Customization guide

Table of contents

1. Overview
2. NodeFeature custom resource
   1. A NodeFeature example
   2. Feature types
3. NodeFeatureRule custom resource
   1. A NodeFeatureRule example
   2. Node tainting
4. NodeFeatureGroup custom resource
   1. A NodeFeatureGroup example
5. Local feature source
   1. An example
   2. Feature files
   3. Hooks
   4. Input format
   5. Mounts
6. Custom feature source
   1. An example custom feature source configuration
   2. Additional configuration directory
7. Node labels
8. Feature rule format
   1. Fields
   2. Available features
   3. Templating
   4. Backreferences
   5. Examples

Overview

NFD provides multiple extension points for vendor and application specific labeling:

• NodeFeature objects can be used to communicate "raw" node features and node labeling requests to nfd-master.
• NodeFeatureRule objects provide a way to deploy custom labeling rules via the Kubernetes API.
• local feature source of nfd-worker creates labels by reading text files and executing hooks.
• custom feature source of nfd-worker creates labels based on user-specified rules.

NodeFeature custom resource

NodeFeature objects provide a way for 3rd party extensions to advertise custom features, both as "raw" features that serve as input to NodeFeatureRule objects and as feature labels directly.

Note that RBAC rules must be created for each extension for them to be able to create and manipulate NodeFeature objects in their namespace.

The NodeFeature CRD API can be disabled with the NodeFeatureAPI feature gate (-feature-gates NodeFeatureAPI=false on command line). The -feature-gates command line flag must be specified for both nfd-master and nfd-worker as it will enable the gRPC communication between them. Note that the gRPC API is DEPRECATED and will be removed in a future release, at which point the NodeFeature API cannot be disabled.

A NodeFeature example

Consider the following referential example:

apiVersion: nfd.k8s-sigs.io/v1alpha1
kind: NodeFeature
metadata:
  labels:
    nfd.node.kubernetes.io/node-name: node-1
  name: vendor-features-for-node-1
spec:
  # Features for NodeFeatureRule matching
  features:
    flags:
      vendor.flags:
        elements:
          feature-x: {}
          feature-y: {}
    attributes:
      vendor.config:
        elements:
          setting-a: "auto"
          knob-b: "123"
    instances:
      vendor.devices:
        elements:
        - attributes:
            model: "dev-1000"
            vendor: "acme"
        - attributes:
            model: "dev-2000"
            vendor: "acme"
  # Labels to be created
  labels:
    vendor.io/feature.enabled: "true"

The object targets node named node-1. It lists two "flag type" features under the vendor.flags domain, two "attribute type" features and under the vendor.config domain and two "instance type" features under the vendor.devices domain. These features will not be directly affecting the node labels but they will be used as input when the NodeFeatureRule objects are evaluated.

In addition, the example requests directly the vendor.io/feature.enabled=true node label to be created.

The nfd.node.kubernetes.io/node-name=<node-name> must be in place for each NodeFeature object as NFD uses it to determine the node which it is targeting.

Feature types

Features have three different types:

• flag features: a set of names without any associated values, e.g. CPUID flags or loaded kernel modules
• attribute features: a set of names each of which has a single value associated with it (essentially a map of key-value pairs), e.g. kernel config flags or os release information
• instance features: a list of instances, each of which has multiple attributes (key-value pairs of their own) associated with it, e.g. PCI or USB devices

NodeFeatureRule custom resource

NodeFeatureRule objects provide an easy way to create vendor or application specific labels and taints. It uses a flexible rule-based mechanism for creating labels and optionally taints based on node features.

