Apps
Yarn Cache
Yarn keeps a local store of downloaded package tarballs and metadata so repeated installs can reuse files instead of fetching every dependency from the registry again. That cache can accumulate outdated or corrupted archives, partial downloads, and package copies for old dependency versions, which leads to checksum mismatches, extract failures, or unexpectedly large cache folders. Kudu removes Yarn’s downloaded package cache contents from its standard cache directories without touching your projects, lockfiles, installed dependencies, accounts, or configuration.
Why clean Yarn Cache?
- A corrupted cached package tarball causes integrity check failures during install, which shows up as repeated "integrity check failed" or extraction errors until the bad archive is removed
- Partial downloads left behind after a network interruption make Yarn reuse incomplete package files, so installs can fail immediately with unexpected end-of-file or invalid tar archive messages
- Old cached copies of many package versions accumulate across projects and workspaces, and the visible symptom is a Yarn cache folder consuming hundreds of megabytes or several gigabytes of disk space
- Registry migrations or package republishing can leave stale metadata or archives in cache, causing Yarn to keep resolving to a bad local copy while fresh downloads succeed only after cleaning
- After long periods of dependency churn, Yarn may spend extra time scanning a bloated cache before installs, which users notice as slower dependency installation even on fast connections
- Broken cache entries can trigger repeated retries against the same package during install, leading to flaky CI or local setup runs that fail on one machine but succeed after the cache is cleared
What gets cleaned
Cache paths Kudu targets
Windows
%LocalAppData%/Yarn/Cache |
macOS
~/Library/Caches/Yarn |
Linux
~/.cache/yarn |
Frequently asked
Common questions about Yarn Cache
Related cleaners
Free & open source
Download Kudu and reclaim your disk space.
Available on Windows, macOS, and Linux. No account required, no feature gates, no telemetry without consent. All cleaning targets are open source and community-auditable.