The road to continuous deployment (PHPCon Poland 2016)

Below are the slides of my talk “The road to continuous deployment: a case study”, as presented at PHPCon Poland in October 2016.

It’s a situation many of us are familiar with: a large legacy application, limited or no tests, slow & manual release process, low velocity, no confidence…. Oh, and management wants new features, fast.

But how to proceed? Using examples and lessons learned from a real-world case, I’ll show you how to strangle the legacy application with a modern service architecture and build a continuous deployment pipeline to deliver value from the first sprint. On the way, we take a look at testing strategies and various (possibly controversial!) tips and best practices.

CD: Continuous Delivery or Continuous Deployment?

I’m writing this post to contribute my take on the acronym CD, and the distinction (or perhaps, confusion) between the phrases Continuous Delivery and Continuous Deployment. These phrases are used interchangeably, and sometimes incorrectly, in various books, blogs and talks. And while these software engineering approaches (or disciplines) share a lot, there is one key difference.

Read more

The road to continuous deployment: a case study

Below are the slides of my talk “The road to continuous deployment: a case study”, as presented at the Dutch PHP Conference in June 2016. I’m planning to dedicate a series of blogposts to this particular topic later this year.

It’s a situation many of us are familiar with: a large legacy application, limited or no tests, slow & manual release process, low velocity, no confidence…. Oh, and management wants new features, fast.

But how to proceed? Using examples and lessons learned from a real-world case, I’ll show you how to strangle the legacy application with a modern service architecture and build a continuous deployment pipeline to deliver value from the first sprint. On the way, we take a look at testing strategies and various (possibly controversial!) tips and best practices.

Building and deploying Java WebSphere applications with Jenkins CI

Jenkins CI (the new name of Hudson) is a very popular continuous integration system. It can be used to monitor the execution of various jobs, including but not limited to compilation, packaging, testing and deploying of software. Also, it is very easy to configure and comes with a great set of (3rd party) plugins.
I use Jenkins in a number of ways: to monitor and prepare & test new releases of Phing, monitor various internal processes (such as backup logs), and build and deploy various other projects that I work on.

In this post I will expand on some of the techniques discussed in an earlier IBM developerWorks article, to (automatically) build and deploy Java J2EE applications to a WebSphere server. The code fragments listed below are contained in a downloadable archive that you’ll be able to download at the end of this post.

Requirements

To get started, you’ll need to have installed:

  • Jenkins CI with the following plugins (can be installed via “Manage Jenkins” -> “Manage Plugins”):
    • Copy Artifact
    • Blame Subversion
    • Parameterized Trigger
    • RAD Builder
  • Ant Contrib
  • IBM Rational Application Developer
  • A test/staging installation of the WebSphere Application Server

This post assumes you have some knowledge of Ant, and are able to install Jenkins and IBM RAD.

Job configuration

For this particular case we will configure two Jenkins CI jobs: one job will build number of artifacts (in this case, .ear files) from source code contained in a version control repository, and another job will deploy the generated artifacts to a WebSphere server. This deployment job will be triggered when the build job completes (successfully).

Jenkins CI Dashboard
Jenkins CI Dashboard

Build job

Create a new “free-style” job, and configure it as you normally would. Make sure you check out the source code to the src directory, within the job workspace.

Build Job - SVN config
Build Job - SVN config

Then, click “Add build step”, and select the IBM RAD plugin. The field “build file” should contain the path to the build file we will use (Builder\build.xml in the archive). The field “RAD workspace” points to a directory (within the job’s workspace) where a RAD workspace will be created, in this case (see the build file below) we use the path “rad-workspace”. The other settings can be left at their default values.

Build Job - RAD builder
Build Job - RAD builder

Build file

The Jenkis RAD builder plugins creates a fresh workspace, similar to the workspace that is used inside RAD (or Eclipse). To prepare this workspace with the right configuration settings, we use the task workspacePreferenceFile. The input for this task is a simple preferences file, either text format (key=value pairs, see sample), or the Eclipse .epf format.

