Move Fast with Stable Build Infrastructure
0 likes31 views
The document discusses the importance of stable build infrastructure in software development, highlighting the challenges faced with on-premise systems, including long waiting times and significant maintenance overhead. It details the journey towards fully automated provisioning, resulting in reduced build times and increased efficiency through homogeneous build agents and effective workload management. The improvements led to faster feedback cycles and enhanced developer productivity.
Move Fast with Stable Build Infrastructure
- 1. Move Fast with Stable Build Infrastructure Sanchit Bahal Move Fast with Stable Build Infrastructure Sanchit Bahal
- 2. Sanchit Bahal • Lead Consultant @ ThoughtWorks • Polyglot full-stack developer • DevOps enthusiast • Current areas of interest: Distributed and Decentralized architectures @sanchit_bahal www.linkedin.com/in/sanchitbahal sbahal@thoughtworks.com
- 7. Tech Stack
- 9. ~100 Pipelines ~80 People On-premise infra Enormous maintenance overhead Significant loss of productivity in managing infrastructure Snowflakes Non-resilient Long waiting times for builds to get assigned an agent (as high as 1.5 days) Broken test pyramid (to add to the woes) Challenges
- 10. Build Agents Work Allocation
- 11. Developer time is precious!
- 12. The Journey
- 14. Pre-baked ‘Golden’ Image Provisioning time reduced from 2 days to 20 minutes No Snowflakes
- 15. Homogeneous Build Agents Build machines capable of running any build type Efficient load balancing of workload Easier allocation of work Resilience
- 16. Easier allocation of work Single build agent per machine
- 17. Android Emulators over Genymotion Spin-up and shut-down Android emulators with every test suite run Run in HEADless mode Save on licensing cost
- 18. Fully automated provisioning No snowflakes ‘Homogeneous’ build agents Resilience State post improvements
- 20. Tech Stack
- 21. Build Wait Time Time it takes to get a build agent assigned to a scheduled job
- 22. Deploy-ready BuildsDeploy Cycle Time Time it takes to get a deployable artifact out
- 24. Impact Machine provisioning time reduced from 2 days to < 20 minutes Build wait times reduced from 6 hours to < 10 seconds Stable, robust, resilient build infrastructure Faster feedback cycle => Increased code commits/day Improved developer/team productivity
- 25. Thank you! We don’t want humans waiting on computers. We want computers waiting on humans. Facebook’s F8 conference, 2015