Running GitLab and GitLab CI in Kubernetes for the Enterprise
summary
Northwestern Mutual moves GitLab onto Kubernetes on AWS to serve ~3,000 users and ~40,000 repos, discovering Omnibus containers are resource hogs and fixing that with Helm charts, EKS, resource requests/limits, and separated services. Deployments dedicate node pools with labels/taints/tolerations and prefer i3 NVMe local disks over EBS for Git data (much higher throughput). Avoid EFS because of EFS 8,900 IOPS limit. For CI, choose the jobs run as pods Kubernetes executor and use KIAM/project-to-IAM mapping to issue time‑bound credentials without per‑runner secrets.
AI-written summary · about
chapters (9)
- 0:00Introduction to GitLab in Kubernetes
- 3:40Early Kubernetes GitLab deployment
- 5:11Common pitfalls to avoid
- 8:45Helm charts for GitLab
- 12:25GitLab CI runners in Kubernetes
- 16:41Dedicated CI node pools
- 17:01Security for CI pipelines
- 20:13Kimesh for role-based access
- 23:04Q&A and final thoughts
related talks
35mGitLab CI and Kubernetes, Build Test and Deploy Your Projects
19mMigrating GitLab.com to Kubernetes: How Multiple Clusters Saved Cost
36mBuilding a CI/CD driven infrastructure for managing Kubernetes clusters
02h 47mA GitOps Architecture from Scratch: Gitlab, Ansible, Terraform, Kub and AWS
55mCI/CD Best Practices with Kubernetes and GitLab
22mEmbedded Software Development Efficiency Improvement with Docker and Kubernetes
18mScalable CI/CD Infrastructure with EKS and Karpenter
47mGitLab CI and Kubernetes: deploying Applications continuously to Kubernetes