The Node Feature Discovery operator is a tool for Openshift administrators that makes it easy to detect and understand the hardware features and configurations of a cluster's nodes. With this operator, administrators can easily gather information about their nodes that can be used for scheduling, resource management, and more by controlling the life cycle of NFD.
Node Feature Discovery – a Kubernetes add-on for detecting hardware features and system configuration.
The Node Feature Discovery and Node Feature Discovery operator are Upstream projects under the kubernetes-Sigs organization
Prerequisite: a running OpenShift cluster 4.6+
Get the source code
git clone https://github.com/openshift/cluster-nfd-operator
Deploy the operator
IMAGE_REGISTRY=quay.io/<your-personal-registry>
make image push deploy
Create a NodeFeatureDiscovery instance
oc apply -f config/samples/nfd.openshift.io_v1_nodefeaturediscovery.yaml
The Operator will deploy NFD based on the information on the NodeFeatureDiscovery CR instance, after a moment you should be able to see
$ oc -n openshift-nfd get ds,deploy
NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE
daemonset.apps/nfd-worker 3 3 3 3 3 <none> 5s
NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/nfd-master 1/1 1 1 17s
Check that NFD feature labels have been created
$ oc get node -o json | jq .items[].metadata.labels
{
"beta.kubernetes.io/arch": "amd64",
"beta.kubernetes.io/os": "linux",
"feature.node.kubernetes.io/cpu-cpuid.ADX": "true",
"feature.node.kubernetes.io/cpu-cpuid.AESNI": "true",
"feature.node.kubernetes.io/cpu-cpuid.AVX": "true",
...
First see upstream documentation of the hook feature and how to create a correct hook file: https://github.com/kubernetes-sigs/node-feature-discovery#local-user-specific-features.
The DaemonSet running on the workers will mount the hostPath: /etc/kubernetes/node-feature-discovery/source.d
. Additional hooks can than be provided by a sidecar container that is as well running on the workers and mounting the same hostpath and writing the hook executable (shell-script, compiled code, ...) to this directory.
NFD will execute any file in this directory, if one needs any configuration for the hook, a separate configuration directory can be created under /etc/kubernetes/node-feature-discovery/source.d
e.g. /etc/kubernetes/node-feature-discovery/source.d/own-hook-conf
, NFD will not recurse deeper into the file hierarchy.
There 2 options:
- Using ARM server - process is the same, but you should use Dockerfile.arm instead of Dockerfile
- Using x86 server/laptop - process is the same but before running build command, Makefile.arm should be copied into Makefile