so i run a few servers for various things. most of them are debian, since they're shared with other people, which means that they mostly autoupdate – but they don't automatically restart updated services.

to be clear, this is the correct default for a server distro. the best-written software is crash-only, but the world runs on more than just well-written code. you can't just blithely reboot whenever there's a package update, you need to let the sysadmin decide how that happens.

but, like, what if it's a simple nginx webserver, and it can safely be shut down whenever?

in my case, i want my servers to reboot once a day or once a week, depending on the system. for my purposes, i wanted to use systemd timers, because cron is old and gross, and systemd is… still old and gross, actually. i don't know where i was going with that. i wanted to use systemd timers because i felt like it.

this turns out to be quite simple! make one file named /etc/systemd/system/reboot.service:

[Unit]
Description=Reboot the machine

[Service]
ExecStart=/usr/bin/systemctl reboot
Type=oneshot

and another named /etc/systemd/system/reboot.timer:

[Unit]
Description=Weekly reboot Mondays at 9:00 UTC

[Timer]
OnCalendar=Monday *-*-* 09:00:00

[Install]
WantedBy=timers.target

there are a couple of things to twiddle for your installation: the title of the timer, the first line in the second file, and the OnCalendar line.

the format is a little weird, described in detail here. the idea is that every single segment has to match the current time, for the thing to fire. this is a pretty simple timespec: it matches Mondays, any year/month/day, and exactly 9:00:00 utc.

once you're done fiddling with the file, you can turn the timer on with a simple systemctl enable --now reboot.timer. if you make updates, don't forget to systemctl daemon-reload!