Infrastructure & Operations › Linux & Servers

systemd

The Linux service manager for starting, stopping and logging services.

Also known as: systemctl, systemd service, unit file

systemd is the init system and service manager on most Linux distributions. It starts as the first process (PID 1), brings up the machine, and manages long-running services. systemctl is how you talk to it.

systemctl start nginx
systemctl stop nginx
systemctl restart nginx
systemctl enable nginx    # start automatically at boot
systemctl disable nginx
systemctl status nginx
systemctl list-units --type=service

The unit of work is a unit, described by a file ending in .service, .socket, .timer, .mount or .target. A simple service:

# /etc/systemd/system/myapp.service
[Unit]
Description=My app
After=network.target

[Service]
ExecStart=/opt/myapp/bin/server
Restart=on-failure
User=app

[Install]
WantedBy=multi-user.target

After editing a unit, run systemctl daemon-reload: systemd caches the files. To customize a distro-provided unit, don’t edit it — add a drop-in override with systemctl edit nginx. And know the difference between start (now) and enable (at boot): a service can be running but not enabled, or enabled but not running.

The classic mistake is starting a background process with nohup and forgetting it. For anything that should survive crashes and reboots, write a unit and let systemd restart it. systemd also logs to the journal (see reading system logs) and provides .timer units, which can replace cron for scheduled jobs while adding logging and ordering.

Units depend on one another through targets, such as multi-user.target for a normal boot, and After= and WantedBy= order startup (see boot process).

systemd is not universal — other init systems exist — and details differ between distributions, so check your platform before assuming. Its trade-off is weight: it also owns logging, timers, sockets and device management, so there’s more to learn than a minimal init system, and a bad unit can wedge startup. For most servers that’s a good deal, because restarting, logging and scheduling come built in.