The task workspacePreferenceFile is then called in the setup-workspace target.

Next, the code that has previously been checked out by Jenkins will need to be copied to this new workspace. The properties copy.from.path and copy.excludes (optional, comma-separated list of excluded patterns) are set in the IBM RAD builder configuration (build job).

Now that the workspace is configured and contains the projects we’d like to build, it’s time to make RAD aware of the contents by actively importing each project – this is done by calling the projectImport task. The list of projects is generated by scanning the workspace for directories that contain a .project file.

The most important part of the build file is the target build-workspace, which calls the task workspaceBuild to perform a full build. By default, this task will fail the build if any (compiler) errors are encountered – this is what we want.

Hopefully, there are no errors, and we are in a situation where all the projects have been built successfully. Time to generate some artifacts!
The target export-ear first updates the (generated) manifest file with a few Jenkins parameters, such as build number, SVN revision, job name, and the current date. This data is a useful (extra) aid to identify the version / origin of deployed code (please note that you can also use the fingerprinting functionality for this, see below).

We then call the earExport task to create a .ear file, identical to choosing “Export” -> “EAR file” within RAD.

When the RAD builder finishes succesfully, the build part of the job is completed and a number of artifacts (.ear files) will have been generated.

Build job - post-build actions
Build job - post-build actions

In the post-build actions we make sure the generated artifacts are scooped up and archived. This makes sure that artifacts are kept even if the original build was (re)moved. Additionally, we enable the recording of fingerprints on each artifact. In essence, this will calculate and store a hash value (MD5 or similar) based on the contents of each file. Should we need to identify a particular artifact at some point in the future, we can simply upload that file to Jenkins, let it calculate a hash value, and match that hash value against its internal fingerprint database. If there’s a match, Jenkins will tell us the job name, build number, date, and any other useful information.

Finally, we call the deploy job using the parameterized trigger plugin. In this case, we do not override any of the default parameters (see below). Should you want to, click “Add parameter”, then “Predefined parameters”. Enter the parameters (key=value pairs) in the text area.

Deploy job

As stated before, the deployment job copies generated artifacts from the build job, and installs the artifacts on a (test/staging) WebSphere server. To achieve this, the job calls the wsadmin tool and executes a single JACL script.

An important part of this job are the predefined parameters, telling the JACL script which SOAP connection to use, and which node / cell / server name / virtual host to install the application to. In this case, each of these parameters has a default value – pointing to a default (local) testing server.

Deploy Job - Build parameters
Deploy Job - Build parameters

The build phase of the job consists of three separate build steps:

  • Remove any artifacts that were left by previous builds
  • Copy the artifacts generated by the last successful run of the build job
  • Execute ws_ant (Ant with WebSphere functionality/classes preloaded), which in turn uses wsadmin to run a JACL script.
Deploy Job - Build steps
Deploy Job - Build steps

The JACL script has two modes of operation. First, it stops and uninstalls the previous version of the application we are trying to install. Errors that occur during this first part are ignored.

set appManager [$AdminControl queryNames cell=$cell,node=$node,type=ApplicationManager,process=$server,*]

catch { $AdminControl invoke $appManager stopApplication $appname } result

$AdminConfig save

$AdminApp uninstall $appname

$AdminConfig save

In the second part of the script, the application is installed on the specified node/cell/server/virtual host. Then, after giving the application server some time to process the installed artifact, the script starts the application. If this completes without errors the application is ready to use!

Conclusion / thoughts

In this post you’ve seen how to use Jenkins CI to build (through IBM RAD) and deploy (through IBM wsadmin) a J2EE application to a WebSphere server. I hope these exampless can serve as a starting point for your forays into the exciting world of Jenkins CI.

Comments and suggestions are very welcome!

Downloads