A NodeFeatureRule example

Consider the following referential example:

apiVersion: nfd.k8s-sigs.io/v1alpha1
kind: NodeFeatureRule
metadata:
  name: my-sample-rule-object
spec:
  rules:
    - name: "my sample rule"
      labels:
        "feature.node.kubernetes.io/my-sample-feature": "true"
      matchFeatures:
        - feature: kernel.loadedmodule
          matchExpressions:
            dummy: {op: Exists}
        - feature: kernel.config
          matchExpressions:
            X86: {op: In, value: ["y"]}

It specifies one rule which creates node label feature.node.kubernetes.io/my-sample-feature=true if both of the following conditions are true (matchFeatures implements a logical AND over the matchers):

• The dummy network driver module has been loaded
• X86 option in kernel config is set to =y

Create a NodeFeatureRule with a yaml file:

kubectl apply -f https://raw.githubusercontent.com/kubernetes-sigs/node-feature-discovery/v0.16.6/examples/nodefeaturerule.yaml

Now, on X86 platforms the feature label appears after doing modprobe dummy on a system and correspondingly the label is removed after rmmod dummy. Note a re-labeling delay up to the sleep-interval of nfd-worker (1 minute by default).

See Feature rule format for detailed description of available fields and how to write labeling rules.

Node tainting

This feature is experimental.

In some circumstances, it is desirable to keep nodes with specialized hardware away from running general workload and instead leave them for workloads that need the specialized hardware. One way to achieve it is to taint the nodes with the specialized hardware and add corresponding toleration to pods that require the special hardware. NFD offers node tainting functionality which is disabled by default. User can define one or more custom taints via the taints field of the NodeFeatureRule CR. The same rule-based mechanism is applied here and the NFD taints only rule matching nodes.

To enable the tainting feature, --enable-taints flag needs to be set to true. If the flag --enable-taints is set to false (i.e. disabled), taints defined in the NodeFeatureRule CR have no effect and will be ignored by the NFD master.

See documentation of the taints field for detailed description how to specify taints in the NodeFeatureRule object.

NOTE: Before enabling any taints, make sure to edit nfd-worker daemonset to tolerate the taints to be created. Otherwise, already running pods that do not tolerate the taint are evicted immediately from the node including the nfd-worker pod.

NodeFeatureGroup custom resource

NodeFeatureGroup API is an alpha feature and disabled by default in NFD version v0.16. Use the NodeFeatureAPI feature gate to enable it.

NodeFeatureGroup objects provide a way to create node groups that share the same set of features. The NodeFeatureGroup object spec consists of a list of NodeFeatureRule that follow the same format as the NodeFeatureRule, but the difference in this case is that nodes that match any of the rules in the NodeFeatureGroup will be listed in the NodeFeatureGroup status.

A NodeFeatureGroup example

Consider the following referential example:

apiVersion: nfd.k8s-sigs.io/v1alpha1
kind: NodeFeatureGroup
metadata:
  name: node-feature-group-example
spec:
  featureGroupRules:
    - name: "kernel version"
      matchFeatures:
        - feature: kernel.version
          matchExpressions:
            major: {op: In, value: ["6"]}
status:
  nodes:
    - name: node-1
    - name: node-2
    - name: node-3

The object specifies a group of nodes that share the same kernel.version.major (Linux kernel v6.x).

Create a NodeFeatureGroup with a yaml file:

kubectl apply -f https://raw.githubusercontent.com/kubernetes-sigs/node-feature-discovery/v0.16.6/examples/nodefeaturegroup.yaml

See Feature rule format for detailed description of available fields and how to write group filtering rules.

Local feature source

NFD-Worker has a special feature source named local which is an integration point for external feature detectors. It provides a mechanism for pluggable extensions, allowing the creation of new user-specific features and even overriding built-in labels.

The local feature source has two methods for detecting features, feature files and hooks (hooks are deprecated and slated for removal in NFD v0.17). The features discovered by the local source can further be used in label rules specified in NodeFeatureRule objects and the custom feature source.

NOTE: Be careful when creating and/or updating hook or feature files while NFD is running. To avoid race conditions you should write into a temporary file, and atomically create/update the original file by doing a file rename operation. NFD ignores dot files, so temporary file can be written to the same directory and renamed (.my.feature -> my.feature) once file is complete. Both file names should (obviously) be unique for the given application.

