1
0
Fork 0
mirror of https://github.com/kubernetes-sigs/node-feature-discovery.git synced 2024-12-14 11:57:51 +00:00
node-feature-discovery/README.md
2020-08-07 10:33:14 +03:00

42 KiB

Node feature discovery for Kubernetes

Build Status Go Report Card

Overview

This software enables node feature discovery for Kubernetes. It detects hardware features available on each node in a Kubernetes cluster, and advertises those features using node labels.

NFD consists of two software components:

  1. nfd-master is responsible for labeling Kubernetes node objects
  2. nfd-worker is detects features and communicates them to nfd-master. One instance of nfd-worker is supposed to be run on each node of the cluster

Command line interface

You can run NFD in stand-alone Docker containers e.g. for testing purposes. This is useful for checking features-detection.

NFD-Master

When running as a standalone container labeling is expected to fail because Kubernetes API is not available. Thus, it is recommended to use --no-publish command line flag. E.g.

$ docker run --rm --name=nfd-test <NFD_CONTAINER_IMAGE> nfd-master --no-publish
2019/02/01 14:48:21 Node Feature Discovery Master <NFD_VERSION>
2019/02/01 14:48:21 gRPC server serving on port: 8080

Command line flags of nfd-master:

$ docker run --rm <NFD_CONTAINER_IMAGE> nfd-master --help
...
nfd-master.

  Usage:
  nfd-master [--no-publish] [--label-whitelist=<pattern>] [--port=<port>]
     [--ca-file=<path>] [--cert-file=<path>] [--key-file=<path>]
     [--verify-node-name] [--extra-label-ns=<list>] [--resource-labels=<list>]
  nfd-master -h | --help
  nfd-master --version

  Options:
  -h --help                       Show this screen.
  --version                       Output version and exit.
  --port=<port>                   Port on which to listen for connections.
                                  [Default: 8080]
  --ca-file=<path>                Root certificate for verifying connections
                                  [Default: ]
  --cert-file=<path>              Certificate used for authenticating connections
                                  [Default: ]
  --key-file=<path>               Private key matching --cert-file
                                  [Default: ]
  --verify-node-name              Verify worker node name against CN from the TLS
                                  certificate. Only has effect when TLS authentication
                                  has been enabled.
  --no-publish                    Do not publish feature labels
  --label-whitelist=<pattern>     Regular expression to filter label names to
                                  publish to the Kubernetes API server.
                                  NB: the label namespace is omitted i.e. the filter
                                  is only applied to the name part after '/'.
                                  [Default: ]
  --extra-label-ns=<list>         Comma separated list of allowed extra label namespaces
                                  [Default: ]
  --resource-labels=<list>        Comma separated list of labels to be exposed as extended resources.
                                  [Default: ]

NFD-Worker

In order to run nfd-worker as a "stand-alone" container against your standalone nfd-master you need to run them in the same network namespace:

$ docker run --rm --network=container:nfd-test <NFD_CONTAINER_IMAGE> nfd-worker
2019/02/01 14:48:56 Node Feature Discovery Worker <NFD_VERSION>
...

If you just want to try out feature discovery without connecting to nfd-master, pass the --no-publish flag to nfd-worker.

Command line flags of nfd-worker:

$ docker run --rm <CONTAINER_IMAGE_ID> nfd-worker --help
...
nfd-worker.

  Usage:
  nfd-worker [--no-publish] [--sources=<sources>] [--label-whitelist=<pattern>]
     [--oneshot | --sleep-interval=<seconds>] [--config=<path>]
     [--options=<config>] [--server=<server>] [--server-name-override=<name>]
     [--ca-file=<path>] [--cert-file=<path>] [--key-file=<path>]
  nfd-worker -h | --help
  nfd-worker --version

