From a Vanished Blog Post to a CNCF Standard: The Origins and Reality of OpenGitOps
In the summer of 2017, the startup Weaveworks published a post titled “GitOps—Operations by Pull Request” on its official blog. Within a few short years, the term swept the cloud-native ecosystem, coalescing into the cross-industry OpenGitOps open-source standard, while Flux and Argo CD each grew into CNCF graduated projects.
Yet in early 2024, Weaveworks shut down due to commercial struggles, taking the original blog offline. Today, that foundational post can only be accessed via the Wayback Machine. The standards and tooling it catalyzed, however, have taken deep root in the community and continue to thrive.
A 45-Minute Disaster Recovery: The Origins of GitOps
On August 7, 2017, Weaveworks CEO Alexis Richardson published the article that formally coined the term “GitOps.”
The most compelling case study in the post was drawn from real-world experience: Weaveworks had wiped its entire infrastructure—including AWS resources and Kubernetes clusters—and then, relying solely on declarative descriptions stored in Git, rebuilt the entire system from scratch in just 45 minutes.
Richardson summarized the core ethos of this operational model:
By using Git as our source of truth, we can operate almost everything.
Under this architecture, all environment changes are driven through pull requests and automated pipelines.
Architecturally, GitOps extends Kubernetes’ foundational control loop (the reconcile loop) outward. Inside a Kubernetes cluster, an accidentally deleted Pod respawns on its own; Weaveworks extended this control mechanism beyond the cluster to a Git repository, continuously comparing actual state against desired state and reconciling the difference.
Semantic Dilution and Community Convergence: The Four Principles of OpenGitOps
As the concept gained traction, the industry ran headlong into the classic problem of semantic dilution. Major cloud vendors rushed to declare GitOps support, but their definitions varied wildly—even traditional setups running kubectl apply directly inside CI scripts were rebranded as GitOps.
To clear up the confusion, the community formed the GitOps Working Group in late 2020 (later launching the OpenGitOps project) and brought it into the CNCF Sandbox in 2021. In October 2021, the working group officially published the GitOps Principles v1.0.0—an effort backed by more than 60 companies and 96 interested parties—replacing vague marketing rhetoric with four verifiable criteria:
| # | Principle | Core Requirements and Mechanisms |
|---|---|---|
| 1 | Declarative | The system’s desired state must be expressed declaratively rather than through imperative steps, ensuring an unambiguous record of what the system should look like at all times. |
| 2 | Versioned and Immutable | The desired state is stored in a medium that enforces version control and preserves a complete history, providing a comprehensive audit trail. |
| 3 | Pulled Automatically | Software agents running inside the target system automatically pull the desired state, upending legacy architectures where external pipelines held privileged credentials to push changes. |
| 4 | Continuously Reconciled | Software agents continuously observe the actual system state and proactively remediate drift, ensuring the live environment constantly converges toward the desired state. |
The third principle uses the phrase from the source rather than mandating Git. Nothing in the wording of the four principles restricts the storage medium to Git tools; any store that is versioned, immutable, and retains a complete history qualifies. Git is simply the industry’s de facto standard.
The accompanying glossary also clearly demarcates the boundaries of applicability:
NOTE
Persistent data is excluded from the GitOps loop: The OpenGitOps Glossary defines desired state as the aggregate configuration data sufficient to recreate the system, explicitly noting that this typically excludes persistent data such as database contents. Stateful database changes are not protected by the GitOps reconciliation loop and must be handled by dedicated migration mechanisms.
Putting Principles into Practice: From Architectural Inversion to Tool Maturity
When the four principles are translated into architecture, the most significant shift is the inversion of the security model. In a traditional push model, external CI servers must hold credentials for production clusters and require inbound network access into them. In contrast, GitOps relies on an agent inside the cluster initiating outbound connections to pull configurations, keeping credentials strictly within cluster boundaries.
Under this architecture, day-to-day operations align directly with Git workflows. If an engineer manually tweaks live settings, the agent detects configuration drift and either remediates it automatically as configured or fires an alert. If a new deployment misbehaves, engineers simply run git revert; in the next reconciliation cycle, the agent automatically restores the environment to the previous state (excluding persistent data like databases).