An example

Consider a plaintext file /etc/kubernetes/node-feature-discovery/features.d/my-features having the following contents (or alternatively a shell script /etc/kubernetes/node-feature-discovery/source.d/my-hook.sh having the following stdout output):

feature.node.kubernetes.io/my-feature.1
feature.node.kubernetes.io/my-feature.2=myvalue
vendor.io/my-feature.3=456

This will translate into the following node labels:

feature.node.kubernetes.io/my-feature.1: "true"
feature.node.kubernetes.io/my-feature.2: "myvalue"
vendor.io/my-feature.3: "456"

Feature files

The local source reads files found in /etc/kubernetes/node-feature-discovery/features.d/. File content is parsed and translated into node labels, see the input format below.

Hooks

DEPRECATED Hooks are deprecated and will be completely removed in NFD v0.17.

The local source executes hooks found in /etc/kubernetes/node-feature-discovery/source.d/. The hook files must be executable and they are supposed to print all discovered features in stdout. Since NFD v0.13 the default container image only supports statically linked ELF binaries.

stderr output of hooks is propagated to NFD log so it can be used for debugging and logging.

NFD tries to execute any regular files found from the hooks directory. Any additional data files the hook might need (e.g. a configuration file) should be placed in a separate directory to avoid NFD unnecessarily trying to execute them. A subdirectory under the hooks directory can be used, for example /etc/kubernetes/node-feature-discovery/source.d/conf/.

NOTE: Starting from release v0.14 hooks are disabled by default and can be enabled via sources.local.hooksEnabled field in the worker configuration.

sources:
  local:
    hooksEnabled: true  # true by default at this point

NOTE: NFD will blindly run any executables placed/mounted in the hooks directory. It is the user's responsibility to review the hooks for e.g. possible security implications.

NOTE: The full image variant provides backwards-compatibility with older NFD versions by including a more expanded environment, supporting bash and perl runtimes.

Input format

The hook stdout and feature files are expected to contain features in simple key-value pairs, separated by newlines:

# This is a comment
<key>[=<value>]

The label value defaults to true, if not specified.

Label namespace must be specified with <namespace>/<name>[=<value>].

NOTE: The feature file size limit it 64kB. The feature file will be ignored if the size limit is exceeded.

Comment lines (starting with #) are ignored.

Adding following line anywhere to feature file defines date when its content expires / is ignored:

# +expiry-time=2023-07-29T11:22:33Z

Also, the expiry-time value would stay the same during the processing of the feature file until another expiry-time directive is encountered. Considering the following file:

# +expiry-time=2012-07-28T11:22:33Z
vendor.io/feature1=featureValue

# +expiry-time=2080-07-28T11:22:33Z
vendor.io/feature2=featureValue2

# +expiry-time=2070-07-28T11:22:33Z
vendor.io/feature3=featureValue3

# +expiry-time=2002-07-28T11:22:33Z
vendor.io/feature4=featureValue4

After processing the above file, only vendor.io/feature2 and vendor.io/feature3 would be included in the list of accepted features.

NOTE: The time format supported is RFC3339. Also, the expiry-time tag is only evaluated in each re-discovery period, and the expiration of node labels is not tracked.

To exclude specific features from the local.feature Feature, you can use the # +no-feature directive. The # +no-label directive causes the feature to be excluded from the local.label Feature and a node label not to be generated.

Considering the following file:

# +no-feature
vendor.io/label-only=value

vendor.io/my-feature=value

vendor.io/foo=bar

# +no-label
foo=baz

Processing the above file would result in the following Features:

local.features:
  foo: baz
  vendor.io/my-feature: value
local.labels:
  vendor.io/label-only: value
  vendor.io/my-feature: value

and the following labels added to the Node:

vendor.io/label-only=value
vendor.io/my-feature=value

NOTE: use of unprefixed label names (like foo=bar) should not be used. In NFD v0.16 unprefixed names will be automatically prefixed with feature.node.kubernetes.io/ but this will change in a future version (see autoDefaultNs config option. Unprefixed names for plain Features (tagged with # +no-label) can be used without restrictions, however.