  Options:
  -h --help                   Show this screen.
  --version                   Output version and exit.
  --config=<path>             Config file to use.
                              [Default: /etc/kubernetes/node-feature-discovery/nfd-worker.conf]
  --options=<config>          Specify config options from command line. Config
                              options are specified in the same format as in the
                              config file (i.e. json or yaml). These options
                              will override settings read from the config file.
                              [Default: ]
  --ca-file=<path>            Root certificate for verifying connections
                              [Default: ]
  --cert-file=<path>          Certificate used for authenticating connections
                              [Default: ]
  --key-file=<path>           Private key matching --cert-file
                              [Default: ]
  --server=<server>           NFD server address to connecto to.
                              [Default: localhost:8080]
  --server-name-override=<name> Name (CN) expect from server certificate, useful
                              in testing
                              [Default: ]
  --sources=<sources>         Comma separated list of feature sources.
                              [Default: cpu,custom,iommu,kernel,local,memory,network,pci,storage,system,usb]
  --no-publish                Do not publish discovered features to the
                              cluster-local Kubernetes API server.
  --label-whitelist=<pattern> Regular expression to filter label names to
                              publish to the Kubernetes API server.
                              NB: the label namespace is omitted i.e. the filter
                              is only applied to the name part after '/'.
                              [Default: ]
  --oneshot                   Label once and exit.
  --sleep-interval=<seconds>  Time to sleep between re-labeling. Non-positive
                              value implies no re-labeling (i.e. infinite
                              sleep). [Default: 60s]

NOTE Some feature sources need certain directories and/or files from the host mounted inside the NFD container. Thus, you need to provide Docker with the correct --volume options in order for them to work correctly when run stand-alone directly with docker run. See the template spec for up-to-date information about the required volume mounts.

Feature discovery

Feature sources

The current set of feature sources are the following:

  • CPU
  • Custom
  • IOMMU
  • Kernel
  • Memory
  • Network
  • PCI
  • Storage
  • System
  • USB
  • Local (hooks for user-specific features)

Feature labels

The published node labels encode a few pieces of information:

  • Namespace, i.e. feature.node.kubernetes.io
  • The source for each label (e.g. cpu).
  • The name of the discovered feature as it appears in the underlying source, (e.g. cpuid.AESNI from cpu).
  • The value of the discovered feature.

Feature label names adhere to the following pattern:

<namespace>/<source name>-<feature name>[.<attribute name>]

The last component (i.e. attribute-name) is optional, and only used if a feature logically has sub-hierarchy, e.g. sriov.capable and sriov.configure from the network source.

{
  "feature.node.kubernetes.io/cpu-<feature-name>": "true",
  "feature.node.kubernetes.io/custom-<feature-name>": "true",
  "feature.node.kubernetes.io/iommu-<feature-name>": "true",
  "feature.node.kubernetes.io/kernel-<feature name>": "<feature value>",
  "feature.node.kubernetes.io/memory-<feature-name>": "true",
  "feature.node.kubernetes.io/network-<feature-name>": "true",
  "feature.node.kubernetes.io/pci-<device label>.present": "true",
  "feature.node.kubernetes.io/storage-<feature-name>": "true",
  "feature.node.kubernetes.io/system-<feature name>": "<feature value>",
  "feature.node.kubernetes.io/usb-<device label>.present": "<feature value>",
  "feature.node.kubernetes.io/<file name>-<feature name>": "<feature value>"
}

The --sources flag controls which sources to use for discovery.

Note: Consecutive runs of nfd-worker will update the labels on a given node. If features are not discovered on a consecutive run, the corresponding label will be removed. This includes any restrictions placed on the consecutive run, such as restricting discovered features with the --label-whitelist option.

