Start jetson-inference container on boot up

[konstantinos.kalogirou]

Dear all,
I would like to run the jetson-inference container when my jetson nano is started(on boot).
So, I created a bash shell script which contains:

gnome-terminal --window -t "jetson session" --working-directory="/home/$username/jetson-inference" -- bash -c '{ echo "Session started successfully..." || echo "Session failed to start.See log file.";docker/run.sh -v /root/scripts:/custom_scripts > '${log_custom_scripts}' ;}  '

where

log_custom_scripts = "log_custom_scripts.txt" 
username = localuser

Then, I created myservice.service as following:

[Unit]
Description=Execute this on boot
After=multi-user.target

[Service]
Environment=XAUTHORITY=/home/localuser/.Xauthority   # Get the stored credentials of localuser to be authenticated
Environment=DISPLAY=:0           # This needs to enable GUI X session, right?
ExecStartPre=/bin/sleep 20       # Delay it just in case
ExecStart=/root/scripts/myscript.sh  

[Install]
WantedBy=multi-user.target

and I put it into the /etc/systemd/system.

I tested the service before enabling to start on boot:
sudo systemctl start myservice.service

and get the status in another terminal by
sudo systemctl status myservice.service

Unfortunately, I received:
the input device is not a TTY

I tried to start the container without the gnome-terminal but without success. More specific,
execute it in background

cd /home/$username/jetson-inference 
docker/run.sh -v /root/scripts:/custom_scripts > $log_custom_scripts &

or in foreground

cd /home/$username/jetson-inference 
docker/run.sh -v /root/scripts:/custom_scripts > $log_custom_scripts

I also tried

startx docker/run.sh -v /root/scripts:/custom_scripts > $log_custom_scripts

but I received

(==) Log file: "/var/log/Xorg.1.log", Time: Thu Apr 18 14:47:05 2024
(==) Using config file: "/etc/X11/xorg.conf"
(==) Using system config directory "/usr/share/X11/xorg.conf.d"
/usr/bin/xterm: No absolute path found for shell: docker/run.sh
xinit: connection to X server lost

waiting for X server to shut down (II) Server terminated successfully (0). Closing log file.

For your information, I have followed these forum discussions:

How to run a script on boot

That is the Canonical (haha) way of doing it now. To start, create a limited user to run your service as by running the following commands: sudo addgroup --system robot sudo adduser --system --no-create-home --disabled-login --disabled-password --ingroup robot robot if your script user needs to be in extra groups you can add them with adduser robot GROUP …where GROUP is the group you want to add the user to (video, for example, to access the GPU). You can see the groups a user such as you…

Run python script (display script) at boot

Hi, You need to add a file to /etc/systemd/system. For more details, please check following topic: Thanks.

Startup/boot script with systemd issues

I’m trying run a TensorRT object detector at boot using the systemd method but I keep running into errors. My Jetson Nano 2GB is configured to auto-login to LXDE so I don’t have to input the password. Build info. Jetpack 4.4.1 head -n 1 /etc/nv_tegra_release #R32 (release), REVISION: 4.4, GCID: 23942405, BOARD: t210ref, EABI: aarch64, DATE: Fri Oct 16 19:44:43 UTC 2020 Here are the steps I took. Created startup_detection_script.py import os os.system("cd /home/ag72/tensorrt_demos && pyt…

Start my jetson-inference application at startup

Hi thanks for the reply. I added the export DISPLAY=:0 before launching the jetson-inference program but it didnt work. I have added some more logging. I get the following error in the log [OpenGL] failed to open X11 server connection.[OpenGL] failed to create X11 Window. I have a main service running as a systemd service and then spawn a new process, and then I get this error. All work great if I just run the application from a shell.

I understand that jeston-inference container needs a GUI X session.
Is it possible to start the container in a GUI session?

I would really appreciate any help.