Mounts

The standard NFD deployments contain hostPath mounts for /etc/kubernetes/node-feature-discovery/source.d/ and /etc/kubernetes/node-feature-discovery/features.d/, making these directories from the host available inside the nfd-worker container.

Injecting labels from other pods

One use case for the feature files and hooks is detecting features in other Pods outside NFD, e.g. in Kubernetes device plugins. By using the same hostPath mounts for /etc/kubernetes/node-feature-discovery/source.d/ and /etc/kubernetes/node-feature-discovery/features.d/ in the side-car (e.g. device plugin) creates a shared area for deploying feature files and hooks to NFD. NFD periodically scans the directories and reads any feature files and runs any hooks it finds.

Custom feature source

The custom feature source in nfd-worker provides a rule-based mechanism for label creation, similar to the NodeFeatureRule objects. The difference is that the rules are specified in the worker configuration instead of a Kubernetes API object.

See worker configuration for instructions how to set-up and manage the worker configuration.

An example custom feature source configuration

Consider the following referential configuration for nfd-worker:

core:
  labelSources: ["custom"]
sources:
  custom:
    - name: "my sample rule"
      labels:
        "feature.node.kubenernetes.io/my-sample-feature": "true"
      matchFeatures:
        - feature: kernel.loadedmodule
          matchExpressions:
            dummy: {op: Exists}
        - feature: kernel.config
          matchExpressions:
            X86: {op: In, value: ["y"]}

It specifies one rule which creates node label feature.node.kubenernetes.io/my-sample-feature=true if both of the following conditions are true (matchFeatures implements a logical AND over the matchers):

• The dummy network driver module has been loaded
• X86 option in kernel config is set to =y

In addition, the configuration only enables the custom source, disabling all built-in labels.

Now, on X86 platforms the feature label appears after doing modprobe dummy on a system and correspondingly the label is removed after rmmod dummy. Note a re-labeling delay up to the sleep-interval of nfd-worker (1 minute by default).

Additional configuration directory

In addition to the rules defined in the nfd-worker configuration file, the custom feature source can read more configuration files located in the /etc/kubernetes/node-feature-discovery/custom.d/ directory. This makes more dynamic and flexible configuration easier.

As an example, consider having file /etc/kubernetes/node-feature-discovery/custom.d/my-rule.yaml with the following content:

- name: "my e1000 rule"
  labels:
    "feature.node.kubenernetes.io/e1000.present": "true"
  matchFeatures:
    - feature: kernel.loadedmodule
      matchExpressions:
        e1000: {op: Exists}

This simple rule will create feature.node.kubenernetes.io/e1000.present=true label if the e1000 kernel module has been loaded.

The samples/custom-rules kustomize overlay sample contains an example for deploying a custom rule from a ConfigMap.

Node labels

Feature labels have the following format:

<namespace>/<name> = <value>

The namespace part (i.e. prefix) of the labels is controlled by nfd:

• All built-in labels use feature.node.kubernetes.io.
• Namespaces may be excluded with the -deny-label-ns command line flag of nfd-master
  • To allow specific namespaces that were denied, you can use -extra-label-ns command line flag of nfd-master. e.g: nfd-master -deny-label-ns="*" -extra-label-ns=example.com

Feature rule format

This section describes the rule format used in NodeFeatureRule objects and in the configuration of the custom feature source.

It is based on a generic feature matcher that covers all features discovered by nfd-worker. The rules rely on a unified data model of the available features and a generic expression-based format. Features that can be used in the rules are described in detail in available features below.

