Back to blog

GSoC Project Intro: External Workspace Manager Plugin

Alexandru Somai
May 23, 2016

About myself

My name is Alexandru Somai. I’m following a major in Software Engineering at the Babes-Bolyai University of Cluj-Napoca, Romania. I have more than two years hands-on experience working in Software Development.

I enjoy writing code in Java, Groovy and JavaScript. The technologies and frameworks that I’m most familiar with are: Spring Framework, Spring Security, Hibernate, JMS, Web Services, JUnit, TestNG, Mockito. As build tools and continuous integration, I’m using Maven and Jenkins. I’m a passionate software developer who is always learning, always looking for new challenges. I want to start contributing to the open source community and Google Summer of Code is a starting point for me.

Project summary

Currently, Jenkins’ build workspace may become very large in size due to the fact that some compilers generate very large volumes of data. The existing plugins that share the workspace across builds are able to do this by copying the files from one workspace to another, process which is inefficient. A solution is to have a Jenkins plugin that is able to manage and reuse the same workspace between multiple builds.

As part of the Google Summer of Code 2016 I will be working on the External Workspace Manager plugin. My mentors for this project are Oleg Nenashev and Martin d’Anjou. This plugin aims to provide an external workspace management system. It should facilitate workspace share and reuse across multiple Jenkins jobs. It should eliminate the need to copy, archive or move files. The plugin will be written for Pipeline jobs.

Usage

Prerequisites

  1. Multiple physical disks accessible from controller.

  2. The same physical disks must be accessible from Jenkins Nodes (renamed to Agents in Jenkins 2.0).

  3. In the Jenkins global configuration, define a disk pool (or many) that will contain the physical disks.

  4. In each Node configuration, define the mounting point from the current node to each physical disk.

The following diagram gives you an overview of how an External Workspace Manager configuration may look like:

ewm config

Example one

Let’s assume that we have one Jenkins job. In this job, we want to use the same workspace on multiple Jenkins nodes. Our pipeline code may look like this:

stage ('Stage 1. Allocate workspace')
def extWorkspace = exwsAllocate id: 'diskpool1'

node ('linux') {
    exws (extWorkspace) {
        stage('Stage 2. Build on the build server')
        git url: '...'
        sh 'mvn clean install'
    }
}

node ('test') {
    exws (extWorkspace) {
        stage('Stage 3. Run tests on a test machine')
        sh 'mvn test'
    }
}

Note: The stage() steps are optional from the External Workspace Manager plugin perspective.

Stage 1. Allocate workspace

The exwsAllocate step selects a disk from diskpool1 (default behavior: the disk with the most available size). On that disk, let’s say disk1, it allocates a directory. The computed directory path is: /physicalPathOnDisk/$JOB_NAME/$BUILD_NUMBER.

For example, Let’s assume that the $JOB_NAME is integration and the $BUILD_NUMBER is 14. Then, the resulting path is: /jenkins-project/disk1/integration/14.

Stage 2. Build on the build server

All the nodes labeled linux must have access to the disks defined in the disk pool. In the Jenkins Node configurations we have defined the local paths that are the mounting points to each disk.

The exws step concatenates the node’s local path with the path returned by the exwsAllocate step. In our case, the node labeled linux has its local path to disk1 defined as: /linux-node/disk1/. So, the complete workspace path is: /linux-node/disk1/jenkins-project/disk1/integration/14.

Stage 3. Run tests on a test machine

Further, we want to run our tests on a different node, but we want to reuse the previously created workspace.

In the node labeled test we have defined the local path to disk1 as: /test-node/disk1/. By applying the exws step, our tests will be able to run in the same workspace as the build. Therefore, the path is: /test-node/disk1/jenkins-project/disk1/integration/14.

Example two

Let’s assume that we have two Jenkins jobs, one called upstream and the other one called downstream. In the upstream job, we clone the repository and build the project, and in the downstream job we run the tests. In the downstream job we don’t want to clone and re-build the project, we need to use the same workspace created in the upstream job. We have to be able to do so without copying the workspace content from one location to another.

The pipeline code in the upstream job is the following:

stage ('Stage 1. Allocate workspace in the upstream job')
def extWorkspace = exwsAllocate id: 'diskpool1'

node ('linux') {
    exws (extWorkspace) {
        stage('Stage 2. Build in the upstream job')
           git url: '...'
           sh 'mvn clean install'
    }
}

And the downstream's pipeline code is:

stage ('Stage 3. Allocate workspace in the downstream job')
def extWorkspace = exwsAllocate id: 'diskpool1', upstream: 'upstream'

node ('test') {
    exws (extWorkspace) {
        stage('Stage 4. Run tests in the downstream job')
        sh 'mvn test'
    }
}
Stage 1. Allocate workspace in the upstream job

The functionality is the same as in example one - stage 1. In our case, the allocated directory on the physical disk is: /jenkins-project/disk1/upstream/14.

Stage 2. Build in the upstream job

Same functionality as example one - stage 2. The final workspace path is: /linux-node/disk1/jenkins-project/disk1/upstream/14.

Stage 3. Allocate workspace in the downstream job

By passing the upstream parameter to the exwsAllocate step, it selects the most recent stable upstream workspace (default behavior). The workspace path pattern is like this: /physicalPathOnDisk/$UPSTREAM_NAME/$MOST_RECENT_STABLE_BUILD. Let’s assume that the last stable build number is 12, then the resulting path is: /jenkins-project/disk1/upstream/12.

Stage 4. Run tests in the downstream job

The exws step concatenates the node’s local path with the path returned by the exwsAllocate step in stage 3. In this scenario, the complete path for running tests is: /test-node/disk1/jenkins-project/disk1/upstream/12. It will reuse the workspace defined in the upstream job.

Additional details

You may find the complete project proposal, along with the design details, features, more examples and use cases, implementation ideas and milestones in the design document. The plugin repository will be available on GitHub.

A prototype version of the plugin should be available in late June and the releasable version in late August. I will be holding plugin functionality demos within the community.

I do appreciate any feedback. You may add comments in the design document. I will be posting updates from time to time about the plugin status on the Jenkins developers mailing list.

link:https://app.gitter.im/#/room/#jenkinsci_external-workspace-manager-title: "Gitter"</a>

About the author

Alexandru Somai

This author has no biography defined. See social media links referenced below.