect to installed Docker (Machine) host and start the localstack container. When it is up and running and services are able to receive requests, the tests are executed. After that container is stopped and destroyed.

TIP1: If you cannot use Arquillian runner you can also use a JUnit Class Rule to define the container as described here http://arquillian.org/arquillian-cube/#_junit_rule

TIP2: If you are planning to use localstack in the whole organization, I suggest you to use Container Object approach instead of DSL because then you can pack the localstack Container Object into a jar file and import in all projects you need to use it. You can read at http://arquillian.org/arquillian-cube/#_arquillian_cube_and_container_object

So now you can write tests for your application running on AWS cloud without having to connect to remote hosts, just using local environment.

We keep learning,
Alex
Tú, tú eres el imán y yo soy el metal , Me voy acercando y voy armando el plan , Solo con pensarlo se acelera el pulso (Oh yeah) (Despacito - Luis Fonsi)
Music: https://www.youtube.com/watch?v=kJQP7kiw5Fk


jueves, junio 22, 2017

Vert.X meets Service Virtualization with Hoverfly



Service Virtualization is a technique using to emulate the behaviour of dependencies of component-based applications.

Hoverfly is a service virtualisation tool written in Go which allows you to emulate HTTP(S) services. It is a proxy which responds to HTTP(S) requests with stored responses, pretending to be it’s real counterpart.

Hoverfly Java is a wrapper around Hoverfly, that let's you use it in Java world. It provides a native Java DSL to write expectations and a JUnit rule to use it together with JUnit.

But apart from being able to program expectations, you can also use Hoverfly to capture traffic between both services (in both cases are real services) and persist it.

Then in next runs Hoverfly will use these persisted scripts to emulate traffic and not touch the remote service. In this way, instead of programming expectations, which means that you are programming how you understand the system, you are using real communication data.

This can be summarised in next figures:


First time the output traffic is sent though Hoverfly proxy, it is redirected to real service and it generates a response. Then when the response arrives to proxy, both request and response are stored, and the real response is sent back to caller.

Then in next calls of same method:


The output traffic of Service A is still sent though Hoverfly proxy, but now the response is returned from previous stored responses instead of redirecting to real service.

So, how to connect from HTTP client of Service A to Hoverfly proxy? The quick answer is nothing.

Hoverfly just overrides Java network system properties (https://docs.oracle.com/javase/7/docs/api/java/net/doc-files/net-properties.html) so you don't need to do anything, all communications (independently of the host you put there) from HTTP client will  go through Hoverfly proxy.

The problem is what's happening if the API you are using as HTTP client does not honor these system properties? Then obviously all outgoing communications will not pass thorough proxy.

One example is Vert.X and its HTTP client io.vertx.rxjava.ext.web.client.WebClient. Since WebClient does not honor these properties, you need to configure the client properly in order to use Hoverfly.

The basic step you need to do is just configure WebClient with proxy options that are set as system properties.

Notice that the only thing that it done here is just checking if system properties regarding network proxy has been configured (by Hoverfly Java) and if it is the case just create a Vert.X ProxyOptions object to configure the HTTP client.

With this change, you can write tests with Hoverfly and Vert.X without any problem:

In previous example Hoverfly is used as in simulate mode and the request/response definitions comes in form of DSL instead of an external JSON script.
Notice that in this case you are programming that when a request by current service (VillainsVerticle), is done to host crimes and port 9090, using GET HTTP method at /crimes/Gru then the response is returned. For sake of simplicity of current post this method is enough.

