51 KiB
Systemd
What is systemd?
Systemd is a collection of programs that aim to unify the service configuration and behavior on most modern Linux distributions.
All of the distributions we've used up until now come with systemd and we've been manipulating most of our servers and services via systemctl
which is the standard command line interface to systemd.
It's worth pointing out that systemd is not just an additional piece of software that is added to your computer.
You should see it as a sort of glue that ties the system together as it's responsible for launching and monitoring all services you run on your server.
Some history
As with most things Linux there are multiple alternatives to systemd and believe it or not, the introduction (around 2015) of systemd to Debian was a controversial moment. A lot of online debates were had to discuss the pro's and cons and Debian was even forked to remove systemd all together.
Devuan GNU+Linux is a fork of Debian without systemd that allows users to reclaim control over their system by avoiding unnecessary entanglements and ensuring Init Freedom.
You can be for or against systemd but the current reality is that it is the most widely used init
system around today.
This can, and probably will, change in the future but for now the world is run by systemd.
The basics
During the numerous hours you've spent using htop
you have probably noticed the first process is often /lib/systemd/systemd --system
on Debian machines.
On Raspberry Pi's that first process is most likely /sbin/init
but a closer look at this program shows the following.
pi@camone:~ $ pgrep -a systemd
1 /sbin/init
122 /lib/systemd/systemd-journald
150 /lib/systemd/systemd-udevd
295 /lib/systemd/systemd-timesyncd
378 /lib/systemd/systemd-logind
16414 /lib/systemd/systemd --user
pi@camone:~ $ which /sbin/init
/sbin/init
pi@camone:~ $ ls -l /sbin/init
lrwxrwxrwx 1 root root 20 Apr 1 14:57 /sbin/init -> /lib/systemd/systemd
pi@camone:~ $
Every running Linux computer must have a first process. But where does this first process come from? Below you can see a nice graph of the boot sequence of a standard Linux machine (taken from the Debian system administrator handbook).
By default the Linux kernel will run the init
program but this can be overridden by passing an argument to the kernel upon boot.
For those who have played around with the broken machines this is probably no real news.
At the last stage of the boot sequence, systemd takes over and launches all services that are enabled
for the requested runlevel
.
The runlevel might be new to you but we'll come back to that in a minute.
Interfacing with systemd
Your main tool to talk to systemd is systemctl
.
It's sort of a client to the systemd server.
The most used commands, that you probably know by hearth, are:
sudo systemctl start sshd.service
sudo systemctl stop sshd.service
sudo systemctl restart sshd.service
sudo systemctl status sshd.service
sudo systemctl enable sshd.service
sudo systemctl disable sshd.service
Just knowing these will get you a long way but there are a few more handy commands to push it all a bit further.
Beyond the basics
A deeper look into what's available
If you invoke just sudo systemctl
it lists all the units that are active.
It's actually a shortcut to sudo systemctl list-units
.
You'll be confronted with an interface, less
, that you know pretty well so have a look around and maybe search for some keywords.
At the bottom of the pager you'll see a few hints that point you to other commands that show even more output.
When we disable a server such as sshd
it's configuration files are not changed at all as the server never tries to start itself.
Systemd is responsible for that so if we want to see all servers available on our system we type sudo systemctl list-unit-files
which gives a clear table, also via less
, that outlines the current state and vendor state.
We can add more command line arguments to systemctl
to narrow down the output a bit.
A handy one is --type service
to only see services.
I advise you to have a read of the man systemctl
to grasp the full scope of it's capabilities.
Inspecting a running service
To inspect a running service we can run sudo systemctl status sshd.service
.
This gives us the following output:
● ssh.service - OpenBSD Secure Shell server
Loaded: loaded (/lib/systemd/system/ssh.service; enabled; vendor preset: enabled)
Active: active (running) since Mon 2021-07-26 12:14:35 CEST; 2 weeks 6 days ago
Docs: man:sshd(8)
man:sshd_config(5)
Main PID: 576 (sshd)
Tasks: 1 (limit: 23851)
Memory: 2.8M
CPU: 112ms
CGroup: /system.slice/ssh.service
└─576 sshd: /usr/sbin/sshd -D [listener] 0 of 10-100 startups
Jul 26 12:14:35 deathstar systemd[1]: Starting OpenBSD Secure Shell server...
Jul 26 12:14:35 deathstar sshd[576]: Server listening on 0.0.0.0 port 22.
Jul 26 12:14:35 deathstar sshd[576]: Server listening on :: port 22.
Jul 26 12:14:35 deathstar systemd[1]: Started OpenBSD Secure Shell server.
Jul 28 20:13:38 deathstar sshd[175321]: Connection closed by authenticating user waldek 192.168.0.222 port 51542 [preauth]
Aug 14 09:05:36 deathstar sshd[1001518]: Connection closed by authenticating user waldek 192.168.0.33 port 35448 [preauth]
Aug 14 09:05:56 deathstar sshd[1001567]: Connection closed by authenticating user waldek 192.168.0.33 port 35636 [preauth]
Aug 14 09:06:20 deathstar sshd[1001648]: Connection closed by authenticating user waldek 192.168.0.236 port 53346 [preauth]
There is quite a bit of interesting information here.
There are two blocks of information.
At the top we see some details and links to the help about the service in question and at the bottom we see the last eight lines of the server logs.
To see how systemd has the sshd service configured we need to have a look at the second line, the one that sais Loaded:
.
The path that follows is the service file that systemd uses to know how, when and where to run the service.
As with most things Linux this is a simple text file we can open up with less
, vim
or even nano
but there is a sweet shortcut supplied by systemd itself!
