Why does Podman fail to run the Computer Vision server image, podman run -p 8501:5000 --hooks-dir=/usr/share/containers/oci/hooks.d/ \--security-opt=label=disable --device=nvidia.com/gpu=all controls_detection eula=accept, with “Error: payload does not match any of the supported formats (oci, oci-archive, dir, docker-archive)"?
Issue Description
When following the minimum software requirements and installing CV server using Podman on Ubuntu 2022.04, running the CV Server image, podman run -p 8501:5000 --hooks-dir=/usr/share/containers/oci/hooks.d/ \--security-opt=label=disable --device=nvidia.com/gpu=all controls_detection eula=accept, will fail with:
"Error: payload does not match any of the supported formats (oci, oci-archive, dir, docker-archive)."
This is caused by running containers accessing NVIDIA GPUs using Container Device Interface (CDI) requires Podman 4.0.1 or later, which is not available on Ubuntu 22.04.
Resolution
Upgrade Ubuntu to 24.04 or later, which will install a compatible version of Podman, such as version 4.9.3, which will load the image as expected.
In order to run containers, such as the CV Server, using NVIDIA GPUS using CDI, Podman 4.01.+ is required. This version of Podman is not accessible on Ubuntu 22.04 but requires a later version, such as Ubuntu 24.04.
Note: If controls_detection.tar fails when loading the server image, podman load -I controls_detection.tar, with the same error, ‘Error: payload does not match any of the supported formats (oci, oci-archive, dir, docker-archive),” follow the guidance in Computer Vision Image Load Failure in Podman Due To Insufficient Space.