CPU Features

Feature name Attribute Description
cpuid <cpuid flag> CPU capability is supported
hardware_multithreading
Hardware multithreading, such as Intel HTT, enabled (number of logical CPUs is greater than physical CPUs)
power sst_bf.enabled Intel SST-BF (Intel Speed Select Technology - Base frequency) enabled
pstate turbo Set to 'true' if turbo frequencies are enabled in Intel pstate driver, set to 'false' if they have been disabled.
rdt RDTMON Intel RDT Monitoring Technology

RDTCMT Intel Cache Monitoring (CMT)

RDTMBM Intel Memory Bandwidth Monitoring (MBM)

RDTL3CA Intel L3 Cache Allocation Technology

RDTL2CA Intel L2 Cache Allocation Technology

RDTMBA Intel Memory Bandwidth Allocation (MBA) Technology

The (sub-)set of CPUID attributes to publish is configurable via the attributeBlacklist and attributeWhitelist cpuid options of the cpu source. If whitelist is specified, only whitelisted attributes will be published. With blacklist, only blacklisted attributes are filtered out. attributeWhitelist has priority over attributeBlacklist. For examples and more information about configurability, see Configuration Options. By default, the following CPUID flags have been blacklisted: BMI1, BMI2, CLMUL, CMOV, CX16, ERMS, F16C, HTT, LZCNT, MMX, MMXEXT, NX, POPCNT, RDRAND, RDSEED, RDTSCP, SGX, SSE, SSE2, SSE3, SSE4.1, SSE4.2 and SSSE3.

NOTE The cpuid features advertise supported CPU capabilities, that is, a capability might be supported but not enabled.

X86 CPUID Attributes (Partial List)

Attribute Description
ADX Multi-Precision Add-Carry Instruction Extensions (ADX)
AESNI Advanced Encryption Standard (AES) New Instructions (AES-NI)
AVX Advanced Vector Extensions (AVX)
AVX2 Advanced Vector Extensions 2 (AVX2)

Arm CPUID Attribute (Partial List)

Attribute Description
IDIVA Integer divide instructions available in ARM mode
IDIVT Integer divide instructions available in Thumb mode
THUMB Thumb instructions
FASTMUL Fast multiplication
VFP Vector floating point instruction extension (VFP)
VFPv3 Vector floating point extension v3
VFPv4 Vector floating point extension v4
VFPD32 VFP with 32 D-registers
HALF Half-word loads and stores
EDSP DSP extensions
NEON NEON SIMD instructions
LPAE Large Physical Address Extensions

Arm64 CPUID Attribute (Partial List)

Attribute Description
AES Announcing the Advanced Encryption Standard
EVSTRM Event Stream Frequency Features
FPHP Half Precision(16bit) Floating Point Data Processing Instructions
ASIMDHP Half Precision(16bit) Asimd Data Processing Instructions
ATOMICS Atomic Instructions to the A64
ASIMRDM Support for Rounding Double Multiply Add/Subtract
PMULL Optional Cryptographic and CRC32 Instructions
JSCVT Perform Conversion to Match Javascript
DCPOP Persistent Memory Support

Custom Features

The Custom feature source allows the user to define features based on a mix of predefined rules. A rule is provided input witch affects its process of matching for a defined feature.

To aid in making Custom Features clearer, we define a general and a per rule nomenclature, keeping things as consistent as possible.

General Nomenclature & Definitions

Rule        :Represents a matching logic that is used to match on a feature.
Rule Input  :The input a Rule is provided. This determines how a Rule performs the match operation.
Matcher     :A composition of Rules, each Matcher may be composed of at most one instance of each Rule.

Custom Features Format (using the Nomenclature defined above)

- name: <feature name>
  matchOn:
  - <Rule-1>: <Rule-1 Input>
    [<Rule-2>: <Rule-2 Input>]
  - <Matcher-2>
  - ...
  - ...
  - <Matcher-N>
- <custom feature 2>
- ...
- ...
- <custom feature M>

Matching process

Specifying Rules to match on a feature is done by providing a list of Matchers. Each Matcher contains one or more Rules.

Logical OR is performed between Matchers and logical AND is performed between Rules of a given Matcher.

Rules

PciId Rule
Nomenclature
Attribute   :A PCI attribute.
Element     :An identifier of the PCI attribute.

The PciId Rule allows matching the PCI devices in the system on the following Attributes: class,vendor and device. A list of Elements is provided for each Attribute.

Format
pciId :
  class: [<class id>, ...]
  vendor: [<vendor id>,  ...]
  device: [<device id>, ...]

