Docker Desktop
Behind Docker Desktop’s UI, the Electron shell builds browser-style caches such as HTTP asset cache entries, V8 Code Cache bytecode, and GPUCache shader or rendering data, while the host service and embedded Linux VM also accumulate rotated diagnostic logs. These files speed up dashboard loading, extension rendering, and hardware-accelerated UI drawing, but they can go stale after Docker Desktop, Electron, or graphics driver updates. Kudu removes those disposable Electron caches and older settled host and VM log files while leaving images, containers, volumes, settings, credentials, and recent active logs alone.
Why clean Docker Desktop?
- Invalidated V8 Code Cache after a Docker Desktop update can make the dashboard open slowly or feel stuck on first launch until bytecode is rebuilt
- Stale GPUCache entries after a graphics driver change can cause blank panes, flickering settings screens, or a black Extensions view until the rendering cache is recreated
- Corrupted HTTP and asset cache files can leave the UI showing outdated status badges, missing icons, or panels that never finish loading even though the backend is running
- Accumulated host and VM logs grow quietly during image pulls, networking errors, and Kubernetes startup retries, eventually consuming gigabytes and triggering low-disk-space warnings on the system drive
- Large settled log sets make troubleshooting harder because current startup or engine errors are buried under old noise, so cleaning them leaves recent failures easier to inspect
- When the Electron cache exceeds a healthy size, opening Docker Desktop can cause extra disk churn and short freezes while the app scans or replaces stale cached resources
Cache paths Kudu targets
Windows
%AppData%/Docker Desktop/cache |
%AppData%/Docker Desktop/Code Cache |
%AppData%/Docker Desktop/GPUCache |
%LocalAppData%/Docker/log/host |
%LocalAppData%/Docker/log/vm |
macOS
~/Library/Caches/com.docker.docker |
Common questions about Docker Desktop
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.