The systemctl command in Linux
systemctl is the interface to systemd, the init system of virtually all current distributions. Services (nginx, postgres, your own apps) are "units"; systemctl starts and stops them now, enables them at boot, and reports their status — including the last log lines, which makes systemctl status the first stop when anything misbehaves.
How systemctl works
systemctl is a client: the work happens in systemd (PID 1), reached over D-Bus. Units — services, timers, mounts, sockets — are declarative files describing what to run and what it depends on; systemd computes the dependency graph and orders operations, which is how enabling one service transparently pulls in what it needs.
systemd's superpower is cgroups: every process a service spawns, including forks of forks, lives in the service's control group. That is why "systemctl stop" is reliable where old init scripts leaked orphans — the whole group is terminated — and why status can list every process belonging to a service and account for its memory and CPU precisely.
Syntax
systemctl COMMAND [UNIT]Common options
| Option | What it does |
|---|---|
status UNIT | State, PID, uptime, and recent log lines — the diagnostic starting point. |
start / stop / restart | Control the service now. |
reload | Re-read configuration without a full restart (if the unit supports it). |
enable / disable | Start (or not) automatically at boot. enable --now also starts it immediately. |
list-units --type=service | Everything currently loaded. |
daemon-reload | Re-read unit files after you edit them. |
How to use systemctl: examples
$ systemctl status nginxIs it running, since when, with what PID — and the last few log lines.
$ sudo systemctl restart myappThe deploy-day classic.
$ sudo systemctl enable --now postgresqlStart now AND at every boot, in one command.
$ systemctl list-units --type=service --state=failedAnything red on this machine?
$ sudo systemctl daemon-reloadAfter editing a unit file — systemd caches them.
Real-world use cases for systemctl
Shipping your own service
Any long-running script deserves a unit file: ten lines in /etc/systemd/system/myapp.service (ExecStart, Restart=on-failure, User=), then daemon-reload and enable --now. You gain supervised restarts, boot persistence and journald logging — retirement papers for nohup-in-a-screen deployments.
The morning-after audit
After patching night: systemctl list-units --state=failed on each host. Empty output is the all-clear; anything listed comes with systemctl status ready to show why. Two commands, complete confidence.
Pro tips and common mistakes
- status is diagnosis, not just state: PID, uptime, memory and the last log lines in one screen. Read it before journalctl.
- Edited a unit file? Nothing happens until daemon-reload — systemd caches units, and forgetting this is the classic "my change was ignored".
- enable --now exists so you stop forgetting one half of the enable/start pair.
Frequently asked questions about systemctl
restart or reload?
restart stops and starts (brief downtime, drops connections). reload asks the service to re-read config gracefully — prefer it when the service supports it (nginx does).
enable vs start?
start affects now; enable affects boot. A started-but-disabled service dies at reboot — the classic "it worked until the maintenance window" story.
Where do my service logs go?
Into the journal: journalctl -u myservice. systemd captures stdout/stderr of units automatically.
Things people do with systemctl
Find which process holds a port and stop it cleanly.
Find a process ID in LinuxLocate a process PID with ps, pgrep and pidof before inspecting or stopping it.
Stop a process by nameFind matching process IDs with pgrep and terminate them safely.
Errors you may hit with systemctl
Something is already listening on the port your program wants.
Connection refusedYou reached the machine, but nothing was listening on that port — or a firewall said no.
Related commands
See what is running, find a process ID, and combine with grep to hunt down a program.
killTerminate processes politely or forcibly — and understand what -9 really does.
journalctlQuery the system journal: by service, time window, priority — and follow live.