Skip to content

Latest commit

 

History

History
149 lines (115 loc) · 4.75 KB

README.md

File metadata and controls

149 lines (115 loc) · 4.75 KB

Kubernetes Support for NETINT VPU

About

The Kubernetes orchestration tool allows NETINT video transcoder devices to be managed as nodes. This section details the configuration and usage of NETINT video transcoder device with Kubernetes.

Prerequisites

This document assumes the user has Kubernetes already installed and is fimilar with the setup, configuration and management of Kubernetes.

Installations

There are two methods to install the device plugin: Install from Source or to Install using pre-built containers.

Install from Source

Complie the NETINT k8 device plugin. This step requires Go to be installed.

make build

Create Docker image for k8s device plugin.

make buildImage

Install Docker image using helm, the k8s device plugin needs privileged mode.

make deploy

Install using pre-built containers

Install Docker image using helm, the k8s device plugin needs privileged mode.

make deploy

Verify the Installation

After successfully deploy Docker image, add a tag of each K8s node to enable NETINT card using below command.

kubectl label nodes xxx netint-device=enable

Verify the device plugin has been successfully installed with the kubectl get pods -n kube-system command. You should find the netint-device-plugin-#####, pod running with ready 1/1 and status should be running.

$ kubectl get pods -n kube-system
NAME                               READY   STATUS    RESTARTS        AGE
coredns-787d4945fb-jmmxn           1/1     Running   1 (3d19h ago)   3d20h
etcd-minikube                      1/1     Running   1 (3d19h ago)   3d20h
kube-apiserver-minikube            1/1     Running   1 (3d19h ago)   3d20h
kube-controller-manager-minikube   1/1     Running   1 (3d19h ago)   3d20h
kube-proxy-xk44j                   1/1     Running   1 (3d19h ago)   3d20h
kube-scheduler-minikube            1/1     Running   1 (3d19h ago)   3d20h
netint-device-plugin-t46v2         1/1     Running   0               29s
netint-pod                         1/1     Running   0               29s
storage-provisioner                1/1     Running   4 (3d19h ago)   3d20h

You can check how many cards have been detected by the plugin with the kubectl describe nodes minikube command. T4xx devices will show up as node resource under netint.ca/ASIC and Quadra devices will show up as a node resource under netint.ca/Quadra.

Allocatable:
  cpu:                12
  ephemeral-storage:  240406076Ki
  hugepages-1Gi:      0
  hugepages-2Mi:      0
  memory:             16006452Ki
  netint.ca/ASIC:     0
  netint.ca/Quadra:   3
  pods:               110

You can view what resources are being used currently under the same command.

Allocated resources:
  (Total limits may be over 100 percent, i.e., overcommitted.)
  Resource           Requests      Limits
  --------           --------      ------
  cpu                1250m (10%)   1 (8%)
  memory             2218Mi (14%)  4266Mi (27%)
  ephemeral-storage  0 (0%)        0 (0%)
  hugepages-1Gi      0 (0%)        0 (0%)
  hugepages-2Mi      0 (0%)        0 (0%)
  netint.ca/ASIC     0             0
  netint.ca/Quadra   1             1
Events:              <none>

Configuration

Using a Different Docker image

Each pod needs to have a container that at minimum has libxcoder installed in order to use the NETINT VPU cards. Optionally FFmpeg or Gstreamer can be installed.

Important

The default container that pulled from Docker Hub only has support for the Quadra VPUs.

Change the value of the image image parameter to use the desired image.

  containers:
  - name: netint-pod
    image: netint/quadra_ubuntu-24.04_ffmpeg:latest

Limit VPUs

To limit the number of VPUs that are available in each pod, the value for netint.ca/ASIC for T4xx and the netint.ca/Quadra for Quadra can be adjusted to the desired amount.

resources:
  requests:
    memory: "128Mi"
    cpu: "500m"
    netint.ca/ASIC: 1
  limits:
    memory: "128Mi"
    cpu: "500m"
    netint.ca/ASIC: 1
  requests:
    memory: "128Mi"
    cpu: "500m"
    netint.ca/Quadra: 1
  limits:
    memory: "128Mi"
    cpu: "500m"
    netint.ca/Quadra: 1

Example

NETINT has provided an example yaml file in example/pod-with-netint.yaml that can be used as a reference design.