

Spilling the git tea (Ship It! #112)
Jul 12, 2024
Discussion includes decentralized Git design, challenges in self-hosting Git servers, backend of gitea.com, use of Sentry for error monitoring, quirky fasting stories, GitT platform exploration, managing large binary files, S3 implementation in CI/CD, frontend development with Retool, changing ownership models in products, enhancing Git workflow with Git Extras
AI Snips
Chapters
Transcript
Episode notes
Gitea's Infrastructure Evolution
- gitea.com started with a complex multi-node Kubernetes setup with Ceph storage.
- They later simplified to a single server setup due to operational complexity and resource constraints.
Balancing Simplicity and Scale
- Single-node servers simplify operations but sacrifice high availability.
- Using cloud services like RDS and S3 offloads storage and scaling complexity effectively.
Git Data Storage Needs
- Git repositories require low disk latency and cannot be stored on object storage like S3.
- Metadata is parsed and optimized via gRPC to improve performance across nodes.