By default OpenShift can schedule a pod to run on any node in the cluster. This can be limited by the use node selectors that limit the pods to only running on nodes that have a matching label. Kernel modules run in the kernel so if they encounter an error they can lead to a kernel panic and the node failing. Loading a kernel module on a node without the hardware it controls is at best pointless and at worst can lead to significant failures.
Therefore discovering what hardware is available on a node and labelling it accordingly is important for the use of Driver Containers, but doing this by hand is time consuming and error prone.
Node Feature Discovery (NFD) is a Kubernetes add-on for detecting hardware features and system configuration, and then labelling the nodes with that information.
It is deployed as an operator, this deploys a worker pod on each node in the cluster. The worker pods periodically check the hardware features and configuration of the node and report back to a master pod. The master pod then adds and removes labels for each node based on this information.
The OpenShift documentation describes the best way to deploy NFD and get started with labelling nodes
Labels are key/value pairs, and must obey some rules about their format.
The keys are made up of an optional prefix and a name, separated by a slash (/).
* The prefix is optional and can be up to 253 characters in length before the slash.
* The name segment is required and must be 63 characters or less,
* They can contain alphanumeric character ([a-z0-9A-Z]), dashes (-), underscores (_), dots (.)
* They must begin and end with an alphanumeric.
The kubernetes.io/ and k8s.io/ prefixes are reserved for Kubernetes core components.
Valid label value:
* must be 63 characters or less,but can be empty (often depicted as "")
* can contain alphanumeric character ([a-z0-9A-Z]), dashes (-), underscores (_), dots (.)
* if not empty must begin and end with an alphanumeric.