Issue with EGL Display in Docker on Jetson Nano in Headless Mode

[s.salehi0026]

Hello,

I have a custom Docker image that I built using the base image from DeepStream L4T. The image includes all the necessary dependencies to run my Velocity code.

When my Jetson Nano is connected to a monitor, everything works perfectly. However, when I connect to the system via SSH without a monitor attached, the program fails to run.

Additionally, I am using the following GStreamer pipeline in my Docker image:

gst-launch-1.0 pylonsrc name=pylon-source imageformat="RGB8" pfs-location=PFS/cur.pfs ! videoconvert ! nvvideoconvert ! capsfilter caps="video/x-raw(memory:NVMM),format=(string)NV12" ! nvmuxer.sink_0 nvstreammux name=nvmuxer gpu-id=0 batch-size=1 nvbuf-memory-type=0 live-source=1 width=1920 height=1200 ! nvinfer name=pgie interval=0 config-file-path=/opt/Models/grill/Grill_config.txt unique-id=1 ! queue ! nvinfer config-file-path=/opt/Models/plate/Plate_config.txt process-mode=2 unique-id=2 infer-on-gie-id=1 ! queue ! nvvideoconvert name=elementWithOCRProb ! queue ! nvdsosd process-mode=HW_MODE ! fakesink

This pipeline works flawlessly when a monitor is connected. However, when running in headless mode (no monitor connected), I encounter the following error:

before parse launchNo EGL Display
nvbufsurftransform: Could not get EGL display connection

Could you please advise on what might be causing this issue and how I can resolve it?

Thank you for your assistance!

[DaneLLL]

Hi,
The docker image of DeepStream has settings for display, since it is expected to have display output. You may need to remove the settings about DISPLAY and X11.

[s.salehi0026]

I have disabled X11 Forwarding in MobaXterm by unchecking the option, and I also created a new Docker image using the following Dockerfile:

FROM deepstream-pro_2

ENV DISPLAY=

WORKDIR /home/nano/Desktop/executable

CMD ["./Velocity"]

However, the issue persists. Additionally, I am running a script (app_runner.sh) which includes the following commands:

#!/bin/bash
xhost + > /home/nano/output.txt

CONTAINER_NAME=my-c

IMAGE_NAME=deepstream-pro_2_dispayoff

while true; do
    sleep 10
    # Check if the container is running
    if ! sudo docker ps --format '{{.Names}}' | grep -q "^$CONTAINER_NAME$"; then
        echo "Starting Docker container: $CONTAINER_NAME"
        sudo docker run -it --rm --net=host --runtime nvidia --privileged \
        -v /home/nano/Desktop/executable:/home/nano/Desktop/executable \
        -v /opt/Models:/opt/Models -v /etc/localtime:/etc/localtime:ro \
        -v /etc/timezone:/etc/timezone:ro -d --name $CONTAINER_NAME $IMAGE_NAME
    else
        echo "Docker container $CONTAINER_NAME is already running"
    fi
    # Sleep for a while before checking again
    sleep 10
done

Despite these changes, the problem remains unresolved. Could there be another setting or step that I might have missed?

Thanks for your help!

[s.salehi0026]

I also tried another solution by building the Docker image with the following Dockerfile:

FROM deepstream-pro_2
RUN apt-get update && apt-get install -y \
    x11-apps \
    x11-xserver-utils \
    xauth \
    xfonts-base

ENV DISPLAY=:0

WORKDIR /home/nano/Desktop/executable

CMD ["./Velocity"]

I ran the Docker container using the following command:

sudo docker run -it --rm --net=host --runtime nvidia --privileged \
-e DISPLAY=$DISPLAY \
-v /tmp/.X11-unix:/tmp/.X11-unix \
-v /home/nano/Desktop/executable:/home/nano/Desktop/executable \
-v /opt/Models:/opt/Models -v /etc/localtime:/etc/localtime:ro \
-v /etc/timezone:/etc/timezone:ro --name my-c my-deepstream-image-x11

However, this did not resolve the issue either.

Any additional suggestions would be greatly appreciated.

[DaneLLL]

Hi,
Please try other console tool instead of MobaXterm. It seems like it initializes its own display environment and may not work properly. Would suggest use pure-command-line console.