[Case Study] Building an In-House Personal Development Environment Image, Part 1: Creating an OpenStack Sandbox Updated via git push

An organization developing OpenStack-based services conducted an experiment to standardize and provide an OpenStack sandbox that individual developers can use to experiment independently. To maintain a structure similar to the production environment while still enabling individual-level experimentation, it was implemented as a single cluster within a single VM.

The core design lies in separating the roles of the image and Git. The image (qcow2) contains the infrastructure skeleton and the initial state of OpenStack—Ubuntu 24.04, Kubernetes, ArgoCD, Flux, and so on—to speed up boot time. On the other hand, settings and component versions that require continuous updates are managed in the Git repository (stack-deployments-sandbox).

The specific mechanism works as follows:

• Image: Includes low-change-frequency areas such as the kernel, runtime, and control plane, and performs minimal initialization tasks like IP replacement at boot time.
• Git: Defines areas that continue to change during operation, such as HelmRelease, ArgoCD Application, and ConfigMap.
• Operation: When a user pushes changes to Git, ArgoCD and Flux, pre-installed in the image, detect this and automatically converge the cluster state to match the definitions in Git.

Through this structure, a foundation was established for efficiently propagating security patches or configuration changes to all sandbox environments without rebuilding the image.