KubeFM Our Journey to GitOps: Migrating to ArgoCD with Zero Downtime, with Andrew Jeffree
Oct 28, 2025
Andrew Jeffree, a Staff Cloud Infrastructure Engineer at SafetyCulture, shares insights from migrating over 250 microservices to GitOps with Argo CD, all while ensuring zero downtime. He discusses the switch from a complex Helm setup to a CUE-based domain-specific language, which enhances developer experience with better validation. Key topics include strategies for seamless migration, automated reconciliation, and the importance of empathy in engineering design. Jeffree also highlights the benefits of codeless approvals and adapting tools to minimize operational pain points.
AI Snips
Chapters
Transcript
Episode notes
Monorepo Helm Worked — Until It Didn’t
- SafetyCulture ran three production clusters (EU, US, AU) and used a monorepo Helm setup with hierarchical YAML configs and per-cluster pipelines.
- That model worked initially but became fragile as services scaled and pipelines proliferated.
YAML Drift Creates Hidden Breakages
- YAML and Helm templating caused hidden errors and cognitive load for developers unfamiliar with Kubernetes fields.
- Inheritance and drift meant changes in one file could silently break other environments.
Make Git The Single Source Of Truth
- Adopt GitOps to make Git the single source of truth and enable automatic reconciliation across clusters.
- Use the history in Git to trace who changed what and avoid ad-hoc pipeline patches.