• bootup — System bootup process

Rerun all generators, reload all unit files and recreate dependency tree:

systemctl daemon-reload

systemctl start servicename

# systemctl get-default
multi-user.target

systemd-analyze blame
systemd-analyze critical-chain
systemd-analyze critical-chain  network.service
systemctl list-dependencies network.service
systemctl --failed

systemd-cgtop

If services are defined with DefaultDependencies=no, they will be started unconditionally even in emergency mode, wchich can be wrong.

local-fs.target is configured to go to OnFailure=emergency.target. File /etc/fstab is parsed by systemd-fstab-generator to generate .mount units, which are Required.

To let mounts fail without switching to emergency mode, special attribute are introduced to fstab. http://www.freedesktop.org/software/systemd/man/systemd.mount.html

Kernel commandline:

systemd.unit=emergency

systemd.log_level=debug systemd.log_target=console console=ttyS0,38400

Enable in u-boot:

setenv bootargs ${bootargs} systemd.unit=emergency.target

Enable debug console on tty9

systemctl enable debug-shell.service

Create directory “dev-mmcblk0.device.wants” and put symlink to service inside it. List of devices:

systemctl --all --full -t device

Value for unit DefaultDependencies is set to yes by default. it will add following dependencies:

• Conflicts=shutdown.target to stop service during shutdown
• Conflicts=umount.target to umount mounts durign shutdown
• Requires=basic.target and After=basic.target

• .service plain service - for running daemons
• .socket socket listeners - like inetd
• .device
• .path filesystem triggers (when file is changed)
• .mount
• .automount
• .swap
• .target
• .timer
• .snapshot
• .slice
• .scope

• Requires - units listed here, are also activated with unit. If any start fail, this unit will be not activated.
• Conflicts - negative dependencies, opposite to Requires

• Requisite - similar to require, but if not met, service will be not started (fail immediately). Usefull to perform action when removable device is inserted.
• Requisite=!local-fs.target

• BindsTo - similar to require, but also declares to stop unit when other units dissapers (aslo nice for removable devices handling)
• PartOf - stops/starts unit when unit listed is stopped/started

• After - start service after other services are started completely.
• Before

• Type
  • Type=simple - default -
  • Type=forking - process will spawn child and exit. Child should live as daemon.
    • PIDFile=/var/run/service.pid - set it to give systemd possibility to known forked child PID
  • Type=oneshot - process should exit. Systemd is not waiting for finish. Like batch file.
    • usefull for setup some things, not starting daemons. See ReamainAfterExit=Yes)
    • Multiple ExecStart= allowed, and will be executed sequentially
  • Type=dbus - systemd assume process started when process name connects to D-Bus
  • Type=notify - systemd assume process started when process explicitly notify systemd using sd_notify()
  • Type=idle

• ExecStart= - first argument MUST be full path to executable
  • ExecStart=-/bin/false - minus sign = ignore exit code and assume always success
  • Multple ExecStart= are allowed, and executed sequentially. If one command fails, other lines are not executed and unit is failed.
• ExecStartPre= - ExecStart= will run only if all ExecStartPre= command exits with success. May not to start long processes.
• ExecStartPost= - will be executed only if service has started successfully
• ExecStop=
• ExecReload=

• Nice=-20
• IOSchedulingClass=0
• IOSchedulingPriority=0
• StandardOutput=tty |none | journal+console

• Restart=always
• RestartSec=10s

• RemainAfterExit=Yes - keep service active status, even all process exits. usefull for Type=oneshot

Condition doesn't skip service, but only service is not started but all requirements (after,before, wants, conflicts) works. So Service is not started but it is still disabling other services by Conflicts definition. If service is not started by Condtiion it is still conflicting

/etc/systemd/system/getty@tty1.service.d/noclear.conf

[Service]
TTYVTDisallocate=no

In udev rules.d add TAG+=“systemd”:

SUBSYSTEM=="input", KERNEL=="event[0-9]*", ENV{ID_INPUT_TOUCHSCREEN}=="1", ATTRS{name}=="edt-ft53*", \
    SYMLINK+="input/lcd_sandwich_touchscreen input/touchscreen_edt", \
    TAG+="systemd", \
    OPTIONS+="link_priority=100"

And now device is visible from systemd:

systemctl -l -a
  dev-input-event0.device                                                                        loaded    active   plugged   /dev/input/event0
  dev-input-lcd_sandwich_touchscreen.device                                                      loaded    active   plugged   /dev/input/lcd_sandwich_touchscreen

You can create refer to device, e.g. create .wants directory

/etc/systemd/system/ev-input-lcd_sandwich_touchscreen.device.wants

To start service if device is inserted:

mu.rule

..., TAG+="systemd", ENV{SYSTEMD_WANTS}="netctl-auto@mywifi.service"

After switching to systemd by command

apt-get install systemd-sysv

, new problem occurs: system boot stops and waits for password for encrypted partition (with user data - no needed to system start).

To prevent automount of crypted volumes, additional parameters 'noauto' and 'noearly' must be added to /etc/crypttab:

/etc/crypttab

/MyMountPoint		/dev/vg_MyName/lv_MyName	none	luks,noauto,noearly

System startup can hang forever if there is no link on ethernet cable:

[    **] A start job is running for LSB: Raise network interf...38s / no limit)

To set timeout, add file:

/etc/systemd/system/networking.service.d/reduce-timeout.conf

[Service]
TimeoutStartSec=15

systemd-sysv-generator[320]: stat() failed on /etc/init.d/traccar, ignoring: No such file or directory
systemd-sysv-generator[320]: stat() failed on /etc/init.d/ts3server, ignoring: No such file or directory
systemd-sysv-generator[320]: stat() failed on /etc/init.d/miio_server, ignoring: No such file or directory

Reason is that some init scripts are symlinked to files located in /opt directory, which is a different FS (in this case BTRFS subvolume), and it was not mounted when systemd-sysv-generator was started.

Solution: How to mount non-root filesystem before systemd-sysv-generator runs?

Under systemd-enabled systems, there's a new-with-systemd mount-option you can place in /etc/fstab — x-initrd.mount.
You also need to make sure that your filesystem's fs_passno is set to “0” …and if your filesystem lives on an LVM2 volume, you need to update your GRUB2 config to ensure that the LVM gets onlined prior to systemd invoking the systemd-sysv-generator utility