Lessons learnt from implementing GitOps and Continuous Delivery at scale
summary
Describes real deployments using Argo CD for ~100–120 industrial clusters and Flux for telco fleets scaling from ~1,200 toward 5,000–6,000, and the design tradeoffs found: hub‑and‑spoke fails on admin privileges and flaky networks, delegation reduces central load but needs plugins, and sharding plus sharded controllers eases scale. Recommends moving targeting rules into Git, using webhook-driven updates to stop polling, and a tiny downstream agent that pulls compressed bundles (avoids CRDs, preserves constrained devices).
AI-written summary · about
chapters (7)
- 0:00Introduction
- 2:22Scale challenges in GitOps
- 9:27Security and networking limits
- 13:38Targeting rules in Git
- 17:50Introducing Fleet tooling
- 21:23Fleet architecture overview
- 28:41Future plans for Fleet
related talks
05mLessons from a Large-Scale GitOps Implementation
37mPanel: Building a GitOps...
35mPanel: GitOps User Stories
31mLeveraging GitOps to wrangle your clusters and projects
29mTales from the Branches: GitOps in the Real World
43mGitOps at Scale with and Argo CD on
56mNavigating the GitOps ecosystem: The principles and in-depth look at Argo and Flux
39mSoftware Rollout at Scale: Using GitOps to Scale Kubernetes Rollouts