Deployment and usage

Table of contents

1. Requirements
2. Image variants
   1. Full
   2. Minimal
3. Deployment options
   1. Operator
   2. Deployment templates
   3. Deployment with Helm
   4. Build your own
4. Usage
   1. NFD-Master
   2. NFD-Worker
   3. Communication security with TLS
5. Worker configuration
6. Using node labels
7. Uninstallation
   1. Operator was used for deployment
   2. Manual
   3. Removing feature labels

Requirements

1. Linux (x86_64/Arm64/Arm)
2. kubectl (properly set up and configured to work with your Kubernetes cluster)

Image variants

NFD currently offers two variants of the container image. The "full" variant is currently deployed by default.

Full

This image is based on debian:buster-slim and contains a full Linux system for running shell-based nfd-worker hooks and doing live debugging and diagnosis of the NFD images.

Minimal

This is a minimal image based on gcr.io/distroless/base and only supports running statically linked binaries.

The container image tag has suffix -minimal (e.g. k8s.gcr.io/nfd/node-feature-discovery:v0.8.2-minimal)

Deployment options

Operator

Deployment using the Node Feature Discovery Operator is recommended to be done via operatorhub.io.

1. You need to have OLM installed. If you don't, take a look at the latest release for detailed instructions.

2. Install the operator:

   kubectl create -f https://operatorhub.io/install/nfd-operator.yaml
   

3. Create NodeFeatureDiscovery object (in nfd namespace here):

   cat << EOF | kubectl apply -f -
   apiVersion: v1
   kind: Namespace
   metadata:
     name: nfd
   ---
   apiVersion: nfd.kubernetes.io/v1
   kind: NodeFeatureDiscovery
   metadata:
     name: my-nfd-deployment
     namespace: nfd
   spec:
     operand:
       namespace: nfd
       image: k8s.gcr.io/nfd/node-feature-discovery:v0.8.2
       imagePullPolicy: IfNotPresent
   EOF
   

In order to deploy the minimal image you need to use

  image: k8s.gcr.io/nfd/node-feature-discovery:v0.8.2-minimal

in the NodeFeatureDiscovery object above.

Deployment templates

The template specs provided in the repo can be used directly:

kubectl apply -f https://raw.githubusercontent.com/kubernetes-sigs/node-feature-discovery/v0.8.2/nfd-master.yaml.template
kubectl apply -f https://raw.githubusercontent.com/kubernetes-sigs/node-feature-discovery/v0.8.2/nfd-worker-daemonset.yaml.template

This will required RBAC rules and deploy nfd-master (as a deployment) and nfd-worker (as a daemonset) in the node-feature-discovery namespace.

Alternatively you can download the templates and customize the deployment manually. For example, to deploy the minimal image.

Master-worker pod

You can also run nfd-master and nfd-worker inside the same pod

kubectl apply -f https://raw.githubusercontent.com/kubernetes-sigs/node-feature-discovery/v0.8.2/nfd-daemonset-combined.yaml.template

This creates a DaemonSet runs both nfd-worker and nfd-master in the same Pod. In this case no nfd-master is run on the master node(s), but, the worker nodes are able to label themselves which may be desirable e.g. in single-node setups.

Worker one-shot

Feature discovery can alternatively be configured as a one-shot job. The Job template may be used to achieve this:

NUM_NODES=$(kubectl get no -o jsonpath='{.items[*].metadata.name}' | wc -w)
curl -fs https://raw.githubusercontent.com/kubernetes-sigs/node-feature-discovery/v0.8.2/nfd-worker-job.yaml.template | \
    sed s"/NUM_NODES/$NUM_NODES/" | \
    kubectl apply -f -

The example above launces as many jobs as there are non-master nodes. Note that this approach does not guarantee running once on every node. For example, tainted, non-ready nodes or some other reasons in Job scheduling may cause some node(s) will run extra job instance(s) to satisfy the request.

Deployment with Helm

Node Feature Discovery Helm chart allow to easily deploy and manage NFD.

Prerequisites

Helm package manager should be installed.

