Running NICE DCV inside a Singularity Container

[rsinghal168]

I need a definition file using any ubuntu or RHEL version of linux, successfully running nice dcv inside a singularity container. I have been stuck on this problem for a while, would really appreciate some help.
Here is my def file
`
Bootstrap: docker
From: ubuntu:22.04

%post
export DEBIAN_FRONTEND=noninteractive

# 1. Base System Setup
apt-get update -y
apt-get install -y wget sudo software-properties-common

# Enhanced GUI dependencies
apt-get install -y \
    libgtk-3-0 \
    libnss3 \
    libxcomposite1 \
    libxrandr2 \
    libatk1.0-0 \
    libxkbcommon-x11-0 \
    libglib2.0-0 \
    libgstreamer1.0-0 \
    libgstreamer-plugins-base1.0-0 \
    libxv1 \
    libxdamage1 \
    libxfixes3 \
    dbus-x11 \
    x11-xserver-utils \
    xauth \
    pulseaudio \
    libpulse0 \
    libxtst6 \
    libglvnd0

# 2. Install Desktop Environment
apt-get install -y ubuntu-desktop gdm3 mesa-utils

# Create required directories
mkdir -p /run/dbus
mkdir -p /var/run/dbus
mkdir -p /tmp/.X11-unix
chmod 1777 /tmp/.X11-unix

# 3. Critical Kernel Components
apt-get install -y linux-aws linux-headers-$(uname -r) build-essential
apt-get install -y awscli

# 4. AWS NVIDIA GRID Driver Setup
aws s3 cp --recursive s3://ec2-linux-nvidia-drivers/latest/ .
chmod +x NVIDIA-Linux-x86_64*.run

# 5. Blacklist Conflicts
cat << EOF > /etc/modprobe.d/blacklist.conf

blacklist nouveau
blacklist lbm-nouveau
options nouveau modeset=0
alias nouveau off
alias lbm-nouveau off
EOF

# 6. Install NVIDIA Driver with additional options
./NVIDIA-Linux-x86_64*.run \
    --silent \
    --no-kernel-module \
    --no-drm \
    --disable-nouveau \
    --no-check-for-alternate-installs \
    --install-compat32-libs

# 7. NVIDIA Configuration
echo "options nvidia NVreg_EnableGpuFirmware=0" > /etc/modprobe.d/nvidia.conf
nvidia-xconfig --preserve-busid --enable-all-gpus --allow-empty-initial-configuration

# Enhanced X11 configuration
cat << EOF > /etc/X11/xorg.conf.d/10-nvidia.conf

Section "ServerLayout"
Identifier "Layout0"
Screen 0 "Screen0"
EndSection

Section "Device"
Identifier "Device0"
Driver "nvidia"
VendorName "NVIDIA Corporation"
BusID "PCI:0:7:0"
EndSection

Section "Screen"
Identifier "Screen0"
Device "Device0"
DefaultDepth 24
Option "AllowEmptyInitialConfiguration" "True"
SubSection "Display"
Depth 24
EndSubSection
EndSection
EOF

# 8. DCV Installation
wget https://d1uj6qtbmh3dt5.cloudfront.net/2024.0/Servers/nice-dcv-2024.0-18131-ubuntu2204-x86_64.tgz
tar -xzf nice-dcv-*.tgz
cd nice-dcv-*/ && apt install -y ./*.deb

# Configure DCV
mkdir -p /etc/dcv
cat << EOF > /etc/dcv/dcv.conf

[display]
enable-gpu-sharing=true

[connectivity]
web-port=8443
enable-quic-frontend=true

[security]
authentication="none"
EOF

# Cleanup
apt-get clean
rm -rf /var/lib/apt/lists/*

%environment
export DISPLAY=:0
export XAUTHORITY=/tmp/.Xauthority
export XDG_RUNTIME_DIR=/tmp/dcv-runtime
export PULSE_SERVER=unix:/tmp/pulse-native
export LD_LIBRARY_PATH="/usr/lib/x86_64-linux-gnu:/usr/lib64:/usr/lib/nvidia:$LD_LIBRARY_PATH"
export NVIDIA_VISIBLE_DEVICES=all
export NVIDIA_DRIVER_CAPABILITIES=compute,utility,graphics,video
`

Here is my singularity run command

sudo singularity run --nv \ --bind /home:/home \ --bind /etc/pam.d:/etc/pam.d \ --bind /etc/passwd:/etc/passwd \ --bind /etc/shadow:/etc/shadow \ --bind /etc/gshadow:/etc/gshadow \ --bind /etc/group:/etc/group \ --bind /tmp/.X11-unix:/tmp/.X11-unix \ --bind /run:/run \ --bind /var/run/dbus:/var/run/dbus \ --bind /sys/fs/cgroup:/sys/fs/cgroup \ --bind /dev/dri:/dev/dri \ --bind $HOME/.Xauthority:/tmp/.Xauthority \ --env DISPLAY=:0 \ dcv.sif

in the host machine I am running this command to start XORG

Xorg :0 & export DISPLAY=:0

then inside the container I am running this command

dcvserver --create-session test --owner=ubuntu

I am able to login to the dcv portal, but after the login I get a black screen with 'x' cursor.
This issue with the X server is really persistent.
If anyone has successfully been able to achieve this. Please help!!