➜ ~ git:(master) ✗ sudo systemctl cat sshd.service
# /lib/systemd/system/ssh.service
[Unit]
Description=OpenBSD Secure Shell server
Documentation=man:sshd(8) man:sshd_config(5)
After=network.target auditd.service
ConditionPathExists=!/etc/ssh/sshd_not_to_be_run
[Service]
EnvironmentFile=-/etc/default/ssh
ExecStartPre=/usr/sbin/sshd -t
ExecStart=/usr/sbin/sshd -D $SSHD_OPTS
ExecReload=/usr/sbin/sshd -t
ExecReload=/bin/kill -HUP $MAINPID
KillMode=process
Restart=on-failure
RestartPreventExitStatus=255
Type=notify
RuntimeDirectory=sshd
RuntimeDirectoryMode=0755
[Install]
WantedBy=multi-user.target
Alias=sshd.service
➜ ~ git:(master) ✗
Modifying a service
What can we do with these unit files you might ask?
Well, we can have a look at the command line arguments for sshd with man sshd
.
This gives us an overview of all options available to us.
One that peaks my interest is the -p
argument which allows us to override the port and ignore all ports specified in the configuration file.
Let's try it out!
To edit the unit file we need a text editor.
There are two ways to do it but we'll go for the most straightforward one first.
I'll be using vim to edit the file via sudo vim /lib/systemd/system/ssh.service
.
Notice the syntax highlighting, nice no?
I modified the tenth line so that it reads:
ExecStart=/usr/sbin/sshd -D $SSHD_OPTS -p 2222
Now, how do we take this change into account? Let's restart the service.
~ git:(master) ✗ sudo systemctl restart sshd.service
Warning: The unit file, source configuration file or drop-ins of sshd.service changed on disk. Run 'systemctl daemon-reload' to reload units.
➜ ~ git:(master) ✗
We can see a warning but did the service restart? Let's have a look at the status.
● ssh.service - OpenBSD Secure Shell server
Loaded: loaded (/lib/systemd/system/ssh.service; enabled; vendor preset: enabled)
Active: active (running) since Sun 2021-08-15 20:00:49 CEST; 38s ago
Docs: man:sshd(8)
man:sshd_config(5)
Process: 1108166 ExecStartPre=/usr/sbin/sshd -t (code=exited, status=0/SUCCESS)
Main PID: 1108167 (sshd)
Tasks: 1 (limit: 23851)
Memory: 1.1M
CPU: 15ms
CGroup: /system.slice/ssh.service
└─1108167 sshd: /usr/sbin/sshd -D [listener] 0 of 10-100 startups
Aug 15 20:00:49 deathstar systemd[1]: Starting OpenBSD Secure Shell server...
Aug 15 20:00:49 deathstar sshd[1108167]: Server listening on 0.0.0.0 port 22.
Aug 15 20:00:49 deathstar sshd[1108167]: Server listening on :: port 22.
Aug 15 20:00:49 deathstar systemd[1]: Started OpenBSD Secure Shell server.
Yes it did, but the service is still running on port 22.
This is what systemd means by loaded
.
A configuration file is loaded into memory and used from there.
To take changes to unit files into account we need to reload the files that have changed, sort of like we restart sshd
when we make changes to it's configuration file but we can't restart systemd
as that would freeze our computer.
Luckily there is a command to do this and it's written in the warning notice.
➜ ~ git:(master) ✗ sudo systemctl daemon-reload
➜ ~ git:(master) ✗ sudo systemctl restart sshd.service
➜ ~ git:(master) ✗ sudo systemctl status sshd.service
● ssh.service - OpenBSD Secure Shell server
Loaded: loaded (/lib/systemd/system/ssh.service; enabled; vendor preset: enabled)
Active: active (running) since Sun 2021-08-15 20:05:53 CEST; 2s ago
Docs: man:sshd(8)
man:sshd_config(5)
Process: 1108694 ExecStartPre=/usr/sbin/sshd -t (code=exited, status=0/SUCCESS)
Main PID: 1108695 (sshd)
Tasks: 1 (limit: 23851)
Memory: 1.1M
CPU: 15ms
CGroup: /system.slice/ssh.service
└─1108695 sshd: /usr/sbin/sshd -D -p 2222 [listener] 0 of 10-100 startups
Aug 15 20:05:53 deathstar systemd[1]: Starting OpenBSD Secure Shell server...
Aug 15 20:05:53 deathstar sshd[1108695]: Server listening on 0.0.0.0 port 2222.
Aug 15 20:05:53 deathstar sshd[1108695]: Server listening on :: port 2222.
Aug 15 20:05:53 deathstar systemd[1]: Started OpenBSD Secure Shell server.
➜ ~ git:(master) ✗
Nice!
Now, let's first undo our changes and explore the alternative way to modify unit files.
Next we do the same changes but in the alternative way.
Just as we have a handy shortcut to cat
unit files we have one to edit
them!
I'll run the sudo -E systemctl edit --full sshd.service
command, notice the -E
, why would I do that?
This opens up my editor of choice and I can go ahead an make my changes to line 10 which I add -p 2200
to this time.
➜ ~ git:(master) ✗ sudo -E systemctl edit sshd.service --full
➜ ~ git:(master) ✗ sudo systemctl daemon-reload
➜ ~ git:(master) ✗ sudo systemctl restart sshd.service
➜ ~ git:(master) ✗ sudo systemctl status sshd.service
● ssh.service - OpenBSD Secure Shell server
Loaded: loaded (/etc/systemd/system/ssh.service; enabled; vendor preset: enabled)
Active: active (running) since Sun 2021-08-15 20:24:47 CEST; 4s ago
Docs: man:sshd(8)
man:sshd_config(5)
Process: 1111232 ExecStartPre=/usr/sbin/sshd -t (code=exited, status=0/SUCCESS)
Main PID: 1111233 (sshd)
Tasks: 1 (limit: 23851)
Memory: 1.1M
CPU: 15ms
CGroup: /system.slice/ssh.service
└─1111233 sshd: /usr/sbin/sshd -D -p 2200 [listener] 0 of 10-100 startups
Aug 15 20:24:47 deathstar systemd[1]: Starting OpenBSD Secure Shell server...
Aug 15 20:24:47 deathstar sshd[1111233]: Server listening on 0.0.0.0 port 2200.
