Introduction To Containerization On NVIDIA Base Command Manager

This page is for version 11 (default). For other versions, use one of these documentation indexes: - 11 (default): https://docs.nvidia.com/base-command-manager/product-manuals/11/llms.txt - 10: https://docs.nvidia.com/base-command-manager/product-manuals/10/llms.txt

Containerization is a technology that allows processes to be isolated by combining cgroups, Linux namespaces, and (container) images.

• Cgroups are introduced in section 7.10 on workload management of the Administrator Manual

• Linux namespaces represent independent spaces for different operating system facilities: process IDs, network interfaces, mount points, inter-process communication resources and others. Cgroups and namespaces allow processes to be isolated from each other by separating the available resources as much as possible.

• A container image is a component of a container, and is a file that contains one or several layers. The layers cannot be altered as far as the container is concerned, and a snapshot of the image can be used for other containers. A union file system is used to combine these layers into a single image. Union file systems allow files and directories of separate file systems to be transparently overlaid, forming a single coherent file system.

Cgroups, namespaces and images are the basis of a container. When the container is created, then a new process can be started within the container. Containerized processes running on a single machine all share the same operating system kernel, so they start immediately, without the delay of requiring a kernel to first boot up. No process is allowed to change the layers of the image. All changes are applied on a temporary layer created on top of the image, and these changes are destroyed when the container is removed.

There are several ways to manage the containers, but the most powerful approaches use Docker, also known as Docker Engine, and Kubernetes.

Docker manages containers on individual hosts, while Kubernetes manages containers across a cluster. BCM integrates both of these solutions, so that setup, configuration and monitoring of containers becomes an easily-managed part of BCM.

Chapter 2 describes how Docker integration with BCM works.

Chapter 3 covers how container registries are integrated.

Chapter 4 covers Kubernetes integration.

Chapter 5 covers Kubernetes application configuration and groups of Kubernetes applications.

Chapter 6 covers Kubernetes operators, which are a way to manage Kubernetes cluster applications.

Chapter 7 covers Kubernetes deployment on edge sites.

Chapter 8 describes the installation and usage of the NVIDIA Base Command Manager CAPI extension called BCM Kubernetes CAPI Infrastructure Provider. Kubernetes Cluster API (CAPI), is an API for managing Kubernetes clusters.