{
  "id": 11163264,
  "title": "Presentation: Keeping the Mainline Green Across Diverse Language Monorepos",
  "url": "https://urgent.news/2026/10/01/presentation-keeping-the-mainline-green-across-diverse-language",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-10-01T09:33:00.000Z",
  "source": {
    "name": "InfoQ",
    "slug": "infoq",
    "url": "https://www.infoq.com/presentations/mergequeue/"
  },
  "original_language": "en",
  "account": "Dhruva Juloori, a Senior Software Engineer at Uber, presented insights on maintaining green mainlines across diverse language monorepos during a conference in San Francisco. At Uber, over 4,500 engineers working in 10 global development centers concurrently commit to six monorepos, handling more than 65,000 changes monthly. The primary repository, Go, processes the highest volume of changes.\n\nMaintaining a green mainline is challenging due to potential conflicts arising from concurrent commits by multiple developers. For instance, Alice and Bob each create branches from the same mainline to introduce changes C1 and C2, respectively. Testing their changes individually in CI results in passing build checks. However, when Bob attempts to merge his change, the mainline fails, indicating instability.\n\nAt scale, with hundreds of developers committing to a single codebase, maintaining mainline stability becomes increasingly difficult. Uber's monorepos manage 65,000+ changes monthly, host 1,000+ business-critical microservices, and facilitate 100,000+ deployments per month, covering over 135 million lines of code.\n\nTo tackle these challenges, Uber developed SubmitQueue, a merge queue that provides developers with a single queue illusion to submit changes efficiently. SubmitQueue guarantees reasonable Service Level Objectives (SLOs) for developers to land changes quickly while optimizing CI resource usage and ensuring green mainlines at scale.",
  "summary": "Dhruva Juloori discusses how Uber maintains green mainlines across massive monorepos handling 65,000+ monthly changes. He explains how SubmitQueue uses binary speculation trees, conflict analysis, and machine learning models to predict build success and execution times. Dhruva shares how bypassing large diffs slashed CI resource usage by 53% while accelerating PR landing times by 37%. By Dhruva…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}