GitOps Workflows Expanding Beyond Kubernetes Clusters

GitOps is emerging as a delivery operating model rather than a Kubernetes-specific convenience, and Fluxโ€™s graduation under the CNCF signals that the method is moving from early adopter practice toward broader infrastructure discipline. The editorial point is not that declarative workflows are novel, but that the center of gravity is shifting from cluster management to a more general approach for reconciling software, infrastructure, and templates from version-controlled sources.

The practical mechanism is controller-based automation: Flux continuously reconciles desired state from repositories, and its integrations with Terraform, Pulumi, and Argo CD show how the same pattern can be extended across infrastructure-as-code and application delivery tools. That matters because many organizations already use CI to build software but have not achieved reliable CD. Loosening the coupling between build and release can reduce reliance on manual installation processes, which do not scale as application backlogs grow. The added interest in IDE-based access also reflects a push to move GitOps closer to day-to-day developer workflows.

The limitations are equally important. The article makes clear that extension beyond Kubernetes is still uneven, depends on integration work, and faces a learning curve that many organizations have not yet absorbed. The enterprise opportunity is tied to that gap, not to guaranteed ubiquity. The significance for practitioners is that GitOps should be judged by whether it can standardize delivery across platforms and lower operational friction, not by promotional claims about universal automation.


Now that the open source Flux version control software has officially graduated to become a mature project with the Cloud Native Computing Foundation (CNCF), the pace at which organizations embrace GitOps best practices should accelerate. GitOps, at its core, defines a declarative approach for pulling software from a repository in a way that is both continuously reconciled and immutable. In that context, GitOps is an opinionated instance of a DevOps workflow that unifies both software deployment and the provisioning of infrastructure using code to create reusable templates. Most of the adoption of GitOps thus far has coincided with adoption of Kubernetes clusters that expose a standard set of application programming interfaces (APIs) that make it simpler to deploy code. Many of the IT teams that have adopted Kubernetes are using dedicated continuous delivery (CD) platforms along with Fluxโ€™s version management capabilities to manage software delivery rather than extending a continuous integration (CI) tool to deploy software. Weaveworks CEO Alexis Richardson said the next GitOps frontier would be to integrate the controllers within platforms such as Flux with other controllers to make it simpler to employ the same platform and processes to deploy software and infrastructure-as-code (IaC) templates on platforms other than just Kubernetes clusters. For example, Weaveworks has worked to integrate Flux with Terraform controllers to unify the management of IaC and application delivery. Now the same approach is being used to integrate Flux with other IaC tools such as Pulumi and the Argo CD platform, notes Richardson. At the other end of the spectrum, more work must be done to make GitOps workflows accessible to developers directly from within the integrated development environments (IDEs) they use to create software, he adds. It remains to be seen how long it might take for GitOps workflows to be extended beyond Kubernetes environments, but as the number of those integrations increases Weaveworks is betting there will more demand for an enterprise edition of a GitOps platform based on Flux. The challenge, of course, is that while many organizations regularly make use of CI to construct a build of an application, the number of organizations that have successfully implemented CD remains relatively small. By more loosely coupling CI and CD processes it should, in theory, become easier for more organizations to automate the delivery of software. Of course, the rate at which applications are currently being constructed will force the CD issue within many organizations. The current manual processes used to install software on platforms today simply doesnโ€™t scale. Thereโ€™s no doubt thereโ€™s a GitOps learning curve but, like it or not, most organizations are not going to allow a growing backlog of applications waiting to be deployed. In fact, organizations simultaneously realize just how dependent they are on software to drive digital processes and that the amount of time it takes to install applications and deliver subsequent updates has a material impact on both the top and bottom line.
https://containerjournal.com/features/gitops-workflows-expanding-beyond-kubernetes-clusters/

Enjoyed this article? Sign up for our newsletter to receive regular insights and stay connected.