Take this rule as a referential example:

    - name: "my feature rule"
      labels:
        "feature.node.kubernetes.io/my-special-feature": "my-value"
      matchFeatures:
        - feature: cpu.cpuid
          matchExpressions:
            AVX512F: {op: Exists}
        - feature: kernel.version
          matchExpressions:
            major: {op: In, value: ["5"]}
            minor: {op: Gt, value: ["1"]}
        - feature: pci.device
          matchExpressions:
            vendor: {op: In, value: ["8086"]}
            class: {op: In, value: ["0200"]}

This will yield feature.node.kubernetes.io/my-special-feature=my-value node label if all of these are true (matchFeatures implements a logical AND over the matchers):

• the CPU has AVX512F capability
• kernel version is 5.2 or later (must be v5.x)
• an Intel network controller is present

Fields

name

The .name field is required and used as an identifier of the rule.

labels

The .labels is a map of the node labels to create if the rule matches.

Take this rule as a referential example:

apiVersion: nfd.k8s-sigs.io/v1alpha1
kind: NodeFeatureRule
metadata:
  name: my-sample-rule-object
spec:
  rules:
    - name: "my dynamic label value rule"
      labels:
        feature.node.kubernetes.io/linux-lsm-enabled: "@kernel.config.LSM"
        feature.node.kubernetes.io/custom-label: "customlabel"

Label linux-lsm-enabled uses the @ notation for dynamic values. The value of the label will be the value of the attribute LSM of the feature kernel.config.

The @<feature-name>.<element-name> format can be used to inject values of detected features to the label. See available features for possible values to use.

This will yield into the following node label:

  labels:
    ...
    feature.node.kubernetes.io/linux-lsm-enabled: apparmor
    feature.node.kubernetes.io/custom-label: "customlabel"

labelsTemplate

The .labelsTemplate field specifies a text template for dynamically creating labels based on the matched features. See templating for details.

NOTE: The labels field has priority over labelsTemplate, i.e. labels specified in the labels field will override anything originating from labelsTemplate.

annotations

The .annotations field is a list of features to be advertised as node annotations.

Take this rule as a referential example:

apiVersion: nfd.k8s-sigs.io/v1alpha1
kind: NodeFeatureRule
metadata:
  name: feature-annotations-example
spec:
  rules:
    - name: "annotation-example"
      annotations:
        feature.node.kubernetes.io/defaul-ns-annotation: "foo"
        custom.vendor.io/feature: "baz"
      matchFeatures:
        - feature: kernel.version
          matchExpressions:
            major: {op: Exists}

This will yield into the following node annotations:

  annotations:
    ...
    feature.node.kubernetes.io/defaul-ns-annotation: "foo"
    custom.vendor.io/feature: "baz"
    ...

NFD enforces some limitations to the namespace (or prefix)/ of the annotations:

• kubernetes.io/ and its sub-namespaces (like sub.ns.kubernetes.io/) cannot generally be used
• the only exception is feature.node.kubernetes.io/ and its sub-namespaces (like sub.ns.feature.node.kubernetes.io)
• unprefixed names (like my-annotation) should not be used. In NFD v0.16 unprefixed names will be automatically prefixed with feature.node.kubernetes.io/ but this will change in a future version (see autoDefaultNs config option.

NOTE: The annotations field has will only advertise features via node annotations the features won't be advertised as node labels unless they are specified in the labels field.

taints

taints is a list of taint entries and each entry can have key, value and effect, where the value is optional. Effect could be NoSchedule, PreferNoSchedule or NoExecute. To learn more about the meaning of these effects, check out k8s documentation.

Example NodeFeatureRule with taints:

apiVersion: nfd.k8s-sigs.io/v1alpha1
kind: NodeFeatureRule
metadata:
  name: my-sample-rule-object
spec:
  rules:
    - name: "my sample taint rule"
      taints:
        - effect: PreferNoSchedule
          key: "feature.node.kubernetes.io/special-node"
          value: "true"
        - effect: NoExecute
          key: "feature.node.kubernetes.io/dedicated-node"
      matchFeatures:
        - feature: kernel.loadedmodule
          matchExpressions:
            dummy: {op: Exists}
        - feature: kernel.config
          matchExpressions:
            X86: {op: In, value: ["y"]}

In this example, if the my sample taint rule rule is matched, feature.node.kubernetes.io/pci-0300_1d0f.present=true:NoExecute and feature.node.kubernetes.io/cpu-cpuid.ADX:NoExecute taints are set on the node.

