See what's using disk space on a server over SSH

When du isn't enough

Every server fills up eventually: a runaway log, a forgotten Docker volume, a backup that never got rotated. The usual fix is an evening of du -sh * | sort -h, one folder at a time.

SquirrelDisk does the same job visually. Pick Remote server on the home screen, choose a host, and the whole file system appears as a sunburst chart that builds up while the scan runs. The biggest folders stand out in seconds, and you can click into them just like on your own disk.

How it works

1. You pick a host. SquirrelDisk reads the hosts in your ~/.ssh/config, or you can type user@host.
2. It connects with your own ssh. Your keys, agent and config apply. One connection is reused for the whole session, so you authenticate once.
3. It scans with the best tool available. SquirrelDisk first tries to use a small helper program in ~/.cache/squirreldisk on the server (it downloads the right build for the server from GitHub, or copies itself when the server runs the same system as your computer). If that isn't possible, it uses GNU find, and as a last resort du.
4. You explore the result on your computer. Zoom into folders, hover to see what's inside, switch between sunburst and treemap. The scan stays on the same file system, so network mounts under it aren't counted twice.

Cleaning up a server

Drag the folders you don't need into the collector at the bottom, then choose:

• Delete permanently: removes them on the server.
• Back up, then delete: copies them to a folder on your computer first (over scp), then removes them from the server.

Before anything is removed, SquirrelDisk checks each path again against its safety rules: system folders like /usr, /etc or /var, and the server's home folder itself, can't be deleted. App and settings folders ask for an extra confirmation.

Remote deletion has no undo. If in doubt, back up first.

Good to know

• Permissions: you see what your SSH user can read. To scan the whole system, connect as a user that can read it.
• Big file systems: millions of files are fine. The helper scans on the server itself and sends back the finished tree in one go, which keeps slow connections usable.
• Cloud storage: S3 buckets, Google Drive, Dropbox and other services are scanned the same way through rclone. See cloud storage.