Aug 15 20:24:47 deathstar sshd[1111233]: Server listening on :: port 2200.
Aug 15 20:24:47 deathstar systemd[1]: Started OpenBSD Secure Shell server.
➜ ~ git:(master) ✗
Notice something different here?
The location of the unit file is no longer /lib/systemd/system/ssh.service
but /etc/systemd/system/ssh.service
.
This is the actual preferred way of modifying unit files supplied by your distribution because if at some point in the future your distro changes it's configuration file and you update, you'll overwrite your custom changes! (see this post on serverfault)
Think of the similar situation we encountered with /etc/dnsmask.d/
when installing a pihole.
What if you want to revert
back to file supplied by Debian?
A quick sudo systemctl revert sshd.service
should do the trick!
Don't forget to daemon-reload
when you want to restart the service.
Writing your own service files
Imagine we want to run a custom server each time the machine boots. Here systemd comes to the rescue, plus we can run them as ourselves and don't need to interfere with the standard system services. Let's give this a go!
A simple example to a server would be a small python3 webserver.
Let's create a directory in our home called website.
We can do this with mkdir ~/website
.
In this folder we'll make an index.html
file where we add the content of our website to.
You can write anything you want, in html or plaintext.
To spin up a quick webserver we can use the http.server
class from the standard library.
I must note that it's not a production proof server and should only be used for small testing purposes (and for our example).
➜ website git:(master) ✗ python3 -m http.server 8080
Serving HTTP on 0.0.0.0 port 8080 (http://0.0.0.0:8080/) ...
127.0.0.1 - - [15/Aug/2021 20:45:30] "GET / HTTP/1.1" 200 -
127.0.0.1 - - [15/Aug/2021 20:45:30] code 404, message File not found
127.0.0.1 - - [15/Aug/2021 20:45:30] "GET /favicon.ico HTTP/1.1" 404 -
The website is now up and running and we can see all requests logged to the command line.
Go to http://localhost:8080
to see your website, or to the IP address of one of the other students to see the logs grow.
Right, we're happy with our service and we would like to offer it permanently.
In order to do so we need to create our own unit file and we can do this in two locations.
The first one is /etc/systemd/system/
which houses most of our system services but users can have their own services!
In order to create your own service, without root privileges, you can add unit files to ~/.local/share/systemd/user
.
You will probably have to create this directory.
In this directory you can add as many .service
files as you want.
For now we'll just make one called website.service
where we need to define some things in.
[Unit]
Description=Our own webserver
[Service]
WorkingDirectory=/home/waldek/website
ExecStart=/usr/bin/python3 -m http.server 8080
[Install]
WantedBy=default.target
Next we need to enable
and start
our service.
Notice that I'm not using sudo
and that I added the --user
argument.
systemctl --user enable website.service
systemctl --user start website.service
And we can inspect the logs via the trusted status
argument as such.
➜ ~ git:(master) ✗ systemctl --user status website.service
● website.service - Our own webserver
Loaded: loaded (/home/waldek/.local/share/systemd/user/website.service; enabled; vendor preset: enabled)
Active: active (running) since Sun 2021-08-15 20:58:15 CEST; 5min ago
Main PID: 1114451 (python3)
Tasks: 1 (limit: 23851)
Memory: 8.8M
CPU: 92ms
CGroup: /user.slice/user-1000.slice/user@1000.service/app.slice/website.service
└─1114451 /usr/bin/python3 -m http.server 8080
Aug 15 20:58:15 deathstar systemd[585]: Started Our own webserver.
Aug 15 20:58:23 deathstar python3[1114451]: 127.0.0.1 - - [15/Aug/2021 20:58:23] "GET / HTTP/1.1" 304 -
Aug 15 20:58:24 deathstar python3[1114451]: 127.0.0.1 - - [15/Aug/2021 20:58:24] "GET / HTTP/1.1" 304 -
Aug 15 20:58:24 deathstar python3[1114451]: 127.0.0.1 - - [15/Aug/2021 20:58:24] "GET / HTTP/1.1" 304 -
Aug 15 20:58:25 deathstar python3[1114451]: 127.0.0.1 - - [15/Aug/2021 20:58:25] "GET / HTTP/1.1" 304 -
➜ ~ git:(master) ✗
For those that want to dive deeper into the syntax of the configuration file you should have a look at the output of systemctl --user show website.service
which list all of the hidden settings that are predefined for a service.
To see what you can change them to, have a look here.
Deep dive into the logs
All logs made you systemd go into the /var/log/daemon.log
file by default.
You can override this but I would highly advise you not to do it as there are special tools that come with systemd to inspect the logs, plus all logs in one place is quite handy for grepping.
Have a look at the file and you should see a similar output.
Aug 15 20:24:44 deathstar systemd[1]: Reloading.
Aug 15 20:24:47 deathstar systemd[1]: Stopping OpenBSD Secure Shell server...
Aug 15 20:24:47 deathstar systemd[1]: ssh.service: Succeeded.
Aug 15 20:24:47 deathstar systemd[1]: Stopped OpenBSD Secure Shell server.
Aug 15 20:24:47 deathstar systemd[1]: Starting OpenBSD Secure Shell server...
Aug 15 20:24:47 deathstar systemd[1]: Started OpenBSD Secure Shell server.
Aug 15 20:57:38 deathstar systemd[585]: Started VTE child process 1114299 launched by gnome-terminal-server process 1027574.
Aug 15 20:58:02 deathstar systemd[585]: Reloading.
Aug 15 20:58:15 deathstar systemd[585]: Started Our own webserver.
Aug 15 20:58:23 deathstar python3[1114451]: 127.0.0.1 - - [15/Aug/2021 20:58:23] "GET / HTTP/1.1" 304 -
Aug 15 20:58:24 deathstar python3[1114451]: 127.0.0.1 - - [15/Aug/2021 20:58:24] "GET / HTTP/1.1" 304 -
Aug 15 20:58:24 deathstar python3[1114451]: 127.0.0.1 - - [15/Aug/2021 20:58:24] "GET / HTTP/1.1" 304 -
Aug 15 20:58:25 deathstar python3[1114451]: 127.0.0.1 - - [15/Aug/2021 20:58:25] "GET / HTTP/1.1" 304 -
Systemd comes with a specialized program to sift through it's logs called journalctl
.
