101.3 Lesson 1

Certificate:

LPIC-1

Version:

5.0

Topic:

101 System Architecture

Objective:

101.3 Change runlevels / boot targets and shutdown or reboot system

Lesson:

1 of 1

Introduction

A common feature among operating systems following Unix design principles is the employment of separate processes to control distinct functions of the system. These processes, called daemons (or, more generally, services), are also responsible for extended features underlying the operating system, like network application services (HTTP server, file sharing, email, etc.), databases, on-demand configuration, etc. Although Linux utilizes a monolithic kernel, many low level aspects of the operating system are affected by daemons, like load balancing and firewall configuration.

Which daemons should be active depend on the purpose of the system. The set of active daemons should also be modifiable at runtime, so services can be started and stopped without having to reboot the whole system. To tackle this issue, every major Linux distribution offers some form of service management utility to control the system.

Services can be controlled by shell scripts or by a program and its supporting configuration files. The first method is implemented by the SysVinit standard, also known as System V or just SysV. The second method is implemented by systemd and Upstart. Historically, SysV based service managers were the most used by Linux distributions. Today, systemd based service managers are more often found in most Linux distributions. The service manager is the first program launched by the kernel during the boot process, so its PID (process identification number) is always 1.

SysVinit

A service manager based on the SysVinit standard will provide predefined sets of system states, called runlevels, and their corresponding service script files to be executed. Runlevels are numbered 0 to 6, being generally assigned to the following purposes:

Runlevel 0: System shutdown.
Runlevel 1, s or single: Single user mode, without network and other non-essential capabilities (maintenance mode).
Runlevel 2, 3 or 4: Multi-user mode. Users can log in by console or network. Runlevels 2 and 4 are not often used.
Runlevel 5: Multi-user mode. It is equivalent to 3, plus the graphical mode login.
Runlevel 6: System restart.

The program responsible for managing runlevels and associated daemons/resources is /sbin/init. During system initialization, the init program identifies the requested runlevel, defined by a kernel parameter or in the /etc/inittab file, and loads the associated scripts listed there for the given runlevel. Every runlevel may have many associated service files, usually scripts in the /etc/init.d/ directory. As not all runlevels are equivalent through different Linux distributions, a short description of the runlevel’s purpose can also be found in SysV based distributions.

The syntax of the /etc/inittab file uses this format:

id:runlevels:action:process

The id is a generic name up to four characters in length used to identify the entry. The runlevels entry is a list of runlevel numbers for which a specified action should be executed. The action term defines how init will execute the process indicated by the term process. The available actions are:

boot: The process will be executed during system initialization. The field runlevels is ignored.
bootwait: The process will be executed during system initialization and init will wait until it finishes to continue. The field runlevels is ignored.
sysinit: The process will be executed after system initialization, regardless of runlevel. The field runlevels is ignored.
wait: The process will be executed for the given runlevels and init will wait until it finishes to continue.
respawn: The process will be restarted if it is terminated.
ctrlaltdel: The process will be executed when the init process receives the SIGINT signal, triggered when the key sequence of Ctrl+Alt+Del is pressed.

The default runlevel — the one that will be chosen if no other is given as a kernel parameter — is also defined in /etc/inittab, in the entry id:x:initdefault. The x is the number of the default runlevel. This number should never be 0 or 6, given that it would cause the system to shutdown or restart as soon as it finishes the boot process. A typical /etc/inittab file is shown below:

# Default runlevel
id:3:initdefault:

# Configuration script executed during boot
si::sysinit:/etc/init.d/rcS

# Action taken on runlevel S (single user)
~:S:wait:/sbin/sulogin

# Configuration for each execution level
l0:0:wait:/etc/init.d/rc 0
l1:1:wait:/etc/init.d/rc 1
l2:2:wait:/etc/init.d/rc 2
l3:3:wait:/etc/init.d/rc 3
l4:4:wait:/etc/init.d/rc 4
l5:5:wait:/etc/init.d/rc 5
l6:6:wait:/etc/init.d/rc 6

# Action taken upon ctrl+alt+del keystroke
ca::ctrlaltdel:/sbin/shutdown -r now

# Enable consoles for runlevels 2 and 3
1:23:respawn:/sbin/getty tty1 VC linux
2:23:respawn:/sbin/getty tty2 VC linux
3:23:respawn:/sbin/getty tty3 VC linux
4:23:respawn:/sbin/getty tty4 VC linux

# For runlevel 3, also enable serial
# terminals ttyS0 and ttyS1 (modem) consoles
S0:3:respawn:/sbin/getty -L 9600 ttyS0 vt320
S1:3:respawn:/sbin/mgetty -x0 -D ttyS1

The telinit q command should be executed every time after the /etc/inittab file is modified. The argument q (or Q) tells init to reload its configuration. Such a step is important to avoid a system halt due to an incorrect configuration in /etc/inittab.