Matching is done by performing a logical OR between Elements of an Attribute and logical AND between the specified Attributes for each PCI device in the system. At least one Attribute must be specified. Missing attributes will not partake in the matching process.

UsbId Rule
Nomenclature
Attribute   :A USB attribute.
Element     :An identifier of the USB attribute.

The UsbId Rule allows matching the USB devices in the system on the following Attributes: class,vendor and device. A list of Elements is provided for each Attribute.

Format
usbId :
  class: [<class id>, ...]
  vendor: [<vendor id>,  ...]
  device: [<device id>, ...]

Matching is done by performing a logical OR between Elements of an Attribute and logical AND between the specified Attributes for each USB device in the system. At least one Attribute must be specified. Missing attributes will not partake in the matching process.

LoadedKMod Rule
Nomenclature
Element     :A kernel module

The LoadedKMod Rule allows matching the loaded kernel modules in the system against a provided list of Elements.

Format
loadedKMod : [<kernel module>, ...]

Matching is done by performing logical AND for each provided Element, i.e the Rule will match if all provided Elements (kernel modules) are loaded in the system.

Example

custom:
  - name: "my.kernel.feature"
    matchOn:
      - loadedKMod: ["kmod1", "kmod2"]
  - name: "my.pci.feature"
    matchOn:
      - pciId:
          vendor: ["15b3"]
          device: ["1014", "1017"]
  - name: "my.usb.feature"
    matchOn:
      - usbId:
          vendor: ["1d6b"]
          device: ["0003"]
  - name: "my.combined.feature"
    matchOn:
      - loadedKMod : ["vendor_kmod1", "vendor_kmod2"]
        pciId:
          vendor: ["15b3"]
          device: ["1014", "1017"] 
  - name: "my.accumulated.feature"
    matchOn:
      - loadedKMod : ["some_kmod1", "some_kmod2"]
      - pciId:
          vendor: ["15b3"]
          device: ["1014", "1017"]

In the example above:

  • A node would contain the label: feature.node.kubernetes.io/custom-my.kernel.feature=true if the node has kmod1 AND kmod2 kernel modules loaded.
  • A node would contain the label: feature.node.kubernetes.io/custom-my.pci.feature=true if the node contains a PCI device with a PCI vendor ID of 15b3 AND PCI device ID of 1014 OR 1017.
  • A node would contain the label: feature.node.kubernetes.io/custom-my.usb.feature=true if the node contains a USB device with a USB vendor ID of 1d6b AND USB device ID of 0003.
  • A node would contain the label: feature.node.kubernetes.io/custom-my.combined.feature=true if vendor_kmod1 AND vendor_kmod2 kernel modules are loaded AND the node contains a PCI device with a PCI vendor ID of 15b3 AND PCI device ID of 1014 or 1017.
  • A node would contain the label: feature.node.kubernetes.io/custom-my.accumulated.feature=true if some_kmod1 AND some_kmod2 kernel modules are loaded OR the node contains a PCI device with a PCI vendor ID of 15b3 AND PCI device ID of 1014 OR 1017.

Statically defined features

Some feature labels which are common and generic are defined statically in the custom feature source. A user may add additional Matchers to these feature labels by defining them in the nfd-worker configuration file.

Feature Attribute Description
rdma capable The node has an RDMA capable Network adapter
rdma enabled The node has the needed RDMA modules loaded to run RDMA traffic

IOMMU Features

Feature name Description
enabled IOMMU is present and enabled in the kernel

Kernel Features

Feature Attribute Description
config <option name> Kernel config option is enabled (set 'y' or 'm').
Default options are NO_HZ, NO_HZ_IDLE, NO_HZ_FULL and PREEMPT
selinux enabled Selinux is enabled on the node
version full Full kernel version as reported by /proc/sys/kernel/osrelease (e.g. '4.5.6-7-g123abcde')

major First component of the kernel version (e.g. '4')

minor Second component of the kernel version (e.g. '5')

revision Third component of the kernel version (e.g. '6')

Kernel config file to use, and, the set of config options to be detected are configurable. See configuration options for more information.

