Using systemd

Introduction

When computers boot up, obviously some software manages that process. On Linux and other Unix or Unix-like systems, this is usually handled via an init system. For example, macOS uses launchd and many Linux distributions, including Ubuntu, use systemd.

systemd does more than handle the startup process, it also manages various services and connects the Linux kernel to various applications. In this section, we'll cover how to use systemd to manage services, and to review log files.

Manage Services

When we install complicated software, like a web server (e.g., Apache2, Nginx), a SSH server (e.g., OpenSSH), or a database server (e.g., mariaDB or MySQL), then it's helpful to have commands that manage that service (the web service, the SSH service, the database service, etc).

For example, the ssh service is installed by default on our gcloud servers, and we can check its status with the following systemctl command:

systemctl status ssh

The output tells us a few things. The line beginning with Loaded tells us that the SSH service is configured. At the end of that line, it also tells us that it is enabled, which means that the service will automatically start when the system gets rebooted or starts up.

The line beginning with Active tells us that the service is active (running) and for how long. It has to say this since I'm connecting to the machine using ssh. If the service was not active (running), then I wouldn't be able to login remotely. We also can see the process ID (PID) for the service as well as how much memory it's using.

At the bottom of the output, we can see the recent log files. We can view more of those log files using the journalctl command. By default, running journalctl by itself will return all log files, but we can specify that we're interested in log files only for the ssh service. We can specify using the PID number. Replace NNN with the PID number attached to your ssh service:

journalctl _PID=NNN

Or we can specify by service, or more specifically, its unit name:

journalctl -u ssh

Use Cases

Later we'll install the Apache web server, and we will use systemctl to manage some aspects of this service.

In particular, we will use the following commands to:

  1. check the state of the Apache service,
  2. configure the Apache service to auto start on reboot,
  3. start the service,
  4. reload the service after editing its configuration files, and
  5. stop the service.

In order, these work out to:

systemctl status apache2
sudo systemctl enable apache2
sudo systemctl start apache2
sudo systemctl reload apache2
sudo systemctl stop apache2

systemctl is a big piece of software, and there are other arguments the command will take. See man systemct for details.

Examine Logs

As mentioned, the journalctl command is part of the systemd software suite, and it is used to monitor system logs.

It's really important to monitor system logs. They help identify any problems in the system or with various services. For example, by monitoring the log entries for ssh, I can see all the attempts to break into the server. Or if the Apache2 web server malfunctions for some reason, which might be because of a configuration error, the logs will indicated how to identify the problem.

If we type journalctl at the command prompt, we are be presented with the logs for the entire system. These logs can be paged through by pressing the space bar, the page up/page down keys, or the up/down arrow keys, and they can also be searched by pressing the forward slash / and then entering a search keyword. To exit out of the pager, press q to quit.

journalctl

It's much more useful to specify the field and to declare an option when using journalctl. See the following man pages for details:

man systemd.journal-fields
man journalctl

There are many fields and options we can use, but as an example, we see that there is an option to view the more recent entries first (which is not the default):

journalctl -r

Or we view log entries in reverse order, for users on the system, and since the last boot with the following options:

journalctl -r --user -b 0

Or for the system:

journalctl -r --system -b 0

I can more specifically look at the logs files for a service by using the -u option with journalctl:

journalctl -u apache2

I can follow the logs in real-time (press ctrl-c to quit the real-time view):

journalctl -f

Timers (Automation)

Linux and Unix operating systems have long provided a way to automate processes. In the past, and still available on most systems, is the cron service. systemd, however, also provides a way to automate jobs using timers.

In our bash exercises, we created a script to examine the /var/log/auth.log file for invalid IP addresses. Whenever we want to check to see what IP addresses are trying to login into our system, we have to run that command.

What if we could have that script run at specific times? For example, what if we wanted to run that script every morning at 8AM and then log the output to a file for us to read? We can do that with systemd timers.

First, let's modify our auth.sh script. In the example below, I've adjusted the location of the auth.log file, created two additional variables to record the start and end dates of the auth.log file, and then modified the end echo statement to add some additional information and save the output in a file called brute.log in our /srv/developers directory. The ${end_date} and ${start_date} variables were created after closely studying the /var/log/auth.log file.

#!/usr/bin/env bash

log_file="/var/log/auth.log"

end_date=$(grep -Eo "^[[:alpha:]]{3}[[:space:]]{1,2}[[:digit:]]{1,2}" "${log_file}" | tail -n1)
start_date=$(grep -Eo "^[[:alpha:]]{3}[[:space:]]{1,2}[[:digit:]]{1,2}" "${log_file}" | head -n1)

total_invalid=$(grep -ic "Invalid user" ${log_file})
invalid_IPs=$(grep "Invalid user" ${log_file} | grep -Eoc "[[:digit:]]{1,3}\.[[:digit:]]{1,3}\.[[:digit:]]{1,3}\.[[:digit:]]{1,3}")

echo "
Log entry created on $(date +%c).
From ${start_date} to ${end_date}, there were ${total_invalid} attempts to login to the system.
These came from ${invalid_IPs} unique IPs.
" >> /srv/developers/brute.log

Next, we need to create two additional files. First we create a service file. This file defines the service that we want to execute. Navigate to the service directory:

cd /etc/systemd/system

And use sudo nano to create a file called brute.service:

sudo nano brute.service

In the above file, we add the following information under two sections, a Unit section and a Service section. The Unit section includes a description of the service and a list of the service's requirements. The Service section declares the type of service, the location of the script to run, and the user to run the script under. Feel free to use this but be sure to change your User information:

[Unit]
Description="Summarize brute login attempts."
Requires=brute.timer

[Service]
Type=simple
ExecStart=/usr/local/bin/auth.sh
User=seanburns

See man 5 systemd.service for more details.

Next we need to create the timer file. Using sudo nano, run the following command in the same directory as above:

sudo nano brute.timer

In this file, add the following:

[Unit]
Description="Timer for the brute login service."

[Timer]
OnCalendar=*-*-* 08:00:00
Persistent=true

[Install]
WantedBy=timers.target

See man 5 systemd.timer for more details.

Next we need to enable and start the timer. To do that, we run two separate systemctl commands:

sudo systemctl daemon-reload

And then enable the timer:

sudo systemctl enable brute.timer

Start the timer:

sudo systemctl start brute.timer

And finally, check the status of all timers:

sudo systemctl list-timers

Or check the status of our specific timer:

sudo systemctl status brute.timer

You can now check that your script was automatically after the next time your system's clock reaches 8AM.

Useful Systemd Commands

You can see more of what systemctl or journalctl can do by reading through their documentation:

man systemctl
man journalctl

You can check if a service if enabled:

systemctl is-enabled apache2

You can reboot, poweroff, or suspend a system (suspending a system mostly makes sense for laptops and not servers):

systemctl reboot
systemctl poweroff
systemctl suspend

To show configuration file changes to the system:

systemd-delta

To list real-time control group process, resource usage, and memory usage:

systemd-cgtop
  • to search failed processes/services:
systemctl --state failed
  • to list services
systemctl list-unit-files -t service
  • to examine boot time:
systemd-analyze

Conclusion

This is a basic introduction to systemd, which is composed of a suite of software to help manage booting a system, managing services, and monitoring logs.

We'll put what we've learned into practice when we set up our LAMP servers.