Web Server Load Balancing with NGINX Plus

When F5’s acquisition of NGINX was finalized in May of this year, one of the first priorities for Engineering was to integrate the control‑plane development teams and adopt a common set of tools. In this post, I describe the changes we made to our code repository, development pipeline, and artifact storage.

Before the acquisition, the NGINX Controller team’s tool suite included GitHub for the code repo, Jenkins for the development pipeline, and Nexus Repository Manager OSS, Docker Registry, and Amazon S3 for artifact storage. The control‑plane team at F5 was using GitLab with its built‑in continuous integration feature (GitLab CI) for the code repo and pipeline, and JFrog Artifactory for all artifacts.

We saw lots of value in adopting the F5 tech stack for the control‑plane team as a whole. First, it’s simpler than the stack the NGINX Controller team was using. Second, the company’s IT team manages it, freeing DevOps teams from having to deal with infrastructure.

Moving the Code Repository

The first step was to move the NGINX Controller code repository from GitHub to GitLab. This turned out to be trivial, requiring just a few scripted git clone and git push operations. We were able to preserve all of the branch rules and merge checks we had been using in GitHub.

Converting to the GitLab Pipeline

This was the most important step, as it enables the integrated engineering team to build, test, and publish code.

We had defined the pipelines for NGINX Controller in Jenkinsfiles written in the Jenkins domain‑specific language (DSL). We had tried to make them as Jenkins‑agnostic as possible, though, by:

  • Storing secrets in HashiCorp Vault wherever possible
  • Limiting the use of plug‑ins
  • Running most commands explicitly in the pipeline, without wrappers
  • Making it possible to run all jobs locally on a developer’s laptop in the same way as on the Jenkins server

The process of translating Jenkinsfiles to GitLab CI pipelines was then mostly a matter of creating GitLab Runners that matched our Jenkins workers, and creating pipelines with the same stages by copying the commands from the previous pipelines.

The biggest task was to change the Jenkins environment variables to GitLab CI variables and to verify all the paths and file locations. To avoid having to do this, I recommend always representing paths as variables. I know that’s obvious, but sometimes we sacrifice best practices for the sake of speed.

Here’s an example of one of our Jenkinsfile stages:

stage ('Test & PushToReg') {
    environment {
        NODE_ENV = 'production'
    steps {
        sh '''
        VERSION=$(git tag --contains | head -1 | tr -d v )
        CTR_VERSION=${VERSION} gulp build
        sh "PLUGIN_NAME=NAME1 yarn webpack-cli --config webpack/platform/plugin.js"
        sh "PLUGIN_NAME=NAME2  yarn webpack-cli --config webpack/platform/plugin.js"
        sh "docker build -t /${BRANCH}:${BUILD_NUMBER} -f docker/Dockerfile ."
        sh "echo 'Running unit tests'"
        sh "yarn test"
        stash includes: 'test-results.xml', name: 'utest', allowEmpty: true
        sh "docker push …..”

This was translated to the following Gitlab CI code:

  stage: build_test_publish
  image: ${BUILD_IMAGE}
    - devops-runner
  script: |
    yarn config set registry ${NPM_REGISTRY}
    yarn install --verbose > yarn.log
    yarn autoclean
    VERSION=$(git tag --contains | head -1 | tr -d v )
    CTR_VERSION=${VERSION} gulp build
    PLUGIN_NAME=NAME1 yarn webpack-cli --config webpack/platform/plugin.js
    PLUGIN_NAME=NAME2  yarn webpack-cli --config webpack/platform/plugin.js
    # Pushing to different locations depending on the branch
    if [[ ${CI_COMMIT_REF_SLUG} == "release-"* ]]; then
    docker build -t ${DOCKER_REG}/${CI_COMMIT_REF_SLUG}:${CI_PIPELINE_ID} -f docker/Dockerfile .
    echo 'Running unit tests'
    yarn test
    docker login ….
    docker push ….
    expire_in: 1 day
    when: always
      - yarn.log
      - test-results
      Junit: test-results.xml

Storing Artifacts

The NGINX Controller team was using separate tools to store different types of artifact:

  • Nexus for deb and rpm packages
  • Amazon S3 for tar balls
  • A private Docker registry for container images

Moving to the F5 infrastructure gave us the opportunity to consolidate all storage under Artifactory, so we went for it.

Artifactory natively supports all of the repository types we were using, so we just needed to configure the repository location and credentials in our new pipelines and we were ready to start publishing.

The are three big advantages to having just one Artifactory repository. Obviously, it makes it management easier. It also means we can scan for vulnerabilities and license compliance in just one place. Lastly, access control is centralized.

Hero image
Managing Kubernetes Traffic with F5 NGINX: A Practical Guide

Learn how to manage Kubernetes traffic with F5 NGINX Ingress Controller and F5 NGINX Service Mesh and solve the complex challenges of running Kubernetes in production.

About The Author

Ismael Serrano

DevOps Engineer

About F5 NGINX

F5, Inc. is the company behind NGINX, the popular open source project. We offer a suite of technologies for developing and delivering modern applications. Together with F5, our combined solution bridges the gap between NetOps and DevOps, with multi-cloud application services that span from code to customer.

Learn more at or join the conversation by following @nginx on Twitter.