GitOps for complex microservices

CI/CD pipelines have a disadvantage

Many organisations use Continuous Integration (CI) and Continuous Deployment (CD) pipelines to build and deploy their applications in Kubernetes clusters, and to ultimately make IT resources available faster. While for a client in the automotive industry CI pipelines were well suited to create immutable artefacts from source code, the use of pipelines for CD had two major disadvantages:

• The pipelines operated on a “fire and forget” principle: a pipeline did not track deployment artefacts (in this case Kubernetes resources) after it expired

• Adding other applications, it became difficult to keep the pipelines in a maintainable condition

Reducing deployment complexity

To make the provisioning of IT resources faster, the client needed to reduce the complexity of service delivery. The client therefore contracted Liquid Reply to achieve improved visibility of the deployment status and to eliminate the environment drift that occurs during manual hotfixes.

The client’s goals

• A reduced deployment complexity of highly dynamic microservices

• The prevention of environment drift

• Establishing live visibility of deployment status and enabling progressive deployment

A GitOps solution

In collaboration with the client, Liquid Reply implemented GitOps solutions such as Flux and ArgoCD across various projects.

A key challenge was to make GitOps interoperate with the client’s source code repositories. As is common in large companies, these were not designed for high interactivity and were not yet part of a dynamic agile system.

Alignment with security requirements

On an organisational level, friction arose with GitOps, as the approach requires access to source code repositories from within a productive system, which is often, if not always, prohibited by security requirements.

Liquid Reply supported the client not only in the technological adaptation, but also in the necessary organisational adjustments.

The achievements

By replacing the CD pipelines with Kubernetes-native GitOps tools, Liquid Reply successfully reduced the overall complexity in the client’s IT. Visibility and confidence in deployment grew, as did confidence in the internal IT. Within the organisation, the alignment with security requirements paved the way for further use of the GitOps paradigm.

What this teaches about GitOps adoption

The technical migration from CD pipelines to Flux and Argo CD was the smaller half of this project. The larger half was organisational: repository structures that were never designed for continuous reconciliation, and security policies that had to be renegotiated before a controller in a productive cluster was allowed to read source repositories at all. Both are typical for large organisations, and both are solvable — but only if they are treated as part of the project scope rather than as obstacles discovered halfway through.

The pattern transfers well beyond automotive: wherever many microservices are deployed through hand-maintained pipelines, replacing push-based deployment with declarative reconciliation reduces maintenance effort, makes the deployment state auditable, and turns a rollback from an emergency procedure into a git revert.