INKHOLD

Operator · How-to

Back up automatically#

Take a backup every night into a folder of plain files (a NAS share, a USB disk, a Dropbox or Google Drive folder) and check that it can be restored. Inkhold keeps serving while it backs up.

In Docker#

Mount the backup folder at /backup, and that is all:

docker run -d --name inkhold -p 8089:8089 \
  -v inkhold:/var/lib/inkhold \
  -v /mnt/nas/inkhold-backup:/backup \
  ghcr.io/m00qek/inkhold:latest

The container then backs up every night at 03:00 UTC, keeps 7 snapshots, and checks each new one. If the computer was off at 03:00, it backs up a minute after it next starts. docker logs inkhold shows each run.

Mount the NAS share itself. A bind mount of a folder whose share has dropped off quietly fills the computer’s own disk instead.

To change the time or the options, copy the image’s crontab, edit it, and mount it with -v ./crontab:/etc/inkhold/crontab:ro. For local time instead of UTC, pass a POSIX zone, for example -e TZ=CET-1CEST,M3.5.0,M10.5.0/3.

With the binary#

The folder must exist; then:

inkhold backup create /mnt/nas/inkhold --keep=7
inkhold backup list   /mnt/nas/inkhold
inkhold backup verify /mnt/nas/inkhold            # --deep re-reads every file

To run it every night, either:

  • add one line with crontab -e:

    0 3 * * *  /usr/local/bin/inkhold --config=/etc/inkhold/config.zon backup create /mnt/nas/inkhold --keep=7 >/dev/null

  • or use the systemd service and timer in contrib/systemd/: set the account, config path and folder at the top of inkhold-backup.service, copy both files to /etc/systemd/system/, then systemctl enable --now inkhold-backup.timer. It backs up at 03:00, makes up for a night the computer was off, and checks each new snapshot.

Worth knowing#

  • The folder is not encrypted. Keep it as private as the notebooks in it.
  • cred.key is left out unless you pass --include-cred-key. Without it a restore still brings back every notebook; each account gets a temporary password instead.
  • Admins are told when backups stop. A line in the library says when the last backup failed, is more than two days old, or never happened, and the admin console’s server panel lists every backup folder.
  • Backing up is slow on purpose: low priority, reading 4 MiB/s (--max-rate changes it), so syncing stays quick meanwhile.
  • On FAT32 a file cannot exceed 4 GB. On “online-only” cloud folders, verify --deep downloads everything to check it.

How backups work explains snapshots and why each night’s backup is small.