Memory Features

Feature Attribute Description
numa
Multiple memory nodes i.e. NUMA architecture detected
nv present NVDIMM device(s) are present
nv dax NVDIMM region(s) configured in DAX mode are present

Network Features

Feature Attribute Description
sriov capable Single Root Input/Output Virtualization (SR-IOV) enabled Network Interface Card(s) present

configured SR-IOV virtual functions have been configured

PCI Features

Feature Attribute Description
<device label> present PCI device is detected
<device label> sriov.capable Single Root Input/Output Virtualization (SR-IOV) enabled PCI device present

<device label> is composed of raw PCI IDs, separated by underscores. The set of fields used in <device label> is configurable, valid fields being class, vendor, device, subsystem_vendor and subsystem_device. Defaults are class and vendor. An example label using the default label fields:

feature.node.kubernetes.io/pci-1200_8086.present=true

Also the set of PCI device classes that the feature source detects is configurable. By default, device classes (0x)03, (0x)0b40 and (0x)12, i.e. GPUs, co-processors and accelerator cards are detected.

USB Features

Feature Attribute Description
<device label> present USB device is detected

<device label> is composed of raw USB IDs, separated by underscores. The set of fields used in <device label> is configurable, valid fields being class, vendor, and device. Defaults are class, vendor and device. An example label using the default label fields:

feature.node.kubernetes.io/usb-fe_1a6e_089a.present=true

See configuration options for more information on NFD config.

Storage Features

Feature name Description
nonrotationaldisk Non-rotational disk, like SSD, is present in the node

System Features

Feature Attribute Description
os_release ID Operating system identifier

VERSION_ID Operating system version identifier (e.g. '6.7')

VERSION_ID.major First component of the OS version id (e.g. '6')

VERSION_ID.minor Second component of the OS version id (e.g. '7')

Feature Detector Hooks (User-specific Features)

NFD has a special feature source named local which is designed for getting the labels from user-specific feature detector. It provides a mechanism for users to implement custom feature sources in a pluggable way, without modifying nfd source code or Docker images. The local feature source can be used to advertise new user-specific features, and, for overriding labels created by the other feature sources.

The local feature source gets its labels by two different ways:

  • It tries to execute files found under /etc/kubernetes/node-feature-discovery/source.d/ directory. The hook files must be executable and they are supposed to print all discovered features in stdout, one per line. With ELF binaries static linking is recommended as the selection of system libraries available in the NFD release image is very limited. Other runtimes currently supported by the NFD stock image are bash and perl.
  • It reads files found under /etc/kubernetes/node-feature-discovery/features.d/ directory. The file content is expected to be similar to the hook output (described above).

These directories must be available inside the Docker image so Volumes and VolumeMounts must be used if standard NFD images are used. The given template files mount by default the source.d and the features.d directories respectively from /etc/kubernetes/node-feature-discovery/source.d/ and /etc/kubernetes/node-feature-discovery/features.d/ from the host. You should update them to match your needs.

In both cases, the labels can be binary or non binary, using either <name> or <name>=<value> format.

Unlike the other feature sources, the name of the file, instead of the name of the feature source (that would be local in this case), is used as a prefix in the label name, normally. However, if the <name> of the label starts with a slash (/) it is used as the label name as is, without any additional prefix. This makes it possible for the user to fully control the feature label names, e.g. for overriding labels created by other feature sources.

You can also override the default namespace of your labels using this format: <namespace>/<name>[=<value>]. You must whitelist your namespace using the --extra-label-ns option on the master. In this case, the name of the file will not be added to the label name. For example, if you want to add the label my.namespace.org/my-label=value, your hook output or file must contains my.namespace.org/my-label=value and you must add --extra-label-ns=my.namespace.org on the master command line.