Deployment

To install the latest stable version:

export NFD_NS=node-feature-discovery
helm repo add nfd https://kubernetes-sigs.github.io/node-feature-discovery/charts
helm repo update
helm install nfd/node-feature-discovery --namespace $NFD_NS --create-namespace --generate-name

To install the latest development version you need to clone the NFD Git repository and install from there.

git clone https://github.com/kubernetes-sigs/node-feature-discovery/
cd node-feature-discovery/deployment
export NFD_NS=node-feature-discovery
helm install node-feature-discovery ./node-feature-discovery/ --namespace $NFD_NS --create-namespace

See the configuration section below for instructions how to alter the deployment parameters.

In order to deploy the minimal image you need to override the image tag:

helm install node-feature-discovery ./node-feature-discovery/ --set image.tag=v0.8.2-minimal --namespace $NFD_NS --create-namespace

Configuration

You can override values from values.yaml and provide a file with custom values:

export NFD_NS=node-feature-discovery
helm install nfd/node-feature-discovery -f <path/to/custom/values.yaml> --namespace $NFD_NS --create-namespace

To specify each parameter separately you can provide them to helm install command:

export NFD_NS=node-feature-discovery
helm install nfd/node-feature-discovery --set nameOverride=NFDinstance --set master.replicaCount=2 --namespace $NFD_NS --create-namespace

Uninstalling the chart

To uninstall the node-feature-discovery deployment:

export NFD_NS=node-feature-discovery
helm uninstall node-feature-discovery --namespace $NFD_NS

The command removes all the Kubernetes components associated with the chart and deletes the release.

Chart parameters

In order to tailor the deployment of the Node Feature Discovery to your cluster needs We have introduced the following Chart parameters.

General parameters

Master pod parameters

Worker pod parameters

Build your own

If you want to use the latest development version (master branch) you need to build your own custom image. See the Developer Guide for instructions how to build images and deploy them on your cluster.

Usage

NFD-Master

NFD-Master runs as a deployment (with a replica count of 1), by default it prefers running on the cluster's master nodes but will run on worker nodes if no master nodes are found.

For High Availability, you should simply increase the replica count of the deployment object. You should also look into adding inter-pod affinity to prevent masters from running on the same node. However note that inter-pod affinity is costly and is not recommended in bigger clusters.

NFD-Master listens for connections from nfd-worker(s) and connects to the Kubernetes API server to add node labels advertised by them.

If you have RBAC authorization enabled (as is the default e.g. with clusters initialized with kubeadm) you need to configure the appropriate ClusterRoles, ClusterRoleBindings and a ServiceAccount in order for NFD to create node labels. The provided template will configure these for you.

NFD-Worker

NFD-Worker is preferably run as a Kubernetes DaemonSet. This assures re-labeling on regular intervals capturing changes in the system configuration and makes sure that new nodes are labeled as they are added to the cluster. Worker connects to the nfd-master service to advertise hardware features.

When run as a daemonset, nodes are re-labeled at an interval specified using the -sleep-interval option. In the template the default interval is set to 60s which is also the default when no -sleep-interval is specified. Also, the configuration file is re-read on each iteration providing a simple mechanism of run-time reconfiguration.

Communication security with TLS

NFD supports mutual TLS authentication between the nfd-master and nfd-worker instances. That is, nfd-worker and nfd-master both verify that the other end presents a valid certificate.

TLS authentication is enabled by specifying -ca-file, -key-file and -cert-file args, on both the nfd-master and nfd-worker instances. The template specs provided with NFD contain (commented out) example configuration for enabling TLS authentication.

The Common Name (CN) of the nfd-master certificate must match the DNS name of the nfd-master Service of the cluster. By default, nfd-master only check that the nfd-worker has been signed by the specified root certificate (-ca-file). Additional hardening can be enabled by specifying -verify-node-name in nfd-master args, in which case nfd-master verifies that the NodeName presented by nfd-worker matches the Common Name (CN) or a Subject Alternative Name (SAN) of its certificate.

Automated TLS certificate management using cert-manager

cert-manager can be used to automate certificate management between nfd-master and the nfd-worker pods. The instructions below describe steps how to set up cert-manager's CA Issuer to sign Certificate requests for NFD components in node-feature-discovery namespace.