NOTE
Further reading: A Practical Guide to GitOps for Kubernetes Users
The “software agent inside the target system” defined by the principles has two primary implementations in the Kubernetes ecosystem: Flux, developed by Weaveworks, and Argo CD, open-sourced by Applatix and later championed by Intuit. As adoption deepened, the tooling ecosystem progressed from early experimentation to enterprise maturity:
| Timeline | Milestone |
|---|---|
| Mid-2016 | Weaveworks starts Flux |
| August 2017 | Alexis Richardson coins the term GitOps |
| 2018 | Argo CD launches |
| July 2019 | Flux enters the CNCF Sandbox |
| Late 2020 | GitOps Working Group is formed |
| October 2021 | OpenGitOps Principles v1.0.0 is released |
| Late 2022 | Flux and Argo CD graduate from the CNCF |
| February 2024 | Weaveworks ceases operations; open-source community takes over Flux |
The CNCF categorizes open-source project maturity into three tiers: Sandbox for early-stage experimentation, Incubating for projects with active communities and real-world adoption, and Graduated for those that have passed rigorous production validation and established independent governance. Both Argo CD and Flux graduated in late 2022, solidifying a duopoly in the ecosystem.
In recent adoption surveys by the CNCF and Octopus Deploy, Argo CD holds the lead with roughly 50% to 52% market share, an advantage the community often credits to its intuitive visual dashboard and multi-tenant access controls. Flux, recognized for its modular GitOps Toolkit, maintains an adoption rate of approximately 11% to 16%.
NOTE
Further reading: A Kubernetes Engineer’s Hands-On Guide to Argo CD
This governance model faced a real-world stress test in 2024. When Weaveworks announced its closure due to business difficulties, the original blog went dark. Yet because Flux had already been transferred to CNCF governance, the open-source community took the reins without interruption, continuing active maintenance and regular releases.
The Placebo Critique and Adoption Realities: When the Buzzword Outpaces Practice
Continuous delivery expert Steve Smith took aim at the trend in his essay “GitOps is a placebo”, bluntly calling GitOps an “unnecessary rebadging” that merely slapped a fresh marketing label onto established declarative IaC and automated delivery patterns.
The critique was partly on target: every individual building block highlighted in the 2017 foundational post had clear historical antecedents.
What the placebo argument overlooks, however, is the closed-loop feedback mechanism. Traditional IaC scripts execute once upon a pipeline trigger, powerless against configuration drift that occurs afterward. GitOps closes this loop via an in-cluster agent that continuously compares actual state against desired state and actively remediates drift, addressing a fundamental blind spot of one-way CI pipelines.
Yet in practice, a stark disconnect remains between the ubiquity of the term and the rigor of its execution.
In CNCF surveys, the share of organizations reporting that they “follow GitOps principles” jumped from 53% in 2022 to 77%. However, according to an Octopus Deploy survey of 660 respondents in 2025 that dug into technical specifics, only 16% of teams meet four of the six practices the study assessed.
While about 90% of teams have adopted version control together with human-readable configuration formats, only about 40% have both declarative configuration and automated pulling in place.
Furthermore, while respondents strongly agreed that GitOps improved auditability (81%) and prevented drift (78%), only 63% agreed that it enhanced security. This disparity highlights a growing awareness among engineering teams: shifting credentials away from CI does not eliminate security threats. When cluster control becomes tightly coupled to a Git repository, repository permissions, gaps in pull request review, and the storage of sensitive secrets instead become the new defensive bottlenecks.
The primary role of the four OpenGitOps principles is to establish unambiguous boundaries for an industry fraught with conflicting interpretations, freeing the discussion from any single vendor’s marketing vocabulary. With nearly 80% of teams claiming adoption while only 16% meet four of the six assessed practices, the standard provides an objective yardstick: it refocuses operations on reconciliation loops and perimeter defense, rather than settling for another superficial GitOps marketing wrapper.