There are some limitations to the namespace part (i.e. prefix/) of the taint key:

• kubernetes.io/ and its sub-namespaces (like sub.ns.kubernetes.io/) cannot generally be used
• the only exception is feature.node.kubernetes.io/ and its sub-namespaces (like sub.ns.feature.node.kubernetes.io)
• unprefixed keys (like foo) keys are disallowed

NOTE: taints field is not available for the custom rules of nfd-worker and only for NodeFeatureRule objects.

vars

The .vars field is a map of values (key-value pairs) to store for subsequent rules to use. In other words, these are variables that are not advertised as node labels. See backreferences for more details on the usage of vars.

extendedResources

The .extendedResources field is a list of extended resources to advertise. See extended resources for more details.

Take this rule as a referential example:

apiVersion: nfd.k8s-sigs.io/v1alpha1
kind: NodeFeatureRule
metadata:
  name: my-extended-resource-rule
spec:
  rules:
    - name: "my extended resource rule"
      extendedResources:
        vendor.io/dynamic: "@kernel.version.major"
        vendor.io/static: "123"
      matchFeatures:
        - feature: kernel.version
          matchExpressions:
            major: {op: Exists}

The extended resource vendor.io/dynamic is defined in the form @feature.attribute. The value of the extended resource will be the value of the attribute major of the feature kernel.version.

The @<feature-name>.<element-name> format can be used to inject values of detected features to the extended resource. See available features for possible values to use. Note that the value must be eligible as a Kubernetes resource quantity.

This will yield into the following node status:

  allocatable:
    ...
    vendor.io/dynamic: "5"
    vendor.io/static: "123"
    ...
  capacity:
    ...
    vendor.io/dynamic: "5"
    vendor.io/static: "123"
    ...

There are some limitations to the namespace part (i.e. prefix)/ of the Extended Resources names:

• kubernetes.io/ and its sub-namespaces (like sub.ns.kubernetes.io/) cannot generally be used
• the only exception is feature.node.kubernetes.io/ and its sub-namespaces (like sub.ns.feature.node.kubernetes.io)
• unprefixed names (like my-er) site.version }} unprefixed names will be automatically prefixed with feature.node.kubernetes.io/ but this will change in a future version (see autoDefaultNs config option.

NOTE: .extendedResources is not supported by the custom feature source – it can only be used in NodeFeatureRule objects.

varsTemplate

The .varsTemplate field specifies a text template for dynamically creating vars based on the matched features. See templating for details on using templates and backreferences for more details on the usage of vars.

NOTE: The vars field has priority over varsTemplate, i.e. vars specified in the vars field will override anything originating from varsTemplate.

matchFeatures

The .matchFeatures field specifies a feature matcher, consisting of a list of feature matcher terms. It implements a logical AND over the terms i.e. all of them must match for the rule to trigger.

      matchFeatures:
        - feature: <feature-name>
          matchExpressions:
            <key>:
              op: <op>
              value:
                - <value-1>
                - ...
          matchName:
            op: <op>
            value:
                - <value-1>
                - ...

The .matchFeatures[].feature field specifies the feature which to evaluate.

NOTE:If both matchExpressions and matchName are specified, they both must match.

matchExpressions

The .matchFeatures[].matchExpressions field is used to match against the value(s) of a feature. The matchExpressions field consists of a set of expressions, each of which is evaluated against all elements of the specified feature.

      matchExpressions:
        <key>:
          op: <op>
          value:
            - <value-1>
            - ...