Regards,

[dusty_nv]

@konstantinos.kalogirou

Unfortunately, I received:
the input device is not a TTY

Hi @konstantinos.kalogirou, I haven’t done this myself so YMMV, but you could try removing the -it (interactive terminal) part of the docker run command that jetson-inference/docker/run.sh makes

@konstantinos.kalogirou

I understand that jeston-inference container needs a GUI X session.

It doesn’t need display or X11, but if one is connected it will try to create a window by default (but should fail gracefully if it can’t)

[konstantinos.kalogirou]

Hi.

I removed the interactive terminal (-it) in the run.sh but without any success.
The docker starts and stops after a second. Specifically, I tried the followings:

cd /home/$localuser/jetson-inference && exec docker/run_play.sh -v /root/scripts:/custom_script

cd /home/$localuser/jetson-inference 
docker/run_play.sh -v /root/scripts:/custom_scripts &

cd /home/$localuser/jetson-inference 
docker/run_play.sh -v /root/scripts:/custom_scripts 

gnome-terminal --window -t "Monitoring session" --working-directory="/home/$localuser/jetson-inference" -- bash -c '{ echo "Session started successfully..." || echo "Session failed to start.See log file.";docker/run_play.sh -v /root/scripts:/custom_scripts > '${log_custom_scripts}' ;}  '

where run_play.sh is the run.sh without the -it.

I would like to run my shell script not from the /home/localuser/jetson-inference.

Regards,

[dusty_nv]

Recommend that you try just using docker run directly or try running another basic container just for testing to figure out your start-up procedure.

My guess is that it is not actually running any command in the container, and without -it there is no interactive terminal attached, so it just exits.

[konstantinos.kalogirou]

Hello again.
I tried:

1. Run as root user , create a new container to host it:

root@kalogirouk-desktop:~/scripts# docker run --name test_container -v /root/scripts:/home/kalogirouk/jetson-inference/ bash -c "docker/run.sh -v root/scripts"

**bash: line 1: docker/run.sh: No such file or directory**

2. Run as root user or as local user (the same result)

kalogirouk@kalogirouk-desktop:~$ /home/$username/jetson-inference/docker/run.sh --volume /root/scripts:/custom_scripts

/home/kalogirouk/jetson-inference/docker/run.sh: line 156: docker/tag.sh: No such file or directory
localuser:root being added to access control list
DATA_VOLUME:      --volume /home/kalogirouk/data:/jetson-inference/data --volume /home/kalogirouk/python/training/classification/data:/jetson-inference/python/training/classification/data --volume /home/kalogirouk/python/training/classification/models:/jetson-inference/python/training/classification/models --volume /home/kalogirouk/python/training/detection/ssd/data:/jetson-inference/python/training/detection/ssd/data --volume /home/kalogirouk/python/training/detection/ssd/models:/jetson-inference/python/training/detection/ssd/models --volume /home/kalogirouk/python/www/recognizer/data:/jetson-inference/python/www/recognizer/data 
USER_VOLUME:      --volume /root/scripts:/custom_scripts 
V4L2_DEVICES:     --device /dev/video0 
DISPLAY_DEVICE:   -e DISPLAY=:0 -v /tmp/.X11-unix/:/tmp/.X11-unix

2. Run as root or as local user,the same result either run.sh contains -it or it does not contain -it

root@kalogirouk-desktop:~/scripts# /home/kalogirouk/jetson-inference/docker/run.sh -v /root/scripts:/custom_scripts
/home/kalogirouk/jetson-inference/docker/run.sh: line 156: docker/tag.sh:No such file or directory
kalogirouk:root being added to access control list
DATA_VOLUME:      --volume /root/scripts/data:/jetson-inference/data --volume /root/scripts/python/training/classification/data:/jetson-inference/python/training/classification/data --volume /root/scripts/python/training/classification/models:/jetson-inference/python/training/classification/models --volume /root/scripts/python/training/detection/ssd/data:/jetson-inference/python/training/detection/ssd/data --volume /root/scripts/python/training/detection/ssd/models:/jetson-inference/python/training/detection/ssd/models --volume /root/scripts/python/www/recognizer/data:/jetson-inference/python/www/recognizer/data 
USER_VOLUME:      --volume /root/scripts:/custom_scripts 
V4L2_DEVICES:     --device /dev/video0 
DISPLAY_DEVICE:   -e DISPLAY=:0 -v /tmp/.X11-unix/:/tmp/.X11-unix 
**/home/kalogirouk/jetson-inference/docker/run_play.sh: line 198: [: =: unary operator expected**
**/home/kalogirouk/jetson-inference/docker/run_play.sh: line 213: [: =: unary operator expected**

