Iscriviti al feed

Andres Martinez, Principal Developer, has been having a rough time lately. While the root cause analysis following the now-infamous two-day outage acquitted his team’s code of wrongdoing, Martinez has spent enough time around the codebase that he knows a major event caused by his team’s code is not a matter of “If,” but “When.” 

His stress is growing every day - he knows that being the only one with extensive experience and knowledge of the codebase is a single point of failure. He is the single thread holding up the technical-debt-Sword of Damocles. He knows that the thread is at risk of breaking, and soon.

A New Hope?

In his conversations over the past few weeks around process improvement with Chief Architect Daniel Mitchell, he’s heard some great initial success stories from the adoption of Infoblox and F5 automation. He finds himself encouraged by their successes -  if Ops can do it, so can we! 

So he begins to craft a plan to modernize the DSE inventory application. Of course, revamping the entire application stack would take way more time and money than they had available. The increasing fragility of the application environments meant his team spends almost all of their time just keeping the lights on. During his discussions with Mitchell, he realizes that he doesn’t have to modernize everything at once, but he needs to start somewhere.

It’s a question with - for once - an easy answer: they need to modernize their version control in the same way the operations team has modernized its version control. Each development team had their own preferred repository, as well as their own preferred way of managing the roll-out of new features due to a lack of coordination during development. This led to three separate git repositories being used -- in addition to a subversion repository that had somehow managed to survive. 

Martinez considers how he might be able to leverage Git to reduce the application’s technical debt - but it quickly becomes apparent to him that it’s not limited to his team’s code. He sees the same potential train wreck happening with the new infrastructure work, so whatever solution they implement needs to be scalable across all aspects of the company. After reviewing his options, he decides to launch a pilot program with GitLab on Red Hat OpenShift.

What is GitOps?

During a brainstorming session in the Dev Team Room, one of Martinez’s direct reports asks him, “What’s GitOps? Git-driven DevOps?” 

“Yes, exactly - let me show you.” He pulls up a diagram on his laptop:

“GitOps is based on three principles: Infrastructure-as-Code (or IaC), merge requests, and CI/CD. Let’s start with IaC - although in reality, it’s more like ‘anything-as-code,’ since in an ideal world, anything that is a policy, config, or any other operations work would be stored in code. The great thing about IaC is that it declares the desired state of your infrastructure in a Git repository, which means you can integrate it into your existing git workflows easily. This in turn leads to way more visibility on changes to infrastructure. More eyes equal fewer chances of a flaw leaking into the system.”
“Now that you’ve got a repository configured for your IaC, it’s time to consider merge requests. They’re your gateway to acting on a change in code. merge requests provide ‘gating’ and collaboration capabilities, code reviews, discussions, and even approvals, allowing many people to contribute to a given change, but limiting the number of people that can approve a change. This helps with both auditing and compliance.”
“Finally, there’s CI/CD. Those letters seem to have a different meaning every day of the week, but for simplicity’s sake, we’re going to say they stand for Continuous Integration (CI) and Continuous Delivery (CD). GitLab provides a runner operator that we can leverage in OpenShift as an agentless-based reconcile loop. When the infrastructure state is out of sync with whatever is defined, the runner will act in order to update configurations to match the definition in the repository.”
“So if we look at the diagram, your Infrastructure-as-Code is modified. In our case, with our existing GitLab Flow process, we first create merge requests, which use feature branches, to do all the work. Merge requests are where all the collaboration takes place among stakeholders. Once it’s reached a state where you as a developer think it’s ready for review, you commit your work to the branch. At this point, an ephemeral review environment is automatically stood up and the updated application is deployed to it so that stakeholders can verify the updates. If the merge request is approved, the feature branch is merged to the main branch and a job automatically kicks off using the appropriate runner, and if all the tasks of that job succeed, the new change is ready to be deployed. Typically this last step is a manual one, but depending on the circumstances it could be automated as well.”
“If you don’t have any other questions, I think it’s time we dive in and see what we can get done.”
Martinez opens up his laptop and connects it to the team room’s big monitor.
“Let’s begin.”

