Using Cron to restart a systemd user service

I don't think systemd allows you to do that, at least the systemd.service(5) manual page doesn't seem to mention anything like that.

However, what you could do is use systemd's builtin watchdog.

You would do that by settings WatchdogSec= and then having your service send WATCHDOG=1 with sd_notify regularly. If you set WatchdogSec=30 then your service must notify systemd at most every 30 seconds.

When your service hangs, it won't notify systemd anymore, and systemd will kill your service as a result.

With restart=on-watchdog (which you already have, but it doesn't do anything without WatchdogSec=), then systemd will restart your service after it's been killed by the watchdog.

Using cron is the correct way to schedule periodic running of tasks on most Unix systems. Using a personal crontab is the most convenient way for a user to schedule their own tasks. System tasks may be scheduled by root (not using the script below!) in the system crontab, which usually has an ever so slightly different format (an extra field with a username).

Here's a simple script for you. Any user may use this to manage their own personal crontab.

• It doesn't do any type of validation of its input except that it will complain if you give it too few arguments. It is therefore completely possible to add improperly formatted crontab entries.

• The remove sub-command takes a line number and will remove what's on that line in the crontab, regardless of what that is. The number is passed, unsanitized, directly to sed.

• The crontab entry, when you add one, has to be quoted. This affects how you must handle quotes inside the crontab entry itself.

Most of those things should be relatively easy for you to fix.

#!/bin/sh

usage () {
    cat <<USAGE_END
Usage:
    $0 add "job-spec"
    $0 list
    $0 remove "job-spec-lineno"
USAGE_END
}

if [ -z "$1" ]; then
    usage >&2
    exit 1
fi

case "$1" in
    add)
        if [ -z "$2" ]; then
            usage >&2
            exit 1
        fi

        tmpfile=$(mktemp)

        crontab -l >"$tmpfile"
        printf '%s\n' "$2" >>"$tmpfile"
        crontab "$tmpfile" && rm -f "$tmpfile"
        ;;
    list)
        crontab -l | cat -n
        ;;
    remove)
        if [ -z "$2" ]; then
            usage >&2
            exit 1
        fi

        tmpfile=$(mktemp)

        crontab -l | sed -e "$2d" >"$tmpfile"
        crontab "$tmpfile" && rm -f "$tmpfile"
        ;;
    *)
        usage >&2
        exit 1
esac

Example of use:

$ ./script
Usage:
    ./script add "job-spec"
    ./script list
    ./script remove "job-spec-lineno"

$ ./script list
     1  */15 * * * * /bin/date >>"$HOME"/.fetchmail.log
     2  @hourly /usr/bin/newsyslog -r -f "$HOME/.newsyslog.conf"
     3  @reboot /usr/local/bin/fetchmail

$ ./script add "0 15 * * * echo 'hello world!'"

$ ./script list
     1  */15 * * * * /bin/date >>"$HOME"/.fetchmail.log
     2  @hourly /usr/bin/newsyslog -r -f "$HOME/.newsyslog.conf"
     3  @reboot /usr/local/bin/fetchmail
     4  0 15 * * * echo 'hello world!'

$ ./script remove 4

$ ./script list
     1  */15 * * * * /bin/date >>"$HOME"/.fetchmail.log
     2  @hourly /usr/bin/newsyslog -r -f "$HOME/.newsyslog.conf"
     3  @reboot /usr/local/bin/fetchmail