Isaac ROS dev container: GLFW driver error, CUVSLAM tracker error

Robotics & Edge Computing Robotics - Isaac Isaac ROS
1

I tried to run the “isaac_ros_visula_slam” following isaac_ros_visual_slam using ros dev container

Firstly, when i tried to run the examples this error popped up GLFW Driver Error: GLX: Failed to create context: BadValue (integer parameter out of range for operation)

I tried to build by own container, when i realised “libcudnn9-cuda” in dockerfile needs to be “libcudnn9-cuda-12” for the image to build properly and after that, this error started popping up on the ros dev container [ERROR] [component_container_mt-1]: process has died [pid 2542, exit code -11, cmd ‘/opt/ros/humble/lib/rclcpp_components/component_container_mt --ros-args -r __node:=container -r __ns:=/isaac_ros_examples’].

Now on my custom container where i installed all the dependencies one by one the node launched but it constantly gives me tracker error even for all the given examples in the documentation. [component_container-1] [ERROR] [1735217102.869027840] [visual_slam_node]: Failed to initialize CUVSLAM tracker: 4

It is very hard to debug what is going wrong. I have tried the rosbag example, the realsense example and the isaac sim example. I have a intel realsense camera aswell.

Tried to install and use v3.1 of issac_ros_visual_slam with ros_dev container built again and that gives me this error: Failed to load node ‘visual_slam_node’ of type ‘nvidia::isaac_ros::visual_slam::VisualSlamNode’ in container ‘/visual_slam_launch_container’: Component constructor threw an exception: loadExtensions Error: GXF_CONTEXT_INVALID

My system has RTX A4000 gpu with all the cuda libs working, I think all the pytorch and cuda libs and paths are properly set in the docker container aswell. I was also able to install the all the rosdep dependencies for the isaac_ros_visual_slam package and colcon build it still the same results.

2

I have also encountered the same problem. Have you solved it？

3

I havn’t solved it yet but i tried a bunch of things. i double checked and reinstalled the cuda and cudnn libs and tested other examples:

Examples working:

Examples not working:

object detection give me this error:
[component_container_mt-1] 2024-12-30 12:07:49.783 ERROR gxf/std/extension_loader.cpp@114: Failed to load extension /opt/ros/humble/share/gxf_isaac_triton/gxf/lib/libgxf_isaac_triton.so Error: libdcgm.so.3: cannot open shared object file: No such file or directory

Now i think it might be still some nvidia driver issue but not sure as the older version of vslam also didn’t work. If you dont mind can you share few details about your setup?

my system has
Cuda Version (in container): 12.6.85 (nvcc --version)
Cudnn (container): 9.6 (cat /usr/include/cudnn_version.h | grep CUDNN_MAJOR -A 2)
nvidia driver (host system): 535.183.01 (nvidia-smi)

I am going through ros logs to get more info on the issue.
logs.txt (78.3 KB)

4

Just updated my host PC to nvidia driver 560.35.03 and reinstalled everything and tried the docker dev container again but still getting the same error of

[ERROR] [component_container_mt-1]: process has died [pid 1249, exit code -11, cmd ‘/opt/ros/humble/lib/rclcpp_components/component_container_mt --ros-args -r __node:=container -r __ns:=/isaac_ros_examples’].