Working with systemd Timers

Create an executable file /usr/local/bin/helloworld.sh with the following content:

#!/bin/sh
  # This is bash program to display Hello World
  echo " Hello World "

This is an executable .sh file containing the commands you want systemd to run and manage.

Create the file /etc/systemd/system/helloworld.service with the following content:

[Unit]
Description="Hello World script"

[Service]
ExecStart=/usr/local/bin/helloworld.sh

This is a systemd service file that tells systemd which application to run.

Create the file /etc/systemd/system/helloworld.timer with the following content:

[Unit]
Description="Run helloworld.service 5min after boot and every 24 hours relative to activation time"

[Timer]
OnBootSec=5min
OnUnitActiveSec=24h
OnCalendar=Mon..Fri *-*-* 10:00:00
Unit=helloworld.service

[Install]
WantedBy=multi-user.target

This is the timer file that controls the activation of the respective service file.

Verify that the files you created above contain no errors:

> systemd-analyze verify /etc/systemd/system/helloworld.*

If the command returns no output, the files have passed the verification successfully.

Start the timer:

> sudo systemctl start helloworld.timer

Activates the timer for the current session only.

Enable the timer to make sure that it is activated on boot:

> sudo systemctl enable helloworld.timer

Create the script /usr/local/bin/send_service_failure_email.sh.

#!/bin/sh

# Send an alert email when a systemd service fails (via failure_email@.service)
# Parameter: Service name (via %i from the service file)
#
# Email address(es) (comma separated, no spaces) must be configured in
# the service file as environment ALERT_EMAILS:
# Environment="ALERT_EMAILS=mail1@DOMAIN,mail2@DOMAIN"
#

UNIT="$1"

# Fetch the environment variables from the failing unit
ENV_VARS=$(systemctl show -p Environment --value "$UNIT")
# Extract the value of ALERT_EMAILS
EMAILS=\
$(printf '%s\n' "$ENV_VARS" | tr ' ' '\n' | grep '^ALERT_EMAILS=' | cut -d= -f2)


# Set fallback email address
if [ -z "$EMAILS" ]; then
    EMAILS="root@DOMAIN"
fi

/usr/sbin/sendmail -t <<EOF
From: $(hostname)@DOMAIN
To: $EMAILS
Subject: [ALERT] Systemd service failure for: $UNIT

The systemd service for $UNIT failed to execute on $(hostname) at $(date).

Check journalctl -u $UNIT for details.
EOF

Replace DOMAIN with the correct domain name. Also verify that hostname delivers the desired result. If not, replace it with a hard coded string. Finally, consider changing the fallback e-mail address.

Make the script executable:

> sudo chmod 755 /usr/local/bin/send_service_failure_email.sh

Create the service file /etc/systemd/system/failure_email@.service.

[Unit]
Description=Send email notification on service failure for %i

[Service]
Type=oneshot
ExecStart=/usr/local/bin/send_service_failure_email.sh %i

Verify the service file and fix the reported issues:

> systemd-analyze verify /etc/systemd/system/failure_email@.service

If the command returns no output, the file has passed the verification successfully.

Test the procedure. Start the service using the dbus instance for testing. (You can use any other service that is currently running. dbus is used in this example because the service is guaranteed to run on any SUSE installation.)

> sudo systemctl start failure_email@dbus.service

If successful, the fallback e-mail address from the send_service_failure_email.sh receives an e-mail with the subject [ALERT] Systemd service failure for: dbus. (This is just a test; there is no problem with the dbus service. You can safely delete the e-mail, and no action is required).

Additionally, inspect the journal:

> journalctl -u failure_email@dbus.service

If the test has passed successfully, proceed with the last step that makes the required changes to your service files.

To receive an e-mail failure notification from a service, add an OnFailure option to the service file and specify the e-mail recipients.

[Unit]
Description="Hello World script"
OnFailure1=failure-email@2%N.service

[Service]
Environment3="ALERT_EMAILS=mail1@DOMAIN,mail2@DOMAIN"
ExecStart=/usr/local/bin/helloworld.sh

Verify the service file and fix the reported issues:

> systemd-analyze verify /etc/systemd/system/helloworld.service

Create a service file executing the script. See Example 1, “The service file” for details.

Create a timer file executing the service file. See Example 2, “The timer file” for general instructions.

1. Convert calendar entries. Time is specified differently in cron and systemd. Use the patterns below as a conversion template:

   Cron:               Minute Hour Day Month DayOfWeek
   systemd: OnCalendar=DayOfWeek Year-Month-Day Hour:Minute:Second

   To test the converted calendar entry, follow the instructions in Section 5, “Testing calendar entries”.

2. Convert cron nicknames (@NICK):

   Cron     : systemd timer
   -------- : ----------------------------
   @reboot  : OnBootSec=1s
   @yearly  : OnCalendar=*-01-01 00:00:00
   @annually: OnCalendar=*-01-01 00:00:00
   @monthly : OnCalendar=*-*-01 00:00:00
   @weekly  : OnCalendar=Sun *-*-* 00:00:00
   @daily   : OnCalendar=*-*-* 00:00:00
   @hourly  : OnCalendar=*-*-* *:00:00

3. Convert variable assignments. The systemd variable assignment must go into the [Service] section. You cannot convert MAILTO this way—refer to the next step for this.

   cron: VARIABLE=VALUE
   systemd: Environment="VARIABLE=VALUE"

4. Set up e-mail notifications to replace cron's MAILTO feature by following the instructions in Section 6, “Getting e-mail notifications when a timer fails”.