Just invoking journalctl
will give you the output of the log file in less.
A very handy argument you'll probably always use is -e
which scrolls to the end of the logs.
As an alternative you can add --no-pager
which will not pipe to less
but just print to STDOUT.
To only view a specific service we can add the --unit
argument, followed by the service name.
For example:
➜ ~ git:(master) ✗ sudo journalctl --unit ssh.service --no-pager --since "1 h 25 min ago"
-- Journal begins at Wed 2021-07-14 22:35:36 CEST, ends at Sun 2021-08-15 21:46:42 CEST. --
Aug 15 20:22:14 deathstar sshd[1110635]: Received signal 15; terminating.
Aug 15 20:22:14 deathstar systemd[1]: Stopping OpenBSD Secure Shell server...
Aug 15 20:22:14 deathstar systemd[1]: ssh.service: Succeeded.
Aug 15 20:22:14 deathstar systemd[1]: Stopped OpenBSD Secure Shell server.
Aug 15 20:22:14 deathstar systemd[1]: Starting OpenBSD Secure Shell server...
Aug 15 20:22:14 deathstar sshd[1110849]: Server listening on 0.0.0.0 port 2222.
Aug 15 20:22:14 deathstar sshd[1110849]: Server listening on :: port 2222.
Aug 15 20:22:14 deathstar systemd[1]: Started OpenBSD Secure Shell server.
Aug 15 20:24:47 deathstar systemd[1]: Stopping OpenBSD Secure Shell server...
Aug 15 20:24:47 deathstar sshd[1110849]: Received signal 15; terminating.
Aug 15 20:24:47 deathstar systemd[1]: ssh.service: Succeeded.
Aug 15 20:24:47 deathstar systemd[1]: Stopped OpenBSD Secure Shell server.
Aug 15 20:24:47 deathstar systemd[1]: Starting OpenBSD Secure Shell server...
Aug 15 20:24:47 deathstar sshd[1111233]: Server listening on 0.0.0.0 port 2200.
Aug 15 20:24:47 deathstar sshd[1111233]: Server listening on :: port 2200.
Aug 15 20:24:47 deathstar systemd[1]: Started OpenBSD Secure Shell server.
➜ ~ git:(master) ✗
To understand the --since
argument I advise you to read the man systemd.time
pages.
An argument you'll often see suggested online is -x
.
It adds more verbose output to debug issues.
The manpage documentation is below for reference purpose for reference purposes.
-x, --catalog
Augment log lines with explanation texts from the message catalog. This will add explanatory help texts to log messages
in the output where this is available. These short help texts will explain the context of an error or log event,
possible solutions, as well as pointers to support forums, developer documentation, and any other relevant manuals. Note
that help texts are not available for all messages, but only for selected ones. For more information on the message
catalog, please refer to the Message Catalog Developer Documentation[5].
Note: when attaching journalctl output to bug reports, please do not use -x.
Last but not least, the -f
argument does a live stream of the log so you can debug on the fly.
This can be very handy in a tmux
session.
For more information I highly advise the man pages with man journalctl
!
A sidetrack into cron
But what if we want to run a quick script or command every day at midnight?
Like an email report of the system status, or a apt update
?
This can also be done with systemd but the classic way of doing this is via cron
.
As always, have a look at man cron
and when you're finished you'll know you want to read the man crontab
as well.
In short, every user can have a crontab which is a list of command to execute at certain intervals.
To inspect your own crontab, just execute crontab -e
which will open your editor of choice.
Read through the comments, it's quite self explanatory no?
Only the timestamp syntax is quite annoying in my opinion but there is a handy website to help you understand it a bit better.
To have a command executed every minute you add the following.
* * * * * echo "helloword" >> /tmp/coucou
The root
user has his own crontab you can edit with sudo crontab -e
To do an apt update
every day at midnight you would add the following.
0 0 * * * apt update
I must note that this is not really the best way to accomplish automatic update and upgrades. Have a look here for better alternatives.
Systemd timers
As you can probably see, cron
is a very basic but powerful way of scheduling actions.
So people really like the simplicity bit for others a bit more control is desired, hence man systemd.timer
.
We can list all current timers with the following command.
➜ ~ git:(master) ✗ systemctl list-timers --no-pager
NEXT LEFT LAST PASSED UNIT ACTIVATES
Mon 2021-08-16 00:00:00 C… 1h 40min left Sun 2021-08-15 00:00:13 C… 22h ago logrotate.timer logrotate.service
Mon 2021-08-16 00:00:00 C… 1h 40min left Sun 2021-08-15 00:00:13 C… 22h ago man-db.timer man-db.service
Mon 2021-08-16 06:52:37 C… 8h left Sun 2021-08-15 06:34:23 C… 15h ago apt-daily-upgrade.timer apt-daily-upgrade.service
Mon 2021-08-16 12:45:13 C… 14h left Sun 2021-08-15 12:45:13 C… 9h ago systemd-tmpfiles-clean.ti… systemd-tmpfiles-clean.ser…
Mon 2021-08-16 12:46:51 C… 14h left Sun 2021-08-15 19:56:29 C… 2h 23min ago apt-daily.timer apt-daily.service
Sun 2021-08-22 03:10:26 C… 6 days left Sun 2021-08-15 03:10:52 C… 19h ago e2scrub_all.timer e2scrub_all.service
6 timers listed.
Pass --all to see loaded but inactive timers, too.
➜ ~ git:(master) ✗
To list your own timers you add the --user
argument.
➜ ~ git:(master) ✗ systemctl --user list-timers --no-pager
NEXT LEFT LAST PASSED UNIT ACTIVATES
0 timers listed.