Configuring Merge Request-based pipelines in GitLab

Martinez opens a local copy of the inventory application and creates a new branch. He then creates a new file in the root directory of the project called .gitlab-ci.yml. He adds this code to the file:

  image: node:latest
      - node_modules

  - prepare_env
  - build
  - package
  - deploy
  image: daviestreetent/gitlab-ci-openshift
    - oc login $OC_URL --token=$OC_TOKEN --insecure-skip-tls-verify
    - oc project $OC_NAMESPACESLUG-dev

"Prepare OpenShift Environment (Dev)":
  extends: .openshift_job
  stage: prepare_env

    - oc apply --filename=./.openshift/dev --recursive=true --wait=true -n $OC_NAMESPACESLUG-dev
    - if: '$CI_COMMIT_REF_NAME == "master"'
      when: always
    - when: never

"Prepare OpenShift Environment (Staging)":
  extends: .openshift_job
  stage: prepare_env
    - oc login $OC_URL --token=$OC_TOKEN --insecure-skip-tls-verify
    - oc project $OC_NAMESPACESLUG-staging
    - oc apply --filename=./.openshift/staging --recursive=true --wait=true -n $OC_NAMESPACESLUG-staging
    - if: '$CI_COMMIT_REF_NAME == "staging"'
      when: always
    - when: never

"Prepare OpenShift Environment (Production)":
  extends: .openshift_job
  stage: prepare_env
    - oc apply --filename=./.openshift/prod --recursive=true --wait=true -n $OC_NAMESPACESLUG-prod
    - if: '$CI_COMMIT_REF_NAME == "prod"'
      when: always
    - when: never

  stage: build
  image: node:latest
    - npm install
  allow_failure: true
    - node_modules
    expire_in: 1 week

"Package Image on OCP":
  extends: .openshift_job
  stage: package
    - oc start-build openshift-spring-custom --follow --from-dir .

"Deploy Application on OCP":
  extends: .openshift_job
  stage: deploy
  when: manual
    - oc project $OC_NAMESPACESLUG-dev
    - oc rollout latest dc/openshift-test-app

“Ok, so that’s a big file - let’s break it down. What we’re doing is establishing some defaults, configuring the stages our pipeline will contain, specifying some global configurations, and then for each stage, we describe the job we want to run. You can see a few common elements from each of these jobs."

"First, they have a human-readable label. Then we see they extend the defaults stage we defined earlier, and define when in the pipeline the job is run. Now we specify our scripts to run for the job, and then optionally establish rules - as an example in the cases above, you can see that the job "Prepare OpenShift Environment (Production)": would only be run if the branch we’re merging into was labeled “prod” - this allows you to be more or less picky about what occurs on different environments."

"In the case of this pipeline, we’re taking the API objects described in the .openshift directory and the appropriate subdirectory and running an oc apply on them. If the apply succeeds, the job proceeds to the next stage, where we build and package a new iteration of the application. The final stage, deploy, is configured to require a manual action in order to roll out the new project.”

“So now that we’ve got that out of the way, let’s commit our pipeline code and see what it looks like on GitLab.”

He commits the code, and pulls up the repository on GitLab. It looks like this:

“Looks like the job failed - that’s completely normal, because we haven’t installed or configured a runner for our project. Let’s do that now.”

Installing the GitLab Runner Operator

“First, we need to install the Operator, which we do through the OpenShift Console.”

“After clicking on the Runner Operator, we’re asked to specify a few things for the installation. We’ll choose a specific namespace to install the operator for, so we can keep our workstreams clean.”

“After clicking on ‘Install,’ the Runner Operator should be online within a few minutes.”

“Now we need to tell GitLab about our runner, and establish a connection. To do this, we’ll need to head to the GitLab project’s Settings page, and click on ‘CI/CD.’ Next we expand the ‘Runners’ section. You can see that we have an option to ‘Set up a specific runner manually.’ We will need to copy the registration token and set up a secret in our OpenShift project.”