I modified the lines 198 and 213 to
[[ $ARCH = “aarch64” ]] and
[[ $ARCH = “x86_64” ]] respectively.

I received the same error without the unary operators:

root@kalogirouk-desktop:~/scripts# /home/kalogirouk/jetson-inference/docker/run.sh -v /root/scripts:/custom_scripts
/home/kalogirouk/jetson-inference/docker/run.sh: line 156: docker/tag.sh: No such file or directory
kalogirouk:root being added to access control list
DATA_VOLUME:      --volume /root/scripts/data:/jetson-inference/data --volume /root/scripts/python/training/classification/data:/jetson-inference/python/training/classification/data --volume /root/scripts/python/training/classification/models:/jetson-inference/python/training/classification/models --volume /root/scripts/python/training/detection/ssd/data:/jetson-inference/python/training/detection/ssd/data --volume /root/scripts/python/training/detection/ssd/models:/jetson-inference/python/training/detection/ssd/models --volume /root/scripts/python/www/recognizer/data:/jetson-inference/python/www/recognizer/data 
USER_VOLUME:      --volume /root/scripts:/custom_scripts 
V4L2_DEVICES:     --device /dev/video0 
DISPLAY_DEVICE:   -e DISPLAY=:0 -v /tmp/.X11-unix/:/tmp/.X11-unix 

Regards,

[dusty_nv]

@konstantinos.kalogirou see if you can get a different container to startup a boot like this as a test, then just take the docker run command that the jetson-inference script produces and substitute that instead. I don’t believe those scripts are independent of working directory, but the tools in jetson-containers are (which is what I use for building jetson-inference docker now)

[konstantinos.kalogirou]

I did some tests and I can start up a container.

Can you pleaes make more clear how to take the “docker run” command that the jetson inferenece produces?

[dusty_nv]

Hi @konstantinos.kalogirou, you can find it here:

dusty-nv/jetson-inference/blob/644f5158a55a7f313f4d43b57ad3ed109fcbf9ab/docker/run.sh#L203

You can also add set +x to this script, and it will print out the exact commands as it runs those, and then you can just take them and run them yourself as desired.

[konstantinos.kalogirou]

After many tests, I realised that any docker container needs a TTY cannot start up on boot.

When I run my script service in the /etc/systed/system (systemctl enable my_script.service, systemctl start)

my_script.service), I receive the following reasonable messages:
my_script.sh[10426]: the input device is not a TTY
systemd[1]: my_script.service: Main process exited, code=exited, status=1/FAILU
systemd[1]: my_script.service: Failed with result 'exit-code'.

Here is the my_script.service:

[Unit]
Description=Execute this on boot
After=multi-user.target

[Service]
Environment=XAUTHORITY=/home/kalogirouk/.Xauthority
Environment=DISPLAY=:0
ExecStartPre=/bin/sleep 20
ExecStart=/root/scripts/my_script.sh 
StandardOutput=journal+console

[Install]
WantedBy=multi-user.target

Unfortuantely, the other docker container I used to run does not need TTY(the -it options were not considered at all). So, my previous answer to you is not correct.

Futermore, I tested running docker in detach mode (-d parameter) but with the same result.

I do not know if it is feasible to start a docker container that needs a TTY in general in Ubuntu OS boot up but I will focus to start the jetson-infernece container in a more indirect way.

[dusty_nv]

jetson-inference container doesn’t need TTY, it just starts it that way by default for ease-of-use for users during development. You should just be able to remove -it from the docker run command and it won’t use TTY anymore.