Pass --all to see loaded but inactive timers, too.
➜ ~ git:(master) ✗
Let's add one!
In order to create a timer, we need a service that we can run so let's do that first.
In the same folder as before I'll create a monitor.service
file and will add the following to it.
➜ user git:(master) ✗ systemctl --user cat monitor.service
# /home/waldek/.local/share/systemd/user/monitor.service
[Unit]
Description=Doing some monitoring
Wants=monitor.timer
[Service]
Type=oneshot
ExecStart=/usr/bin/ps u
[Install]
WantedBy=multi-user.target
➜ user git:(master) ✗
Now, let's test the service.
➜ user git:(master) ✗ systemctl --user start monitor.service
➜ user git:(master) ✗ systemctl --user status monitor.service
● monitor.service - Doing some monitoring
Loaded: loaded (/home/waldek/.local/share/systemd/user/monitor.service; disabled; vendor preset: enabled)
Active: inactive (dead)
Aug 15 22:29:08 deathstar ps[1122931]: waldek 1099261 0.0 0.0 16300 9880 pts/2 Ss 17:54 0:00 zsh
Aug 15 22:29:08 deathstar ps[1122931]: waldek 1101477 0.0 0.0 17468 11496 pts/3 Ss 18:31 0:06 zsh
Aug 15 22:29:08 deathstar ps[1122931]: waldek 1117521 0.0 0.0 16744 10068 pts/4 Ss 21:38 0:01 zsh
Aug 15 22:29:08 deathstar ps[1122931]: waldek 1118424 0.0 0.0 17120 10052 pts/5 Ss+ 21:44 0:00 zsh
Aug 15 22:29:08 deathstar ps[1122931]: waldek 1119048 0.0 0.0 40788 7956 pts/4 S+ 21:50 0:00 journalctl -f
Aug 15 22:29:08 deathstar ps[1122931]: waldek 1121204 1.9 0.1 45280 31108 pts/2 S+ 22:11 0:20 vim learning_systemd.md
Aug 15 22:29:08 deathstar ps[1122931]: waldek 1122718 0.2 0.0 12848 8716 pts/6 Ss+ 22:27 0:00 zsh
Aug 15 22:29:08 deathstar ps[1122931]: waldek 1122930 0.0 0.0 10072 1168 pts/3 S+ 22:29 0:00 systemctl --user start monit>
Aug 15 22:29:08 deathstar systemd[585]: monitor.service: Succeeded.
Aug 15 22:29:08 deathstar systemd[585]: Finished Doing some monitoring.
OK, that seems to work well, so now we want to create the timer that will run this service every minute.
In order to do this we create a monitor.timer
file with the following content.
The onCalendar=*-*-* *:*:00
make it run every minute.
You can read more about the syntax in the man systemd.time
pages.
➜ user git:(master) ✗ systemctl --user cat monitor.timer
# /home/waldek/.local/share/systemd/user/monitor.timer
[Unit]
Description=Doing some timely monitoring
Requires=monitor.service
[Timer]
Unit=monitor.service
OnCalendar=*-*-* *:*:00
[Install]
WantedBy=timers.target
➜ user git:(master) ✗
Once the timer is in place you should start
the service with systemctl --user start monitor.service
.
There is no need to start or enable the monitor.timer
file as the link between them is in the monitor.service
file via the Wants=monitor.timer
configuration line.
If you now watch your log in real time with journalctl -f --user-unit monitor.service
you should see your service executing every minute!
Pro's and cons
The following advice was taken from the arch wiki.
As a cron replacement
Although cron is arguably the most well-known job scheduler, systemd timers can be an alternative.
Benefits
The main benefits of using timers come from each job having its own systemd service. Some of these benefits are:
- Jobs can be easily started independently of their timers. This simplifies debugging.
- Each job can be configured to run in a specific environment (see systemd.exec(5)).
- Jobs can be attached to cgroups.
- Jobs can be set up to depend on other systemd units.
- Jobs are logged in the systemd journal for easy debugging.
Caveats
Some things that are easy to do with cron are difficult to do with timer units alone:
- Creation: to set up a timed job with systemd you need to create two files and run systemctl commands, compared to adding a single line to a crontab.
- Emails: there is no built-in equivalent to cron's MAILTO for sending emails on job failure. See the next section for an example of setting up a similar functionality using OnFailure=.
Also note that user timer units will only run during an active user login session by default. However, lingering can enable services to run at boot even when the user has no active login session.
A sidetrack into runlevels
The world of Linux has a concept called runlevels which determines a target state the machine is in, or to which you want the manche to go to. It's a complicated way of saying fully operational with graphical interface, a root only rescue mode, a reboot, halted etc. The official specification of the runlevels defines them as such.
- Runlevel 0 or Halt is used to shift the computer from one state to another. It shut down the system.
- Runlevel 1, s, S or Single-User Mode is used for administrative and recovery functions. It has only enough daemons to allow one user (the root user) to log in and perform system maintenance tasks. All local file systems are mounted. Some essential services are started, but networking remains disabled.
- Runlevel 2 or Multi-user Mode is used for most daemons running and allows multiple users the ability to log in and use system services but without networking. On Debian and its derivatives, a full multi-user mode with X running and a graphical login. Most other distributions leave this runlevel undefined.
- Runlevel 3 or Extended Multi-user Mode is used for a full multi-user mode with a console (without GUI) login screen with network services available
- Runlevel 4 is not normally used and undefined so it can be used for a personal customization
- Runlevel 5 or Graphical Mode is same as Runlevel 3 with graphical login (such as GDN).
- Runlevel 6 or Reboot is a transitional runlevel to reboot the system.
You can inspect the runlevel your system is currenty at by ececuting the following command.
➜ ~ git:(master) ✗ sudo runlevel
N 5
➜ ~ git:(master) ✗
You can change your runlevel with the sudo telinit
, followed by the level number, command.
You'll probably won't see that much difference between levels but try to change it to level 6
and see what happens.
