nvidia-container-toolkit

mirror of https://github.com/NVIDIA/nvidia-container-toolkit synced 2024-11-21 15:57:49 +00:00

History

Evan Lezar 9c2476c98d Expose opt-in features in toolkit-container This change enables opt-in (off-by-default) features to be opted into. These features can be toggled by name by specifying the (repeated) --opt-in-feature command line argument or as a comma-separated list in the NVIDIA_CONTAINER_TOOLKIT_OPT_IN_FEATURES environment variable. Signed-off-by: Evan Lezar <elezar@nvidia.com>		2024-09-18 22:30:27 +02:00
..
containerd	Extract options from default runtime if runc does not exist	2024-07-04 12:44:08 +02:00
crio	run goimports -local against the entire codebase	2023-12-01 11:13:17 +01:00
docker	run goimports -local against the entire codebase	2023-12-01 11:13:17 +01:00
nvidia-toolkit	Allow toolkit.pid path to be specified	2024-06-17 11:26:23 +02:00
operator	avoid using map pointers as maps are always passed by reference	2024-07-02 17:35:44 -07:00
toolkit	Expose opt-in features in toolkit-container	2024-09-18 22:30:27 +02:00
container.go	Add support for adding additional containerd configs	2024-05-17 12:58:08 +02:00
README.md	Remove NVIDIA experimental runtime from toolkit container	2023-07-10 11:44:55 +02:00

README.md

Introduction

This repository contains tools that allow docker, containerd, or cri-o to be configured to use the NVIDIA Container Toolkit.

Note: These were copied from the container-config repository are being migrated to commands installed with the NVIDIA Container Toolkit.

These will be migrated into an upcoming nvidia-ctk CLI as required.

Docker

After building the docker binary, run:

docker setup \
    --runtime-name NAME \
        /run/nvidia/toolkit

Configure the nvidia-container-runtime as a docker runtime named NAME. If the --runtime-name flag is not specified, this runtime would be called nvidia.

Since --set-as-default is enabled by default, the specified runtime name will also be set as the default docker runtime. This can be disabled by explicityly specifying --set-as-default=false.

The following table describes the behaviour for different --runtime-name and --set-as-default flag combinations.

Flags	Installed Runtimes	Default Runtime
NONE SPECIFIED	`nvidia`	`nvidia`
`--runtime-name nvidia`	`nvidia`	`nvidia`
`--runtime-name NAME`	`NAME`	`NAME`
`--set-as-default`	`nvidia`	`nvidia`
`--set-as-default --runtime-name nvidia`	`nvidia`	`nvidia`
`--set-as-default --runtime-name NAME`	`NAME`	`NAME`
`--set-as-default=false`	`nvidia`	NOT SET
`--set-as-default=false --runtime-name NAME`	`NAME`	NOT SET
`--set-as-default=false --runtime-name nvidia`	`nvidia`	NOT SET

These combinations also hold for the environment variables that map to the command line flags: DOCKER_RUNTIME_NAME, DOCKER_SET_AS_DEFAULT.

Containerd

After running the containerd binary, run:

containerd setup \
    --runtime-class NAME \
        /run/nvidia/toolkit

Configure the nvidia-container-runtime as a runtime class named NAME. If the --runtime-class flag is not specified, this runtime would be called nvidia.

Adding the --set-as-default flag as follows:

containerd setup \
    --runtime-class NAME \
    --set-as-default \
        /run/nvidia/toolkit

will set the runtime class NAME (or nvidia if not specified) as the default runtime class.

The following table describes the behaviour for different --runtime-class and --set-as-default flag combinations.

Flags	Installed Runtime Classes	Default Runtime Class
NONE SPECIFIED	`nvidia`	NOT SET
`--runtime-class NAME`	`NAME`	NOT SET
`--runtime-class nvidia`	`nvidia`	NOT SET
`--set-as-default`	`nvidia`	`nvidia`
`--set-as-default --runtime-class NAME`	`NAME`	`NAME`
`--set-as-default --runtime-class nvidia`	`nvidia`	`nvidia`

These combinations also hold for the environment variables that map to the command line flags.