mjhjmtu

I have two services :

First starts X, it is /etc/systemd/system/startx@.service

[Unit]
Description=startx automatique pour l'utilisateur %I
After=graphical.target systemd-user-sessions.service

[Service]
User=%I
Environment=XAUTHORITY=/tmp/.Xauthority_%I
PAMName=login
Type=simple
ExecStart=/bin/bash -l -c startx

[Install]
WantedBy=graphical.target

Second one, app.service requires X to be running as it uses display :0.0

[Unit]
Description=mY service
Requires= serviceA serviceB
After= serviceA serviceB

[Service]
Environment=DISPLAY=:0.0
User=USER_NAME
ExecStart=/bin/bash -c "my_app params ..."
Restart=always

[Install]
WantedBy=multi-user.target

When enabling both services

systemctl enable startx@remi.service
systemctl enable my_app.service

X server is started, my_app is started, its windows are displayed.

After, in order to prevent my_app to be started when X server not yet started
I added :

Requires=serviceA serviceB startx@remi.service
After=serviceA serviceB startx@remi.service

to app.service.

I disabled units, rebooted and tested modif with :

systemctl start my_app.service

X server is started, my_app is started, its windows are displayed.

Then I enabled app.service :
on next boot, X is started but my_app is not started. After some tests, I can see X is often started (not always), serviceA, serviceB are always started and my_app is never started.

Here is critical chain for startx :

systemd-analyze critical-chain startx@remi.service 
The time after the unit is active or started is printed after the "@" character. 
The time the unit takes to start is printed after the "+" character. 

startx@remi.service @5.296s 
└─graphical.target @5.293s 
 └─multi-user.target @5.292s 
 └─getty.target @5.292s 
 └─getty@tty1.service @5.267s 
 └─systemd-user-sessions.service @5.221s +36ms 
 └─network.target @5.214s 
 └─netctl-auto@wlan1.service @4.195s +1.013s 
 └─basic.target @4.096s 
 └─sockets.target @4.091s 
 └─dbus.socket @4.086s 
 └─sysinit.target @4.067s 
 └─haveged.service @3.982s +74ms 
 └─systemd-tmpfiles-setup.service @3.854s +96ms 
 └─systemd-journal-flush.service @3.557s +284ms 
 └─var-log.mount @3.474s +64ms 
 └─local-fs-pre.target @3.436s 
 └─lvm2-monitor.service @1.214s +2.098s 
 └─lvm2-lvmetad.service @1.434s
 └─systemd-journald.socket @1.173s
 └─-.mount @649ms
 └─system.slice @649ms
 └─-.slice @649ms

And critical chain for my_app :

systemd-analyze critical-chain detect_smile.service
The time after the unit is active or started is printed after the "@" character.
The time the unit takes to start is printed after the "+" character.

└─startx@remi.service @5.296s
 └─graphical.target @5.293s
 └─multi-user.target @5.292s
 └─getty.target @5.292s
 └─getty@tty1.service @5.267s
 └─systemd-user-sessions.service @5.221s +36ms
 └─network.target @5.214s
 └─netctl-auto@wlan1.service @4.195s +1.013s
 └─basic.target @4.096s
 └─sockets.target @4.091s
 └─dbus.socket @4.086s
 └─sysinit.target @4.067s
 └─haveged.service @3.982s +74ms
 └─systemd-tmpfiles-setup.service @3.854s +96ms
 └─systemd-journal-flush.service @3.557s +284ms
 └─var-log.mount @3.474s +64ms
 └─local-fs-pre.target @3.436s
 └─lvm2-monitor.service @1.214s +2.098s
 └─lvm2-lvmetad.service @1.434s
 └─systemd-journald.socket @1.173s
 └─-.mount @649ms
 └─system.slice @649ms
 └─-.slice @649ms

As you can see, my_app service is not shown above startx@remi.service

Why?