 “Using the OpenShift command line makes this pretty straightforward. We just log in to our project…”


“And then create a secret file with the token we just copied.”

cat > gitlab-runner-secret.yml <<< EOF
apiVersion: v1
kind: Secret
  name: gitlab-runner-secret
type: Opaque
  runner-registration-token: YOUR_TOKEN_HERE # your project runner secret

oc apply -f gitlab-runner-secret.yml

"Now let’s create the runner itself.”

cat > gitlab-runner.yml <<< EOF
kind: Runner
  name: example
  imagePullPolicy: Always
  gitlabUrl: ‘  tags: ‘openshift, test’
  token: gitlab-runner-secret

oc apply -f gitlab-runner.yml
“Just to be sure, let’s confirm that the GitLab runner and its pod are both running:
oc get runners
NAME               AGE
gitlab-runner      5m

oc get pods
NAME                                 READY   STATUS     RESTARTS     AGE
gitlab-runner-bf9894bdb-wplxn        1/1     Running    0            5m

“Now that it’s reporting as running, we can check back with GitLab to see that the Runner has phoned home and is now available.”

Deploying a requested change using the GitLab Runner Operator

“Let’s test it out. I’ve modified the version number for NodeJS for our application - take a look at the following commit:”

“Looks good - let’s now create a merge request, and approve it. After a few minutes, we can start to see our pipeline jobs have all successfully passed, up to the final ‘deploy’ stage, which requires manual intervention:”

“Take a peek behind the scenes - this is what the Prepare_env job is reporting. You can see that the new configurations have been successfully applied, specifically in this case, the build config.”

“All we need to do now is deploy - we’ll click on the ‘play’ button to kick it off... Looks like it’s running nicely!”

“And now it’s finished. If we look in the logs for the build we can see that it did indeed rollout the new version of our application. Where’s my ‘Easy Button?’”

Tune in next time…

A few days and several pizzas later, Martinez’s team steps away from their keyboards and joins him in a conference room. Already seated at the table is Zachary L. Tureaud, Director of Security Engineering. Martinez motions to his team to take a seat.

“I think we can all agree that we’ve had amazing success in improving our development pipeline. I have already seen an increase in velocity through some of the issues that have been worked on for the past couple of days. It seems evident that the new pipeline has taken a very large bite out of our workload, and freed us up to move faster and even have some time to brainstorm.”

“It’s because of this success that I’ve asked Zachary to take a look at what we’ve done to see if we might be able to leverage this system to increase our security as well. Our relationship with the security team has not been well-maintained over the past few years. We haven’t had the time to work with them on ensuring we maintain our security, and need to do a better job shifting that security burden left. I’ll let Zachary explain further.”

Tureaud stands up and grabs a whiteboard marker. “Alright, is everyone sitting comfortably? Then we’ll begin.”

To be continued

If you’d like to see a step-by-step demonstration of the methods described in this post, please take a look at the video on our YouTube channel.



Phillip Lamb is the DevOps Solution Architect for Red Hat’s Global Partners ISV team. He has more than 15 years of development experience, and led a GitOps-based digital transformation effort with the State of North Carolina’s Department of Information Technology prior to joining Red Hat. He is currently based out of Raleigh, North Carolina, after moving from France with his wife and five children. In his spare time, he is an avid outdoorsman but also enjoys playing Minecraft with his kids and building weird things on the Raspberry Pi platform.
Read full bio

Ricerca per canale

automation icon


Novità sull'automazione IT di tecnologie, team e ambienti

AI icon

Intelligenza artificiale

Aggiornamenti sulle piattaforme che consentono alle aziende di eseguire carichi di lavoro IA ovunque

open hybrid cloud icon

Hybrid cloud open source

Scopri come affrontare il futuro in modo più agile grazie al cloud ibrido

security icon


Le ultime novità sulle nostre soluzioni per ridurre i rischi nelle tecnologie e negli ambienti

edge icon

Edge computing

Aggiornamenti sulle piattaforme che semplificano l'operatività edge

Infrastructure icon


Le ultime novità sulla piattaforma Linux aziendale leader a livello mondiale

application development icon


Approfondimenti sulle nostre soluzioni alle sfide applicative più difficili

Original series icon

Serie originali

Raccontiamo le interessanti storie di leader e creatori di tecnologie pensate per le aziende