In each MatchExpression the key specifies the name of of the feature element (flag and attribute features) or name of the attribute (instance features) which to look for. The behavior of MatchExpression depends on the feature type:

• for flag and attribute features the MatchExpression operates on the feature element whose name matches the <key>
• for instance features all MatchExpressions are evaluated against the attributes of each instance separately

The op field specifies the operator to apply. Valid values are described below.

The value field of MatchExpression is a list of string arguments to the operator.

matchName

The .matchFeatures[].matchName field is used to match against the name(s) of a feature (whereas the matchExpressions field matches against the value(s). The matchName field consists of a single expression which is evaulated against the name of each element of the specified feature.

      matchName:
        op: <op>
        value:
          - <value-1>
          - ...

The behavior of matchName depends on the feature type:

• for flag and attribute features the expression is evaluated against the name of each element
• for instance features the expression is evaluated against the name of each attribute, for each element (instance) separately (matches if the attributes of any of the elements satisfy the expression)

The op field specifies the operator to apply. Same operators as for matchExpressions above are available.

Other operators are not practical with matchName (DoesNotExist never matches; Gt,Lt and GtLt are only usable if feature names are integers; IsTrue and IsFalse are only usable if the feature name is true or false).

The value field is a list of string arguments to the operator.

An example:

      matchFeatures:
        - feature: cpu.cpuid
          matchName: {op: InRegexp, value: ["^AVX"]}

The snippet above would match if any CPUID feature starting with AVX is present (e.g. AVX1 or AVX2 or AVX512F etc).

matchAny

The .matchAny field is a list of of matchFeatures matchers. A logical OR is applied over the matchers, i.e. at least one of them must match for the rule to trigger.

Consider the following example:

      matchAny:
        - matchFeatures:
            - feature: kernel.loadedmodule
              matchExpressions:
                kmod-1: {op: Exists}
            - feature: pci.device
              matchExpressions:
                vendor: {op: In, value: ["0eee"]}
                class: {op: In, value: ["0200"]}
        - matchFeatures:
            - feature: kernel.loadedmodule
              matchExpressions:
                kmod-2: {op: Exists}
            - feature: pci.device
              matchExpressions:
                vendor: {op: In, value: ["0fff"]}
                class: {op: In, value: ["0200"]}

This matches if kernel module kmod-1 is loaded and a network controller from vendor 0eee is present, OR, if kernel module kmod-2 has been loaded and a network controller from vendor 0fff is present (OR both of these conditions are true).

Available features

The following features are available for matching:

Intel RDT flags

Templating

Rules support template-based creation of labels and vars with the .labelsTemplate and .varsTemplate fields. These makes it possible to dynamically generate labels and vars based on the features that matched.

The template must expand into a simple format with <key>=<value> pairs separated by newline.

Consider the following example:

    labelsTemplate: |
      {{ range .pci.device }}vendor-{{ .class }}-{{ .device }}.present=true
      {{ end }}
    matchFeatures:
      - feature: pci.device
        matchExpressions:
          class: {op: InRegexp, value: ["^02"]}
          vendor: ["0fff"]

The rule above will create individual labels feature.node.kubernetes.io/vendor-<class-id>-<device-id>.present=true for each network controller device (device class starting with 02) from vendor 0fff.

All the matched features of each feature matcher term under matchFeatures fields are available for the template engine. Matched features can be referenced with {{ .<feature-name> }} in the template, and the available data could be described in yaml as follows:

.
  <key-feature>:
    - Name: <matched-key>
    - ...

  <value-feature>:
    - Name: <matched-key>
      Value: <matched-value>
    - ...

  <instance-feature>:
    - <attribute-1-name>: <attribute-1-value>
      <attribute-2-name>: <attribute-2-value>
      ...
    - ...

