Closed peterbandi closed 1 year ago
There is already an existing ticket for this issue: https://github.com/docker/for-win/issues/13324
Closed issues are locked after 30 days of inactivity. This helps our team focus on active issues.
If you have found a problem that seems similar to this, please open a new issue.
/lifecycle locked
Actual behavior
Upgrading to Docker Desktop 4.17.1 (101757) breaks the nVidia GPU support on WSL2. After the upgrade from 4.17.0 to 4.17.1 starting a docker container with
--gpus=all
option makes the docker container hang. It is not possible to interact with it or exit it. The container is now shown in Docker Desktop or withdocker container list
ordocker ps
commands.The symptoms are the same if the container is started with docker compose. It sticks in a state where it is not listed with the aforementioned docker command line commands. Docker Dekstop shows the container with orange color in the swarm in "Created" state even after the swarm is shut down. It is not possible to shut it down or delete it without restarting Docker Desktop.
Expected behavior
Being able to run docker containers with attached GPUs (
--gpus
option) as before.Information
Output of
& "C:\Program Files\Docker\Docker\resources\com.docker.diagnose.exe" check
Steps to reproduce the behavior
docker pull nvidia/cuda:11.8.0-base-ubuntu20.04
docker run --interactive --tty --rm nvidia/cuda:11.8.0-base-ubuntu20.04 echo "Hello"
Hello
and exitsdocker run --interactive --tty --rm --gpus=all nvidia/cuda:11.8.0-base-ubuntu20.04 echo "Hello"
docker container list
command