Skip to content

Latest commit

 

History

History
175 lines (134 loc) · 6.65 KB

File metadata and controls

175 lines (134 loc) · 6.65 KB

Intel Device Plugins Operator

Table of Contents

Introduction

Intel Device Plugins Operator is a Kubernetes custom controller whose goal is to serve the installation and lifecycle management of Intel device plugins for Kubernetes. It provides a single point of control for GPU, QAT, SGX, FPGA, DSA and DLB devices to a cluster administrators.

Installation

Install NFD (if it's not already installed) and node labelling rules (requires NFD v0.10+):

# either with default NFD installation
$ kubectl apply -k https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/nfd?ref=<RELEASE_VERSION>
# or when setting up with SGX
$ kubectl apply -k https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/nfd/overlays/sgx?ref=<RELEASE_VERSION>
# and finally, NodeFeatureRules
$ kubectl apply -k https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/nfd/overlays/node-feature-rules?ref=<RELEASE_VERSION>

Make sure both NFD master and worker pods are running:

$ kubectl get pods -n node-feature-discovery
NAME                          READY   STATUS    RESTARTS   AGE
nfd-master-599c58dffc-9wql4   1/1     Running   0          25h
nfd-worker-qqq4h              1/1     Running   0          25h

Note that labelling is not performed immediately. Give NFD 1 minute to pick up the rules and label nodes.

As a result all found devices should have correspondent labels, e.g. for Intel DLB devices the label is intel.feature.node.kubernetes.io/dlb:

$ kubectl get no -o json | jq .items[].metadata.labels |grep intel.feature.node.kubernetes.io/dlb
  "intel.feature.node.kubernetes.io/dlb": "true",

Full list of labels can be found in the deployments/operator/samples directory:

$ grep -r feature.node.kubernetes.io/ deployments/operator/samples/
deployments/operator/samples/deviceplugin_v1_dlbdeviceplugin.yaml:    intel.feature.node.kubernetes.io/dlb: 'true'
deployments/operator/samples/deviceplugin_v1_qatdeviceplugin.yaml:    intel.feature.node.kubernetes.io/qat: 'true'
deployments/operator/samples/deviceplugin_v1_sgxdeviceplugin.yaml:    intel.feature.node.kubernetes.io/sgx: 'true'
deployments/operator/samples/deviceplugin_v1_gpudeviceplugin.yaml:    intel.feature.node.kubernetes.io/gpu: "true"
deployments/operator/samples/deviceplugin_v1_fpgadeviceplugin.yaml:    intel.feature.node.kubernetes.io/fpga-arria10: 'true'
deployments/operator/samples/deviceplugin_v1_dsadeviceplugin.yaml:    intel.feature.node.kubernetes.io/dsa: 'true'

The default operator deployment depends on cert-manager running in the cluster. See installation instructions here.

Make sure all the pods in the cert-manager namespace are up and running:

$ kubectl get pods -n cert-manager
NAME                                      READY   STATUS    RESTARTS   AGE
cert-manager-7747db9d88-bd2nl             1/1     Running   0          21d
cert-manager-cainjector-87c85c6ff-59sb5   1/1     Running   0          21d
cert-manager-webhook-64dc9fff44-29cfc     1/1     Running   0          21d

Also if your cluster operates behind a corporate proxy make sure that the API server is configured not to send requests to cluster services through the proxy. You can check that with the following command:

$ kubectl describe pod kube-apiserver --namespace kube-system | grep -i no_proxy | grep "\.svc"

In case there's no output and your cluster was deployed with kubeadm open /etc/kubernetes/manifests/kube-apiserver.yaml at the control plane nodes and append .svc and .svc.cluster.local to the no_proxy environment variable:

apiVersion: v1
kind: Pod
metadata:
  ...
spec:
  containers:
  - command:
    - kube-apiserver
    - --advertise-address=10.237.71.99
    ...
    env:
    - name: http_proxy
      value: http://proxy.host:8080
    - name: https_proxy
      value: http://proxy.host:8433
    - name: no_proxy
      value: 127.0.0.1,localhost,.example.com,10.0.0.0/8,.svc,.svc.cluster.local
    ...

Note: To build clusters using kubeadm with the right no_proxy settings from the very beginning, set the cluster service names to $no_proxy before kubeadm init:

$ export no_proxy=$no_proxy,.svc,.svc.cluster.local

Finally deploy the operator itself:

$ kubectl apply -k https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/operator/default?ref=<RELEASE_VERSION>

Now you can deploy the device plugins by creating corresponding custom resources. The samples for them are available here.

Usage

Deploy your device plugin by applying its custom resource, e.g. GpuDevicePlugin with

$ kubectl apply -f https://raw.githubusercontent.com/intel/intel-device-plugins-for-kubernetes/main/deployments/operator/samples/deviceplugin_v1_gpudeviceplugin.yaml

Observe it is up and running:

$ kubectl get GpuDevicePlugin
NAME                     DESIRED   READY   NODE SELECTOR   AGE
gpudeviceplugin-sample   1         1                       5s

In order to limit the deployment to a specific device type, use one of kustomizations under deployments/operator/device.

For example, to limit the deployment to FPGA, use:

$ kubectl apply -k deployments/operator/device/fpga

Operator also supports deployments with multiple selected device types. In this case, create a new kustomization with the necessary resources that passes the desired device types to the operator using --device command line argument multiple times.

Upgrade

The upgrade of the deployed plugins can be done by simply installing a new release of the operator.

The operator auto-upgrades operator-managed plugins (CR images and thus corresponding deployed daemonsets) to the current release of the operator.

The [registry-url]/[namespace]/[image] are kept intact on the upgrade.

No upgrade is done for:

  • Non-operator managed deployments
  • Operator deployments without numeric tags

Known issues

When the operator is run with leader election enabled, that is with the option --leader-elect, make sure the cluster is not overloaded with excessive number of pods. Otherwise a heart beat used by the leader election code may trigger a timeout and crash. We are going to use different clients for the controller and leader election code to alleviate the issue. See more details in intel#476.

In case the deployment is limited to specific device type(s), the CRDs for other device types are still created, but no controllers for them are registered.