If you change the runlevel to 1
your machine will probably freeze.
This has to do with the fact we haven't set a root
password on most of our machines so the single user mode can't be accessed.
Try setting a root password and reset the level to one and see what happens.
Systemd targets
Systemd take the concept of runlevels a bit further and they are renamed to targets. The mapping of runlevels to targets is as follows.
- poweroff.target (runlevel 0): shutdown and power off the system
- rescue.target (runlevel 1): launch the rescue shell session
- multi-user.target (runlevel 2,3,4): set the system in non graphical (console) multi-user system
- graphical.target (runlevel 5): use a graphical multi-user system with network services
- reboot.target (runlevel 6): shutdown and reboot the system
But, there are a lot more targets available on a machine running systemd.
Luckily systemctl
offers a nice way to inspect them.
➜ ~ git:(master) ✗ sudo systemctl list-units --type target
UNIT LOAD ACTIVE SUB DESCRIPTION
basic.target loaded active active Basic System
cryptsetup.target loaded active active Local Encrypted Volumes
getty.target loaded active active Login Prompts
local-fs-pre.target loaded active active Local File Systems (Pre)
local-fs.target loaded active active Local File Systems
multi-user.target loaded active active Multi-User System
network.target loaded active active Network
nfs-client.target loaded active active NFS client services
paths.target loaded active active Paths
remote-fs-pre.target loaded active active Remote File Systems (Pre)
remote-fs.target loaded active active Remote File Systems
rpcbind.target loaded active active RPC Port Mapper
slices.target loaded active active Slices
sockets.target loaded active active Sockets
swap.target loaded active active Swap
sysinit.target loaded active active System Initialization
time-set.target loaded active active System Time Set
time-sync.target loaded active active System Time Synchronized
timers.target loaded active active Timers
LOAD = Reflects whether the unit definition was properly loaded.
ACTIVE = The high-level unit activation state, i.e. generalization of SUB.
SUB = The low-level unit activation state, values depend on unit type.
19 loaded units listed. Pass --all to see loaded but inactive units, too.
To show all installed unit files use 'systemctl list-unit-files'.
Notice how some of the mappings, such as rescue.target, are missing?
We can show the inactive ones as well if we add the --all
argument.
➜ ~ git:(master) ✗ sudo systemctl list-units --type target --all --no-pager
UNIT LOAD ACTIVE SUB DESCRIPTION
basic.target loaded active active Basic System
blockdev@dev-disk-by\x2duuid-4a77d180\x2dfc64\x2d4057… loaded inactive dead Block Device Preparation for /dev/disk/by-uuid/4a77d18…
blockdev@dev-dm\x2d1.target loaded inactive dead Block Device Preparation for /dev/dm-1
blockdev@dev-mapper-deathstar\x2d\x2dvg\x2droot.target loaded inactive dead Block Device Preparation for /dev/mapper/deathstar--vg…
blockdev@dev-mapper-deathstar\x2d\x2dvg\x2dswap_1.tar… loaded inactive dead Block Device Preparation for /dev/mapper/deathstar--vg…
blockdev@dev-sda1.target loaded inactive dead Block Device Preparation for /dev/sda1
bluetooth.target loaded inactive dead Bluetooth
cryptsetup.target loaded active active Local Encrypted Volumes
emergency.target loaded inactive dead Emergency Mode
first-boot-complete.target loaded inactive dead First Boot Complete
getty-pre.target loaded inactive dead Login Prompts (Pre)
getty.target loaded active active Login Prompts
graphical.target loaded inactive dead Graphical Interface
local-fs-pre.target loaded active active Local File Systems (Pre)
local-fs.target loaded active active Local File Systems
multi-user.target loaded active active Multi-User System
network-online.target loaded inactive dead Network is Online
network-pre.target loaded inactive dead Network (Pre)
network.target loaded active active Network
nfs-client.target loaded active active NFS client services
nss-user-lookup.target loaded inactive dead User and Group Name Lookups
paths.target loaded active active Paths
remote-fs-pre.target loaded active active Remote File Systems (Pre)
remote-fs.target loaded active active Remote File Systems
rescue.target loaded inactive dead Rescue Mode
rpcbind.target loaded active active RPC Port Mapper
shutdown.target loaded inactive dead Shutdown
slices.target loaded active active Slices
sockets.target loaded active active Sockets
sound.target loaded inactive dead Sound Card
swap.target loaded active active Swap
sysinit.target loaded active active System Initialization
time-set.target loaded active active System Time Set
time-sync.target loaded active active System Time Synchronized
timers.target loaded active active Timers
umount.target loaded inactive dead Unmount All Filesystems
LOAD = Reflects whether the unit definition was properly loaded.
ACTIVE = The high-level unit activation state, i.e. generalization of SUB.
SUB = The low-level unit activation state, values depend on unit type.
36 loaded units listed.
To show all installed unit files use 'systemctl list-unit-files'.
➜ ~ git:(master) ✗
That's better but still, some other ones such as poweroff.target seem to be missing. Those are both not active and not loaded, but still available. We can list all unit files known to our system with a different command.
