At present, when automatic CDI spec generation fails, the error is logged and execution continues handling the automatic CDI request as a regular CDI request. This leads to cryptic errors such as:
sudo docker run --rm --runtime=nvidia --gpus all ubuntu nvidia-smi
docker: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: could not apply required modification to OCI specification: error modifying OCI spec: failed to inject CDI devices: unresolvable CDI de vices runtime.nvidia.com/gpu=all: unknown
We should rather return the error from automatic spec generation directly to surface errors to the user.
At present, when automatic CDI spec generation fails, the error is logged and execution continues handling the automatic CDI request as a regular CDI request. This leads to cryptic errors such as:
We should rather return the error from automatic spec generation directly to surface errors to the user.