Running a nightly backup script after logout with systemd user units – no nohup needed
I set up a little home‑lab server last year and wanted a way to snap my home directory every night without tying the job to a terminal. The old nohup trick works, but it leaves orphaned processes and you end up chasing logs in the dark. Systemd user units give a cleaner, more reliable solution that plays nicely with the rest of the system.
Why use a systemd user unit?
- Automatic cleanup – When you log out, systemd kills any processes started by the unit, so you never end up with stray backup jobs.
- Fine‑grained control – You can set
OnCalendar,ExecStart,Restart, andStandardOutputjust like a system service. - No extra daemons – The user instance of systemd is already running for every logged‑in user, so you don’t need to start a separate service manager.
- Security isolation – The unit runs in the user namespace, limiting what it can see and do compared to a root‑level service.
Quick setup
Assume your backup script sits at ~/scripts/backup.sh and produces its own logs.
chmod +x ~/scripts/backup.sh
1. Create the unit file
User units live in ~/.config/systemd/user/. Drop a file called backup-nightly.service there:
[Unit]
Description=Nightly backup of home directory
After=network-online.target
[Service]
Type=oneshot
ExecStart=/home/youruser/scripts/backup.sh
StandardOutput=append:/home/youruser/backup.log
StandardError=append:/home/youruser/backup.err
# If you want the job to retry on failure
Restart=on-failure
RestartSec=5s
[Install]
WantedBy=default.target
After=network-online.targetmakes sure the network is up before the backup starts.StandardOutput/StandardErrorsend output straight to log files;append:keeps old data.Restart=on-failureis optional; if the script exits non‑zero, systemd will retry after 5 s.
2. Create a timer unit
The timer triggers the service. In the same directory, create backup-nightly.timer:
[Unit]
Description=Run nightly backup at 02:30
[Timer]
OnCalendar=*-*-* 02:30:00
Persistent=true
[Install]
WantedBy=timers.target
Persistent=true guarantees the job runs if the machine was off at the scheduled time.
3. Reload, enable, and start
systemctl --user daemon-reload
systemctl --user enable --now backup-nightly.timer
Verify the timer:
systemctl --user list-timers | grep backup-nightly
The first run will happen at the next 02:30. After that, systemd takes care of the rest.
What the backup script should look like
A minimal example that uses rsync:
#!/usr/bin/env bash
set -euo pipefail
SRC="$HOME/"
DST="/mnt/backup/home-$(date +%Y%m%d)/"
mkdir -p "$DST"
rsync -aAXv --delete "$SRC" "$DST"
-aAXvpreserves permissions, ACLs, and extended attributes.--deletekeeps the destination clean.- The script exits with a non‑zero status if
rsyncfails, triggering theRestart=on-failurelogic.
If you prefer btrfs snapshots, swap the rsync line for a btrfs subvolume snapshot command. The unit file stays the same.
Security considerations
- Least‑privilege – The unit runs as your user, not root. If the backup needs to touch root‑only files, add a dedicated
sudorule that only allows the script to run as root. - File permissions – Keep the script and logs readable only by you:
chmod 700 backup.shandchmod 600 backup.log backup.err. - Network – If the backup writes to a remote server via SSH, use key‑based authentication and restrict the key’s permissions (
chmod 600 ~/.ssh/id_rsa). - Audit – Log the start and end times inside the script. A simple
echo "$(date) – backup finished" >> ~/backup.loghelps spot silent failures.
Common pitfalls and troubleshooting
| Symptom | Likely cause | Fix |
|---|---|---|
| Timer never fires | systemd --user not started |
Log in via a graphical session or run systemctl --user enable --now after login. |
| Service fails with “Permission denied” | Script or log file permissions | Ensure the script is executable and logs are writable. |
| Backup runs but no data copied | Wrong source path | Verify $SRC expands correctly; add echo "$SRC" to the script for debugging. |
| Logs grow unchecked | StandardOutput set to append: |
Rotate logs with logrotate or change to StandardOutput=journal to use systemd’s journal. |
Use journalctl --user -u backup-nightly.service to inspect recent runs. The -b flag shows only the current boot, which is handy for debugging after a reboot.
Trade‑offs compared to nohup
| Feature | nohup |
Systemd user unit |
|---|---|---|
| Process cleanup | Manual kill |
Automatic on logout |
| Logging | Manual redirection | Built‑in StandardOutput |
| Scheduling | cron or at |
systemd.timer |
| Security isolation | None | User namespace |
| Resource limits | None | LimitCPU, LimitFSIZE, etc. |
If you only need a one‑off backup, nohup is fine. For recurring jobs, systemd units are more maintainable and integrate with the rest of the system.
Happy backing up, and keep your logs tidy!