➜ ~ git:(master) ✗ sudo systemctl list-unit-files --type target --all --no-pager
UNIT FILE STATE VENDOR PRESET
basic.target static -
blockdev@.target static -
bluetooth.target static -
boot-complete.target static -
cryptsetup-pre.target static -
cryptsetup.target static -
ctrl-alt-del.target alias -
default.target alias -
emergency.target static -
exit.target disabled disabled
final.target static -
first-boot-complete.target static -
getty-pre.target static -
getty.target static -
graphical.target static -
halt.target disabled disabled
hibernate.target static -
hybrid-sleep.target static -
initrd-fs.target static -
initrd-root-device.target static -
initrd-root-fs.target static -
initrd-switch-root.target static -
initrd.target static -
kexec.target disabled disabled
local-fs-pre.target static -
local-fs.target static -
multi-user.target static -
network-online.target static -
network-pre.target static -
network.target static -
nfs-client.target enabled enabled
nss-lookup.target static -
nss-user-lookup.target static -
paths.target static -
poweroff.target disabled disabled
printer.target static -
reboot.target disabled enabled
remote-cryptsetup.target disabled enabled
remote-fs-pre.target static -
remote-fs.target enabled enabled
rescue-ssh.target static -
rescue.target static -
rpcbind.target static -
runlevel0.target alias -
runlevel1.target alias -
runlevel2.target alias -
runlevel3.target alias -
runlevel4.target alias -
runlevel5.target alias -
runlevel6.target alias -
shutdown.target static -
sigpwr.target static -
sleep.target static -
slices.target static -
smartcard.target static -
sockets.target static -
sound.target static -
suspend-then-hibernate.target static -
suspend.target static -
swap.target static -
sysinit.target static -
system-update-pre.target static -
system-update.target static -
time-set.target static -
time-sync.target static -
timers.target static -
umount.target static -
usb-gadget.target static -
68 unit files listed.
➜ ~ git:(master) ✗
That seems to be complete.
Now, how do we switch form one target to an other in a modern systemd-like fashion?
For this we use the isolate
argument to systemctl
.
A quick test of a this can be done as such, sudo systemctl isolate reboot.target
.
On a Linux system where root has a password set you could try the rescue.target
as well.
You can get and set the default runlevel of you system with the following commands.
➜ ~ git:(master) ✗ sudo systemctl get-default
graphical.target
➜ ~ git:(master) ✗ sudo systemctl set-default multi-user.target
Created symlink /etc/systemd/system/default.target → /lib/systemd/system/multi-user.target.
➜ ~ git:(master) ✗ sudo systemctl get-default
multi-user.target
➜ ~ git:(master) ✗
A deeper look into targets
What is included in all of these targets?
We can inspect their dependencies by invoking the list-dependencies
argument to systemctl
.
Let's start with the most basic one, the rescue.target
.
➜ ~ git:(master) ✗ sudo systemctl list-dependencies rescue.target --no-pager
rescue.target
● ├─rescue.service
● ├─systemd-update-utmp-runlevel.service
● └─sysinit.target
● ├─apparmor.service
● ├─blk-availability.service
● ├─dev-hugepages.mount
● ├─dev-mqueue.mount
● ├─keyboard-setup.service
● ├─kmod-static-nodes.service
● ├─lvm2-lvmpolld.socket
● ├─lvm2-monitor.service
● ├─proc-sys-fs-binfmt_misc.automount
● ├─sys-fs-fuse-connections.mount
● ├─sys-kernel-config.mount
● ├─sys-kernel-debug.mount
● ├─sys-kernel-tracing.mount
● ├─systemd-ask-password-console.path
● ├─systemd-binfmt.service
● ├─systemd-boot-system-token.service
● ├─systemd-hwdb-update.service
● ├─systemd-journal-flush.service
● ├─systemd-journald.service
● ├─systemd-machine-id-commit.service
● ├─systemd-modules-load.service
● ├─systemd-pstore.service
● ├─systemd-random-seed.service
● ├─systemd-sysctl.service
● ├─systemd-sysusers.service
● ├─systemd-timesyncd.service
● ├─systemd-tmpfiles-setup-dev.service
● ├─systemd-tmpfiles-setup.service
● ├─systemd-udev-trigger.service
● ├─systemd-udevd.service
● ├─systemd-update-utmp.service
● ├─cryptsetup.target
● ├─local-fs.target
● │ ├─-.mount
● │ ├─boot.mount
● │ ├─systemd-fsck-root.service
● │ └─systemd-remount-fs.service
● └─swap.target
● └─dev-mapper-deathstar\x2d\x2dvg\x2dswap_1.swap
➜ ~ git:(master) ✗
As you can see, it's quite basic.
All of these services and additional targets will try to be loaded and started when we enter rescue mode.
Now, let's compare it to the most elaborate runlevel, 5
.
➜ ~ git:(master) ✗ sudo systemctl list-dependencies graphical.target --no-pager
graphical.target
● ├─display-manager.service
● ├─systemd-update-utmp-runlevel.service
● ├─udisks2.service
● └─multi-user.target
● ├─avahi-daemon.service
● ├─binfmt-support.service
● ├─blueman-mechanism.service
● ├─chrony.service
● ├─console-setup.service
● ├─cron.service
● ├─cups-browsed.service
● ├─cups.path
● ├─dbus.service
● ├─e2scrub_reap.service
● ├─ModemManager.service
● ├─networking.service
● ├─rpcbind.service
● ├─rsyslog.service
● ├─ssh.service
● ├─systemd-ask-password-wall.path
● ├─systemd-logind.service
● ├─systemd-update-utmp-runlevel.service
● ├─systemd-user-sessions.service
● ├─wpa_supplicant.service
● ├─basic.target
● │ ├─-.mount
● │ ├─tmp.mount
● │ ├─paths.target
● │ ├─slices.target
● │ │ ├─-.slice
● │ │ └─system.slice
● │ ├─sockets.target
● │ │ ├─avahi-daemon.socket
● │ │ ├─cups.socket
● │ │ ├─dbus.socket
● │ │ ├─dm-event.socket
● │ │ ├─pcscd.socket
● │ │ ├─rpcbind.socket