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

Injecting Labels from Other Pods

One use case for the hooks and/or feature files is detecting features in other Pods outside NFD, e.g. in Kubernetes device plugins. It is possible to mount the source.d and/or features.d directories common with the NFD Pod and deploy the custom hooks/features there. NFD will periodically scan the directories and run any hooks and read any feature files it finds. The example nfd-worker deployment template contains hostPath mounts for sources.d and features.d directories. By using the same mounts in the secondary Pod (e.g. device plugin) you have created a shared area for delivering hooks and feature files to NFD.

A Hook Example

User has a shell script /etc/kubernetes/node-feature-discovery/source.d/my-source which has the following stdout output:

MY_FEATURE_1
MY_FEATURE_2=myvalue
/override_source-OVERRIDE_BOOL
/override_source-OVERRIDE_VALUE=123
override.namespace/value=456

which, in turn, will translate into the following node labels:

feature.node.kubernetes.io/my-source-MY_FEATURE_1=true
feature.node.kubernetes.io/my-source-MY_FEATURE_2=myvalue
feature.node.kubernetes.io/override_source-OVERRIDE_BOOL=true
feature.node.kubernetes.io/override_source-OVERRIDE_VALUE=123
override.namespace/value=456

A File Example

User has a file /etc/kubernetes/node-feature-discovery/features.d/my-source which contains the following lines:

MY_FEATURE_1
MY_FEATURE_2=myvalue
/override_source-OVERRIDE_BOOL
/override_source-OVERRIDE_VALUE=123
override.namespace/value=456

which, in turn, will translate into the following node labels:

feature.node.kubernetes.io/my-source-MY_FEATURE_1=true
feature.node.kubernetes.io/my-source-MY_FEATURE_2=myvalue
feature.node.kubernetes.io/override_source-OVERRIDE_BOOL=true
feature.node.kubernetes.io/override_source-OVERRIDE_VALUE=123
override.namespace/value=456

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

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! Be careful when creating and/or updating hook or feature files while NFD is running. In order to avoid race conditions you should write into a temporary file (outside the source.d and features.d directories), and, atomically create/update the original file by doing a filesystem move operation.

Extended resources (experimental)

This feature is experimental and by no means a replacement for the usage of device plugins.

Labels which have integer values, can be promoted to Kubernetes extended resources by listing them to the master --resource-labels command line flag. These labels won't then show in the node label section, they will appear only as extended resources.

An example use-case for the extended resources could be based on a hook which creates a label for the node SGX EPC memory section size. By giving the name of that label in the --resource-labels flag, that value will then turn into an extended resource of the node, allowing PODs to request that resource and the Kubernetes scheduler to schedule such PODs to only those nodes which have a sufficient capacity of said resource left.

Similar to labels, the default namespace feature.node.kubernetes.io is automatically prefixed to the extended resource, if the promoted label doesn't have a namespace.

Example usage of the command line arguments, using a new namespace: nfd-master --resource-labels=my_source-my.feature,sgx.some.ns/epc --extra-label-ns=sgx.some.ns

The above would result in following extended resources provided that related labels exist:

  sgx.some.ns/epc: <label value>
  feature.node.kubernetes.io/my_source-my.feature: <label value>

Getting started

For a stable version with ready-built images see the latest released version (release notes).

If you want to use the latest development version (master branch) you need to build your own custom image.

System requirements

  1. Linux (x86_64/Arm64/Arm)
  2. kubectl (properly set up and configured to work with your Kubernetes cluster)
  3. Docker (only required to build and push docker images)

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.

You can use the template spec provided to deploy nfd-master, or use nfd-master.yaml generated by Makefile. The latter includes image: and namespace: definitions that match the latest built image. Example:

make IMAGE_TAG=<IMAGE_TAG>
docker push <IMAGE_TAG>
kubectl create -f nfd-master.yaml

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. There is an example spec (nfd-worker-daemonset.yaml.template) that can be used as a template, or, as is when just trying out the service. Similarly to nfd-master above, the Makefile also generates nfd-worker-daemonset.yaml from the template that you can use to deploy the latest image. Example:

make IMAGE_TAG=<IMAGE_TAG>
docker push <IMAGE_TAG>
kubectl create -f nfd-worker-daemonset.yaml

Nfd-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.