The scripts used by init to setup each runlevel are stored in the directory /etc/init.d/. Every runlevel has an associated directory in /etc/, named /etc/rc0.d/, /etc/rc1.d/, /etc/rc2.d/, etc., with the scripts that should be executed when the corresponding runlevel starts. As the same script can be used by different runlevels, the files in those directories are just symbolic links to the actual scripts in /etc/init.d/. Furthermore, the first letter of the link filename in the runlevel’s directory indicates if the service should be started or terminated for the corresponding runlevel. A link’s filename starting with letter K determines that the service will be killed when entering the runlevel (kill). Starting with letter S, the service will be started when entering the runlevel (start). The directory /etc/rc1.d/, for example, will have many links to network scripts beginning with letter K, considering that the runlevel 1 is the single user runlevel, without network connectivity.

The command runlevel shows the current runlevel for the system. The runlevel command shows two values, the first is the previous runlevel and the second is the current runlevel:

$ runlevel
N 3

The letter N in the output shows that the runlevel has not changed since last boot. In the example, the runlevel 3 is the current runlevel of the system.

The same init program can be used to alternate between runlevels in a running system, without the need to reboot. The command telinit can also be used to alternate between runlevels. For example, commands telinit 1, telinit s or telinit S will change the system to runlevel 1.

systemd

Currently, systemd is the most widely used set of tools to manage system resources and services, which are referred to as units by systemd. A unit consists of a name, a type and a corresponding configuration file. For example, the unit for a httpd server process (like the Apache web server) will be httpd.service on Red Hat based distributions and its configuration file will also be called httpd.service (on Debian based distributions this unit is named apache2.service).

There are seven distinct types of systemd units:

service: The most common unit type, for active system resources that can be initiated, interrupted and reloaded.
socket: The socket unit type can be a filesystem socket or a network socket. All socket units have a corresponding service unit, loaded when the socket receives a request.
device: A device unit is associated with a hardware device identified by the kernel. A device will only be taken as a systemd unit if a udev rule for this purpose exists. A device unit can be used to resolve configuration dependencies when certain hardware is detected, given that properties from the udev rule can be used as parameters for the device unit.
mount: A mount unit is a mount point definition in the filesystem, similar to an entry in /etc/fstab.
automount: An automount unit is also a mount point definition in the filesystem, but mounted automatically. Every automount unit has a corresponding mount unit, which is initiated when the automount mount point is accessed.
target: A target unit is a grouping of other units, managed as a single unit.
snapshot: A snapshot unit is a saved state of the systemd manager (not available on every Linux distribution).

The main command for controlling systemd units is systemctl. Command systemctl is used to execute all tasks regarding unit activation, deactivation, execution, interruption, monitoring, etc. For a fictitious unit called unit.service, for example, the most common systemctl actions will be:

systemctl start unit.service: Starts unit.
systemctl stop unit.service: Stops unit.
systemctl restart unit.service: Restarts unit.
systemctl status unit.service: Shows the state of unit, including if it is running or not.
systemctl is-active unit.service: Shows active if unit is running or inactive otherwise.
systemctl enable unit.service: Enables unit, that is, unit will load during system initialization.
systemctl disable unit.service: unit will not start with the system.
systemctl is-enabled unit.service: Verifies if unit starts with the system. The answer is stored in the variable $?. The value 0 indicates that unit starts with the system and the value 1 indicates that unit does not starts with the system.

Note	Newer installations of systemd will actually list a unit’s configuration for boot time. For example: $ systemctl is-enabled apparmor.service enabled

If no other units with the same name exists in the system, then the suffix after the dot can be dropped. If, for example, there is only one httpd unit of type service, then only httpd is enough as the unit parameter for systemctl.