$ kubectl apply -f https://github.com/jetstack/cert-manager/releases/download/v1.2.0/cert-manager.yaml
$ make yamls
$ openssl genrsa -out ca.key 2048
$ openssl req -x509 -new -nodes -key ca.key -subj "/CN=nfd-ca" -days 10000 -out ca.crt
$ sed s"/tls.key:.*/tls.key: $(cat ca.key|base64 -w 0)/" -i nfd-cert-manager.yaml
$ sed s"/tls.crt:.*/tls.crt: $(cat ca.crt|base64 -w 0)/" -i nfd-cert-manager.yaml
$ kubectl apply -f nfd-cert-manager.yaml

Finally, deploy nfd-master.yaml and nfd-worker-daemonset.yaml with the Secrets (nfd-master-cert and nfd-worker-cert) mounts enabled.

Worker configuration

NFD-Worker supports dynamic configuration through a configuration file. The default location is /etc/kubernetes/node-feature-discovery/nfd-worker.conf, but, this can be changed by specifying the-config command line flag. Configuration file is re-read whenever it is modified which makes run-time re-configuration of nfd-worker straightforward.

Worker configuration file is read inside the container, and thus, Volumes and VolumeMounts are needed to make your configuration available for NFD. The preferred method is to use a ConfigMap which provides easy deployment and re-configurability.

The provided nfd-worker deployment templates create an empty configmap and mount it inside the nfd-worker containers. Configuration can be edited with:

kubectl -n ${NFD_NS} edit configmap nfd-worker-conf

See nfd-worker configuration file reference for more details. The (empty-by-default) example config contains all available configuration options and can be used as a reference for creating creating a configuration.

Configuration options can also be specified via the -options command line flag, in which case no mounts need to be used. The same format as in the config file must be used, i.e. JSON (or YAML). For example:

-options='{"sources": { "pci": { "deviceClassWhitelist": ["12"] } } }'

Configuration options specified from the command line will override those read from the config file.

Using node labels

Nodes with specific features can be targeted using the nodeSelector field. The following example shows how to target nodes with Intel TurboBoost enabled.

apiVersion: v1
kind: Pod
metadata:
  labels:
    env: test
  name: golang-test
spec:
  containers:
    - image: golang
      name: go1
  nodeSelector:
    feature.node.kubernetes.io/cpu-pstate.turbo: 'true'

For more details on targeting nodes, see node selection.

Uninstallation

Operator was used for deployment

If you followed the deployment instructions above you can simply do:

kubectl -n nfd delete NodeFeatureDiscovery my-nfd-deployment

Optionally, you can also remove the namespace:

kubectl delete ns nfd

See the node-feature-discovery-operator and OLM project documentation for instructions for uninstalling the operator and operator lifecycle manager, respectively.

Manual

Simplest way is to invoke kubectl delete on the deployment files you used. Beware that this will also delete the namespace that NFD is running in. For example:

kubectl delete -f https://raw.githubusercontent.com/kubernetes-sigs/node-feature-discovery/v0.8.2/nfd-worker-daemonset.yaml.template
kubectl delete -f https://raw.githubusercontent.com/kubernetes-sigs/node-feature-discovery/v0.8.2/nfd-master.yaml.template

Alternatively you can delete create objects one-by-one, depending on the type of deployment, for example:

NFD_NS=node-feature-discovery
kubectl -n $NFD_NS delete ds nfd-worker
kubectl -n $NFD_NS delete deploy nfd-master
kubectl -n $NFD_NS delete svc nfd-master
kubectl -n $NFD_NS delete sa nfd-master
kubectl delete clusterrole nfd-master
kubectl delete clusterrolebinding nfd-master

Removing feature labels

NFD-Master has a special -prune command line flag for removing all nfd-related node labels, annotations and extended resources from the cluster.

kubectl apply -f https://raw.githubusercontent.com/kubernetes-sigs/node-feature-discovery/v0.8.2/nfd-prune.yaml.template
kubectl -n node-feature-discovery wait job.batch/nfd-prune --for=condition=complete && \
    kubectl delete -f https://raw.githubusercontent.com/kubernetes-sigs/node-feature-discovery/v0.8.2/nfd-prune.yaml.template

NOTE: You must run prune before removing the RBAC rules (serviceaccount, clusterrole and clusterrolebinding).