--- title: "Give shared cache writes deliberate ownership" description: "Control which runs save shared cache entries so parallel jobs and pull requests do not compete to publish the same cache." url: https://starsling.dev/ci-playbooks/techniques/cache-write-ownership canonicalUrl: https://starsling.dev/ci-playbooks/techniques/cache-write-ownership --- # Give shared cache writes deliberate ownership Control which runs save shared cache entries so parallel jobs and pull requests do not compete to publish the same cache. ## How to choose an approach Compare the examples by repository setup, prerequisites, and risks before adopting one. Each company implemented this in its own environment; measure each change against your own baseline. - **[Anthropic](https://starsling.dev/ci-playbooks/anthropic#cache-write-hygiene)**: Use it when Pull request or merge-queue runs write cache entries that no later run can restore, while the repository is near its cache quota. Check before adopting: Every run still restores the cache and builds cold when it misses. - **[OpenAI](https://starsling.dev/ci-playbooks/openai#isolate-cache-writers)**: Use it when Parallel jobs report cache reservation collisions or restore an incomplete first writer's archive. Check before adopting: Separate namespaces add storage and cold misses. A shared producer can be better when the contents are identical; measure both. ## Apply this technique ````text Evaluate and apply this CI technique to my repository: Give shared cache writes deliberate ownership. Control which runs save shared cache entries so parallel jobs and pull requests do not compete to publish the same cache. Read the technique and its source-backed company examples: https://starsling.dev/ci-playbooks/techniques/cache-write-ownership.md Inspect this repository’s CI workflows, build and test commands, dependencies, and recent run timings. Identify where this technique applies and explain the expected benefit using this repository’s evidence. Measure a baseline from recent comparable runs: elapsed time, queue time, runner minutes and failures where available. If data is unavailable, say what cannot be measured. Compare relevant company approaches, checking prerequisites, platform differences and tradeoffs. Rank applicable options by evidence, fit and measurable opportunity; do not present company results as predictions for this repo. Recommend the smallest safe change, show what checks and fallback paths remain, and ASK ME WHICH CHANGE TO APPROVE BEFORE EDITING ANY FILES. Only after approval, work on a new feature branch, make one reversible change, and open a reviewable pull request. Never merge or push to a default or protected branch. Keep required checks, test coverage, deployment and release behavior intact. Validate the change and compare equivalent runs where possible; report limitations and what should be monitored after merge. ```` ## How companies used this technique Read what each company changed and follow the original sources. ### Anthropic: Save shared caches only from the default branch Buffa's tools-image build cache now saves only on pushes to main. Pull request and merge-queue runs restore from it, so their writes stop evicting the main entries every pull request starts from. Evidence: Source PR/commit evidence Use it when: Pull request or merge-queue runs write cache entries that no later run can restore, while the repository is near its cache quota. Before applying it: Every run still restores the cache and builds cold when it misses. [Read more in the Anthropic playbook](https://starsling.dev/ci-playbooks/anthropic#cache-write-hygiene) [Primary source](https://github.com/anthropics/buffa/pull/288) ### OpenAI: Give concurrent cache writers deliberate ownership Codex test, clippy and release-verification jobs competed to reserve the same immutable Bazel repository-cache key. Each job now gets a logical namespace with a matching fallback, shares one computed key between restore and save, and skips the save on an exact hit. Evidence: Source PR/commit evidence Each job saves under its own namespace and skips saving on an exact hit ([Source](https://github.com/openai/codex/pull/18366)) Use it when: Parallel jobs report cache reservation collisions or restore an incomplete first writer's archive. Before applying it: Separate namespaces add storage and cold misses. A shared producer can be better when the contents are identical; measure both. [Read more in the OpenAI playbook](https://starsling.dev/ci-playbooks/openai#isolate-cache-writers) [Primary source](https://github.com/openai/codex/pull/18366) ## CI optimization is continuous work. Apply this technique yourself, or use StarSling Runners for faster GitHub Actions and optional agent-driven optimization PRs, which teams on eligible paid plans can turn on. [Get self-improving CI](https://starsling.dev/products/runners) ## Company Playbooks ### [Anthropic](https://starsling.dev/ci-playbooks/anthropic) Run only affected tests ### [DoorDash](https://starsling.dev/ci-playbooks/doordash) Run build and test in parallel, sharded across CI agents ### [Linear](https://starsling.dev/ci-playbooks/linear) Benchmark the same job on faster hardware ### [Mastra](https://starsling.dev/ci-playbooks/mastra) StarSling runners for the slowest test workflow ### [OpenAI](https://starsling.dev/ci-playbooks/openai) Rank CI fixes by the minutes they save ### [Ramp](https://starsling.dev/ci-playbooks/ramp) Move a CPU-bound test suite to a faster CPU ### [Shopify](https://starsling.dev/ci-playbooks/shopify) Let an agent iterate against one CI metric ### [Spotify](https://starsling.dev/ci-playbooks/spotify) Build arm64 on native arm64 runners ### [Stripe](https://starsling.dev/ci-playbooks/stripe) Run only the tests a change can reach ### [Uber](https://starsling.dev/ci-playbooks/uber) Build the merge-queue speculations most likely to be needed ## Related guides Implementation guides for this technique. ### [Speed up fork pull request builds in GitHub Actions](https://starsling.dev/github-actions/optimizations/speed-up-fork-pr-builds) A fork pull request reads your base-branch cache but writes only its own scope, so each new PR starts cold. Publish warm artifacts from a trusted producer job.