Running a service under systemd

Goal

Start a long-running process at boot, restart it on failure, confine it, and collect its output — without custom init scripts.

Prerequisites

A Linux host with systemd and a service that runs in the foreground and logs to standard output.

Steps

1. Write /etc/systemd/system/<name>.service with [Unit] (Description, After=network-online.target), [Service] and [Install] (WantedBy=multi-user.target).
2. In [Service]: ExecStart= with an absolute path, User= a dedicated account, WorkingDirectory=, EnvironmentFile= for secrets (mode 0600), Restart=on-failure and RestartSec=.
3. Add confinement: NoNewPrivileges=yes, ProtectSystem=strict, ProtectHome=yes, PrivateTmp=yes, ReadWritePaths= for the few directories that need writes.
4. Bound resources: MemoryMax=, TasksMax=, and a TimeoutStopSec= that matches graceful shutdown.
5. systemctl daemon-reload, enable --now, then verify with systemctl status and journalctl -u <name>.

Expected result

The service survives reboots and crashes, cannot write outside its designated paths, and its logs are searchable with the rest of the system's.

Limits and test basis

Confinement directives can break services that write to unexpected locations; add them one at a time and test. Timers (.timer units) replace cron for scheduled jobs with the same confinement. Directives follow the cited manual page.