The systemctl command can also control system targets. The multi-user.target unit, for example, combines all units required by the multi-user system environment. It is similar to the runlevel number 3 in a system utilizing SysV.

Command systemctl isolate alternates between different targets. So, to manually alternate to target multi-user:

# systemctl isolate multi-user.target

There are corresponding targets to SysV runlevels, starting with runlevelO.target up to runlevel6.target. However, systemd does not use the /etc/inittab file. To change the default system target, the option systemd.unit can be added to the kernel parameters list. For example, to use multi-user.target as the standard target, the kernel parameter should be systemd.unit=multi-user.target. All kernel parameters can be made persistent by changing the bootloader configuration.

Another way to change the default target is to modify the symbolic link /etc/systemd/system/default.target so it points to the desired target. The redefinition of the link can be done with the systemctl command by itself:

# systemctl set-default multi-user.target

Likewise, you can determine what your system’s default boot target is with the following command:

$ systemctl get-default
graphical.target

Similar to systems adopting SysV, the default target should never point to shutdown.target, as it corresponds to the runlevel 0 (shutdown).

The configuration files associated with every unit can be found in the /lib/systemd/system/ directory. The command systemctl list-unit-files lists all available units and shows if they are enabled to start when the system boots. The option --type will select only the units for a given type, as in systemctl list-unit-files --type=service and systemctl list-unit-files --type=target.

Active units or units that have been active during the current system session can be listed with command systemctl list-units. Like the list-unit-files option, the systemctl list-units --type=service command will select only units of type service and command systemctl list-units --type=target will select only units of type target.

systemd is also responsible for triggering and responding to power related events. The systemctl suspend command will put the system in low power mode, keeping current data in memory. Command systemctl hibernate will copy all memory data to disk, so the current state of the system can be recovered after powering it off. The actions associated with such events are defined in the file /etc/systemd/logind.conf or in separate files inside the directory /etc/systemd/logind.conf.d/. However, this systemd feature can only be used when there is no other power manager running in the system, like the acpid daemon. The acpid daemon is the main power manager for Linux and allows finer adjustments to the actions following power related events, like closing the laptop lid, low battery or battery charging levels.

Upstart

The initialization scripts used by Upstart are located in the directory /etc/init/. System services can be listed with command initctl list, which also shows the current state of the services and, if available, their PID number.

# initctl list
avahi-cups-reload stop/waiting
avahi-daemon start/running, process 1123
mountall-net stop/waiting
mountnfs-bootclean.sh start/running
nmbd start/running, process 3085
passwd stop/waiting
rc stop/waiting
rsyslog start/running, process 1095
tty4 start/running, process 1761
udev start/running, process 1073
upstart-udev-bridge start/running, process 1066
console-setup stop/waiting
irqbalance start/running, process 1842
plymouth-log stop/waiting
smbd start/running, process 1457
tty5 start/running, process 1764
failsafe stop/waiting

Every Upstart action has its own independent command. For example, command start can be used to initiate a sixth virtual terminal:

# start tty6

The current state of a resource can be verified with command status:

# status tty6
tty6 start/running, process 3282

And the interruption of a service is done with the command stop:

# stop tty6

Upstart does not use the /etc/inittab file to define runlevels, but the legacy commands runlevel and telinit can still be used to verify and alternate between runlevels.

Note	Upstart was developed for the Ubuntu Linux distribution to help facilitate parallel startup of processes. Ubuntu has stopped using Upstart since 2015 when it switched from Upstart to systemd.

Shutdown and Restart

A very traditional command used to shutdown or restart the system is unsurprisingly called shutdown. The shutdown command adds extra functions to the power off process: it automatically notifies all logged-in users with a warning message in their shell sessions and new logins are prevented. Command shutdown acts as an intermediary to SysV or systemd procedures, that is, it executes the requested action by calling the corresponding action in the services manager adopted by the system.

After shutdown is executed, all processes receive the SIGTERM signal, followed by the SIGKILL signal, then the system shuts down or changes its runlevel. By default, when neither options -h or -r are used, the system alternates to runlevel 1, that is, the single user mode. To change the default options for shutdown, the command should be executed with the following syntax:

$ shutdown [option] time [message]

Only the parameter time is required. The time parameter defines when the requested action will be executed, accepting the following formats:

hh:mm: This format specifies the execution time as hour and minutes.
+m: This format specifies how many minutes to wait before execution.
now or +0: This format determines immediate execution.