You can see source code at https://github.com/arquillian-testing-microservices/villains-service and read about Hoverfly Java at http://hoverfly-java.readthedocs.io/en/latest/

We keep learning,
Alex
No vull veure't, vull mirar-te. No vull imaginar-te, vull sentir-te. Vull compartir tot això que sents. No vull tenir-te a tu: vull, amb tu, tenir el temps. (Una LLuna a l'Aigua - Txarango)
Music: https://www.youtube.com/watch?v=BeH2eH9iPw4

miércoles, mayo 24, 2017

Deploying Docker Images to OpenShift



OpenShift is RedHat's cloud development Platform as a Service (PaaS). It uses Kubernetes as container orchestration (so you can use OpenShift as Kubernetes implementation), but providing some features missed in Kubernates such as automation of the build process of the containers, health management, dynamic provision storage or multi-tenancy to cite a few.

In this post I am going to explain how you can deploy a Docker image from Docker Hub into an OpenShift instance. 

It is important to note that OpenShift offers other ways to create and deploy a container into its infrastructure, you can read more about at https://docs.openshift.com/enterprise/3.2/dev_guide/builds.html but as read in previous paragraph, in this case I am going to show you how to deploy already created Docker images from Docker Hub.

First thing to do is create an account in OpenShift Online. It is free and for the sake of this post is enough. Of course you can use any other OpenShift approach like OpenShift Origin.

After that you need to login into OpenShift cluster. In case of OpenShift Online using the token provided:

oc login https://api.starter-us-east-1.openshift.com --token=xxxxxxx

Then you need to create a new project inside OpenShift.

oc new-project villains

You can understand a project as a Kubernetes namespace with additional features. 

Then let's create a new application within previous project based on a Docker image published at Docker Hub. This example is a VertX application where you can get crimes from several fictional villains from Lex Luthor to Gru.

oc new-app lordofthejars/crimes:1.0 --name crimes

In this case a new app called crimes is created based on lordofthejars/crimes:1.0 image. After running previous command, a new pod running previous image + a service +  a replication controller is created.

After that we need to create a route so the service is available to public internet.

oc expose svc crimes --name=crimeswelcome

The last step is just get the version of the service from the browser, in my case was: http://crimeswelcome-villains.1d35.starter-us-east-1.openshiftapps.com/version notice that you need to change the public host with the one generated by your router and then append version. You can find the public URL by going to OpenShift dashboard, at top of pods definition.


Ok now you'll get a 1.0 which is the version we have deployed. Now suppose you want to update to next version of the service, to version 1.1, so you need to run next commands to deploy next version of crimes service container, which is pushed at Docker Hub.

oc import-image crimes:1.1 --from=lordofthejars/crimes:1.1

With previous command you are configuring internal OpenShift Docker Registry with next Docker image to release.

Then let's prepare the application so when next rollout command is applied, the new image is deployed:

oc patch dc/crimes -p '{"spec": { "triggers":[ {"type": "ConfigChange", "type": "ImageChange" , "imageChangeParams": {"automatic": true, "containerNames":["crimes"],"from": {"name":"crimes:1.1"}}}]}}'

And finally you can do the rollout of the application by using:

oc rollout latest dc/crimes

After a few seconds you can go again to http://crimeswelcome-villains.1d35.starter-us-east-1.openshiftapps.com/version (of course change the host with your host) and the version you'll get is 1.1.

Finally what's happening if this new version contains a bug, and you want to do a rollback of the deployment to previous version? Easy just run next command:

oc rollback crimes-1

And previous version is going to be deployed again, so after a few seconds you can go again to /version and you'll see 1.0 version again.

Finally if you want to delete the application to have a clean cluster run:

oc delete all --all

So as you can see, it is really easy to deploy to deploy container images from Docker Hub to OpenShift. Notice that there are other ways to deploy our application into OpenShift (https://docs.openshift.com/enterprise/3.2/dev_guide/builds.html), in this post I have just shown you one.

Commands: https://gist.github.com/lordofthejars/9fb5f08e47775a185a9b1f80f4af7aff

We keep learning,
Alex.
Yo listen up here's a story, About a little guy that lives in a blue world, And all day and all night and, everything he sees is just blue, Like him inside and outside (Blue - Eiffel 65)
Music: https://www.youtube.com/watch?v=68ugkg9RePc