From ae5df6944ddda555b75351d7a35a3aef5100bf18 Mon Sep 17 00:00:00 2001 From: "Christopher M. Luciano" Date: Mon, 22 May 2017 12:58:55 -0400 Subject: [PATCH] Add GPU mountpath warning The GPU path is not permissive by default and requires a bit of additional setup if the operator does not allow for privileged containers. Related kubernetes/kubernetes#46007 Signed-off-by: Christopher M. Luciano --- docs/tasks/manage-gpus/scheduling-gpus.md | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/docs/tasks/manage-gpus/scheduling-gpus.md b/docs/tasks/manage-gpus/scheduling-gpus.md index aeb92a3e2b..f9105308dd 100644 --- a/docs/tasks/manage-gpus/scheduling-gpus.md +++ b/docs/tasks/manage-gpus/scheduling-gpus.md @@ -109,6 +109,10 @@ The API presented here **will change** in an upcoming release to better support As of now, CUDA libraries are expected to be pre-installed on the nodes. +The NVIDIA drivers will require privileged containers due to the permissions on ``/usr/lib/nvidia-367``. + +To mitigate this, you can copy the libraries to a more permissive folder in ``/var/lib/`` or change the permissions directly. (Future releases will automatically perform this operation) + Pods can access the libraries using `hostPath` volumes. ```yaml