The message parameter is the warning text sent to all terminal sessions of logged-in users.

The SysV implementation allows for the limiting of users that will be able to restart the machine by pressing Ctrl+Alt+Del. This is possible by placing option -a for the shutdown command present at the line regarding ctrlaltdel in the /etc/inittab file. By doing this, only users whose usernames are in the /etc/shutdown.allow file will be able to restart the system with the Ctrl+Alt+Del keystroke combination.

The systemctl command can also be used to turn off or to restart the machine in systems employing systemd. To restart the system, the command systemctl reboot should be used. To turn off the system, the command systemctl poweroff should be used. Both commands require root privileges to run, as ordinary users can not perform such procedures.

Note	Some Linux distributions will link `poweroff` and `reboot` to `systemctl` as individual commands. For example: $ sudo which poweroff /usr/sbin/poweroff $ sudo ls -l /usr/sbin/poweroff lrwxrwxrwx 1 root root 14 Aug 20 07:50 /usr/sbin/poweroff -> /bin/systemctl

Not all maintenance activities require the system to be turned off or restarted. However, when it is necessary to change the system’s state to single-user mode, it is important to warn logged-in users so that they are not harmed by an abrupt termination of their activities.

Similar to what the shutdown command does when powering off or restarting the system, the wall command is able to send a message to terminal sessions of all logged-in users. To do so, the system administrator only needs to provide a file or directly write the message as a parameter to command wall.

Guided Exercises

How could the telinit command be used to reboot the system?
What will happen to the services related to the file /etc/rc1.d/K90network when the system enters runlevel 1?
Using command systemctl, how could a user verify if the unit sshd.service is running?
In a systemd based system, what command must be executed to enable activation of the unit sshd.service during system initialization?

Explorational Exercises

In a SysV based system, suppose the default runlevel defined in /etc/inittab is 3, but the system always starts in runlevel 1. What is the probable cause for that?
Although the file /sbin/init can be found in systemd based systems, it is only a symbolic link to another executable file. In such systems, what is the file pointed by /sbin/init?
How can the default system target be verified in a systemd based system?
How can a system reboot scheduled with the shutdown command be canceled?

Summary

This lesson covers the main utilities used as service managers by Linux distributions. The SysVinit, systemd and Upstart utilities each have their own approach to control system services and system states. The lesson goes through the following topics:

What system services are and their role in the operating system.
Concepts and basic usage of SysVinit, systemd and Upstart commands.
How to properly start, stop and restart system services and the system itself.

The commands and procedures addressed were:

Commands and files related to SysVinit, like init, /etc/inittab and telinit.
The main systemd command: systemctl.
Upstart commands: initctl, status, start, stop.
Traditional power management commands, like shutdown, and command wall.

Answers to Guided Exercises

How could the telinit command be used to reboot the system?

The command telinit 6 will alternate to runlevel 6, that is, reboot the system.
What will happen to the services related to the file /etc/rc1.d/K90network when the system enters runlevel 1?

Due to the letter K in the beginning of the file name, the related services will be stopped.
Using command systemctl, how could a user verify if the unit sshd.service is running?

With the command systemctl status sshd.service or systemctl is-active sshd.service.
In a systemd based system, what command must be executed to enable activation of the unit sshd.service during system initialization?

Command systemctl enable sshd.service, executed by root.

Answers to Explorational Exercises

In a SysV based system, suppose the default runlevel defined in /etc/inittab is 3, but the system always starts in runlevel 1. What is the probable cause for that?

The parameters 1 or S may be present in the kernel’s parameter list.
Although file /sbin/init can be found in systemd based systems, it is only a symbolic link to another executable file. In such systems, what is the file pointed by /sbin/init?

The main systemd binary: /lib/systemd/systemd.
How can the default system target be verified in a systemd based system?

The symbolic link /etc/systemd/system/default.target will point to the unit file defined as the default target. Command systemctl get-default can also be used.
How can a system reboot scheduled with the shutdown command be canceled?

The command shutdown -c should be used.