That is, the per-feature data is a list of objects whose data fields depend on the type of the feature:

• for flag features only ‘Name' is available
• for value features ‘Name' and ‘Value' are available
• for instance features all attributes of the matched instance are available

A simple example of a template utilizing name and value from an attribute feature:

    labelsTemplate: |
      {{ range .system.osrelease }}system-{{ .Name }}={{ .Value }}
      {{ end }}
    matchFeatures:
      - feature: system.osRelease
        matchExpressions:
          ID: {op: Exists}
          VERSION_ID.major: {op: Exists}

NOTE:If both matchExpressions and matchName for a feature matcher term (see matchFeatures) is specified, the list of matched features (for the template engine) is the union from both of these. NOTE: In case of matchAny is specified, the template is executed separately against each individual matchFeatures field and the final set of labels will be superset of all these separate template expansions. E.g. consider the following:

  - name: <name>
    labelsTemplate: <template>
    matchFeatures: <matcher#1>
    matchAny:
      - matchFeatures: <matcher#2>
      - matchFeatures: <matcher#3>

In the example above (assuming the overall result is a match) the template would be executed on matcher#1 as well as on matcher#2 and/or matcher#3 (depending on whether both or only one of them match). All the labels from these separate expansions would be created, i.e. the end result would be a union of all the individual expansions.

Rule templates use the Golang text/template package and all its built-in functionality (e.g. pipelines and functions) can be used. An example template taking use of the built-in len function, advertising the number of PCI network controllers from a specific vendor:

    labelsTemplate: |
      num-intel-network-controllers={{ .pci.device | len }}
    matchFeatures:
      - feature: pci.device
        matchExpressions:
          vendor: {op: In, value: ["8086"]}
          class: {op: In, value: ["0200"]}

Imaginative template pipelines are possible, but care must be taken to produce understandable and maintainable rule sets.

Backreferences

Rules support referencing the output of preceding rules. This enables sophisticated scenarios where multiple rules are combined together to for more complex heuristics than a single rule can provide. The labels and vars created by the execution of preceding rules are available as a special rule.matched feature.

Consider the following configuration:

  - name: "my kernel label rule"
    labels:
      kernel-feature: "true"
    matchFeatures:
      - feature: kernel.version
        matchExpressions:
          major: {op: Gt, value: ["4"]}

  - name: "my var rule"
    vars:
      nolabel-feature: "true"
    matchFeatures:
      - feature: cpu.cpuid
        matchExpressions:
          AVX512F: {op: Exists}
      - feature: pci.device
        matchExpressions:
          vendor: {op: In, value: ["0fff"]}
          device: {op: In, value: ["1234", "1235"]}

  - name: "my high level feature rule"
    labels:
      high-level-feature: "true"
    matchFeatures:
      - feature: rule.matched
        matchExpressions:
          kernel-feature: {op: IsTrue}
          nolabel-feature: {op: IsTrue}

The feature.node.kubernetes.io/high-level-feature = true label depends on the two previous rules.

Note that when referencing rules across multiple NodeFeatureRule objects attention must be paid to the ordering. NodeFeatureRule objects are processed in alphabetical order (based on their .metadata.name).

Examples

Some more configuration examples below.

Match certain CPUID features:

  - name: "example cpuid rule"
    labels:
      my-special-cpu-feature: "true"
    matchFeatures:
      - feature: cpu.cpuid
        matchExpressions:
          AESNI: {op: Exists}
          AVX: {op: Exists}

Require a certain loaded kernel module and OS version:

  - name: "my multi-feature rule"
    labels:
      my-special-multi-feature: "true"
    matchFeatures:
      - feature: kernel.loadedmodule
        matchExpressions:
          e1000: {op: Exists}
      - feature: system.osrelease
        matchExpressions:
          NAME: {op: InRegexp, values: ["^openSUSE"]}
          VERSION_ID.major: {op: Gt, values: ["14"]}

Require a loaded kernel module and two specific PCI devices (both of which must be present):

  - name: "my multi-device rule"
    labels:
      my-multi-device-feature: "true"
    matchFeatures:
      - feature: kernel.loadedmodule
        matchExpressions:
          my-driver-module: {op: Exists}
      - pci.device:
          vendor: "0fff"
          device: "1234"
      - pci.device:
          vendor: "0fff"
          device: "abcd"