● │ │ ├─systemd-initctl.socket
● │ │ ├─systemd-journald-audit.socket
● │ │ ├─systemd-journald-dev-log.socket
● │ │ ├─systemd-journald.socket
● │ │ ├─systemd-udevd-control.socket
● │ │ └─systemd-udevd-kernel.socket
● │ ├─sysinit.target
● │ │ ├─apparmor.service
● │ │ ├─blk-availability.service
● │ │ ├─dev-hugepages.mount
● │ │ ├─dev-mqueue.mount
● │ │ ├─keyboard-setup.service
● │ │ ├─kmod-static-nodes.service
● │ │ ├─lvm2-lvmpolld.socket
● │ │ ├─lvm2-monitor.service
● │ │ ├─proc-sys-fs-binfmt_misc.automount
● │ │ ├─sys-fs-fuse-connections.mount
● │ │ ├─sys-kernel-config.mount
● │ │ ├─sys-kernel-debug.mount
● │ │ ├─sys-kernel-tracing.mount
● │ │ ├─systemd-ask-password-console.path
● │ │ ├─systemd-binfmt.service
● │ │ ├─systemd-boot-system-token.service
● │ │ ├─systemd-hwdb-update.service
● │ │ ├─systemd-journal-flush.service
● │ │ ├─systemd-journald.service
● │ │ ├─systemd-machine-id-commit.service
● │ │ ├─systemd-modules-load.service
● │ │ ├─systemd-pstore.service
● │ │ ├─systemd-random-seed.service
● │ │ ├─systemd-sysctl.service
● │ │ ├─systemd-sysusers.service
● │ │ ├─systemd-timesyncd.service
● │ │ ├─systemd-tmpfiles-setup-dev.service
● │ │ ├─systemd-tmpfiles-setup.service
● │ │ ├─systemd-udev-trigger.service
● │ │ ├─systemd-udevd.service
● │ │ ├─systemd-update-utmp.service
● │ │ ├─cryptsetup.target
● │ │ ├─local-fs.target
● │ │ │ ├─-.mount
● │ │ │ ├─boot.mount
● │ │ │ ├─systemd-fsck-root.service
● │ │ │ └─systemd-remount-fs.service
● │ │ └─swap.target
● │ │ └─dev-mapper-deathstar\x2d\x2dvg\x2dswap_1.swap
● │ └─timers.target
● │ ├─apt-daily-upgrade.timer
● │ ├─apt-daily.timer
● │ ├─e2scrub_all.timer
● │ ├─logrotate.timer
● │ ├─man-db.timer
● │ └─systemd-tmpfiles-clean.timer
● ├─getty.target
● │ ├─getty-static.service
● │ └─getty@tty1.service
● ├─nfs-client.target
● │ ├─auth-rpcgss-module.service
● │ ├─nfs-blkmap.service
● │ └─remote-fs-pre.target
● └─remote-fs.target
● └─nfs-client.target
● ├─auth-rpcgss-module.service
● ├─nfs-blkmap.service
● └─remote-fs-pre.target
➜ ~ git:(master) ✗
You immediately see, and probably recognise a lot of very useful services that get launched when we enter the graphical target. Mind you that the output above is from a pretty lean system running a minimal i3 graphical environment.
We can also use the list-dependencies
to inspect services such as sshd.service
.
The list below is everything sshd depends on to succesfully run as a systemd service.
sshd.service
● ├─-.mount
● ├─system.slice
● └─sysinit.target
● ├─apparmor.service
● ├─blk-availability.service
● ├─dev-hugepages.mount
● ├─dev-mqueue.mount
● ├─keyboard-setup.service
● ├─kmod-static-nodes.service
● ├─lvm2-lvmpolld.socket
● ├─lvm2-monitor.service
● ├─proc-sys-fs-binfmt_misc.automount
● ├─sys-fs-fuse-connections.mount
● ├─sys-kernel-config.mount
● ├─sys-kernel-debug.mount
● ├─sys-kernel-tracing.mount
● ├─systemd-ask-password-console.path
● ├─systemd-binfmt.service
● ├─systemd-boot-system-token.service
● ├─systemd-hwdb-update.service
● ├─systemd-journal-flush.service
● ├─systemd-journald.service
● ├─systemd-machine-id-commit.service
● ├─systemd-modules-load.service
● ├─systemd-pstore.service
● ├─systemd-random-seed.service
● ├─systemd-sysctl.service
● ├─systemd-sysusers.service
● ├─systemd-timesyncd.service
● ├─systemd-tmpfiles-setup-dev.service
● ├─systemd-tmpfiles-setup.service
● ├─systemd-udev-trigger.service
● ├─systemd-udevd.service
● ├─systemd-update-utmp.service
● ├─cryptsetup.target
● ├─local-fs.target
● │ ├─-.mount
● │ ├─boot.mount
● │ ├─systemd-fsck-root.service
● │ └─systemd-remount-fs.service
● └─swap.target
● └─dev-mapper-deathstar\x2d\x2dvg\x2dswap_1.swap
➜ ~ git:(master) ✗
A very clever reverse dependency list can be show by adding the --reverse
argument.
The output below show the dependencies of the networking.service first.
You can see it needs the ifupdown-pre.service, system.slice and the network.target.
The second command shows the reverse, which services or targets depend on the networking.service to be up and running.
➜ ~ git:(master) ✗ sudo systemctl list-dependencies networking.service --no-pager
networking.service
● ├─ifupdown-pre.service
● ├─system.slice
● └─network.target
➜ ~ git:(master) ✗ sudo systemctl list-dependencies networking.service --no-pager --reverse
networking.service
● ├─multi-user.target
● │ └─graphical.target
● └─network-online.target
➜ ~ git:(master) ✗
The combination of both can give you a solid understanding of how all services and targets are interconnected. Just as with services we can inspect what a target is doing by looking at it's unit file.
➜ ~ git:(master) ✗ sudo systemctl cat network-online.target
# /lib/systemd/system/network-online.target
# SPDX-License-Identifier: LGPL-2.1-or-later
#
# This file is part of systemd.
#
# systemd is free software; you can redistribute it and/or modify it
# under the terms of the GNU Lesser General Public License as published by
# the Free Software Foundation; either version 2.1 of the License, or
# (at your option) any later version.
[Unit]
Description=Network is Online
Documentation=man:systemd.special(7)
Documentation=https://www.freedesktop.org/wiki/Software/systemd/NetworkTarget
After=network.target
➜ ~ git:(master) ✗
This might not tell you all that much on first sight, but I urge you to take the time out to really read the man systemd.special
.
It will explain you all the intricacies of the different standard targets and how you can use them to your benefit.