Feature discovery can alternatively be configured as a one-shot job. There is an example script in this repo that demonstrates how to deploy the job in the cluster.

./label-nodes.sh [<IMAGE_TAG>]

The label-nodes.sh script tries to launch as many jobs as there are Ready nodes. Note that this approach does not guarantee running once on every node. For example, if some node is tainted NoSchedule or fails to start a job for some other reason, then some other node will run extra job instance(s) to satisfy the request and the tainted/failed node does not get labeled.

nfd-master and nfd-worker in the same Pod

You can also run nfd-master and nfd-worker inside a single pod (skip the sed part if running the latest released version):

sed -E s',^(\s*)image:.+$,\1image: <YOUR_IMAGE_REPO>:<YOUR_IMAGE_TAG>,' nfd-daemonset-combined.yaml.template > nfd-daemonset-combined.yaml
kubectl apply -f nfd-daemonset-combined.yaml

Similar to the nfd-worker setup above, this creates a DaemonSet that schedules an NFD Pod an all worker nodes, with the difference that the Pod also also contains an nfd-master instance. In this case no nfd-master service is run on the master node(s), but, the worker nodes are able to label themselves.

This may be desirable e.g. in single-node setups.

TLS authentication

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) of its certificate. This means that each nfd-worker requires a individual node-specific TLS certificate.

Usage demo

asciicast

Configuration options

Nfd-worker supports 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 on each labeling pass (determined by --sleep-interval) which makes run-time re-configuration of nfd-worker possible.

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. For example, create a config map using the example config as a template:

cp nfd-worker.conf.example nfd-worker.conf
vim nfd-worker.conf  # edit the configuration
kubectl create configmap nfd-worker-config --from-file=nfd-worker.conf

Then, configure Volumes and VolumeMounts in the Pod spec (just the relevant snippets shown below):

...
  containers:
      volumeMounts:
        - name: nfd-worker-config
          mountPath: "/etc/kubernetes/node-feature-discovery/"
...
  volumes:
    - name: nfd-worker-config
      configMap:
        name: nfd-worker-config
...

You could also use other types of volumes, of course. That is, hostPath if different config for different nodes would be required, for example.

The (empty-by-default) example config is used as a config in the NFD Docker image. Thus, this can be used as a default configuration in custom-built images.

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.

Currently, the only available configuration options are related to the CPU, PCI and Kernel feature sources.

Building from source

Download the source code:

git clone https://github.com/kubernetes-sigs/node-feature-discovery

Build the container image:
See customizing the build below for altering the container image registry, for example.

cd <project-root>
make

Push the container image:
Optional, this example with Docker.

docker push <IMAGE_TAG>

Change the job spec to use your custom image (optional):

To use your published image from the step above instead of the k8s.gcr.io/nfd/node-feature-discovery image, edit image attribute in the spec template(s) to the new location (<registry-name>/<image-name>[:<version>]).

Customizing the Build

There are several Makefile variables that control the build process and the name of the resulting container image.

Variable Description Default value
IMAGE_BUILD_CMD Command to build the image docker build
IMAGE_BUILD_EXTRA_OPTS Extra options to pass to build command empty
IMAGE_PUSH_CMD Command to push the image to remote registry docker push
IMAGE_REGISTRY Container image registry to use k8s.gcr.io/nfd
IMAGE_NAME Container image name node-feature-discovery
IMAGE_TAG_NAME Container image tag name <nfd version>
IMAGE_REPO Container image repository to use <IMAGE_REGISTRY>/<IMAGE_NAME>
IMAGE_TAG Full image:tag to tag the image with <IMAGE_REPO>/<IMAGE_NAME>
K8S_NAMESPACE nfd-master and nfd-worker namespace kube-system
KUBECONFIG Kubeconfig for running e2e-tests empty
E2E_TEST_CONFIG Parameterization file of e2e-tests (see example) empty

For example, to use a custom registry:

make IMAGE_REGISTRY=<my custom registry uri>

Or to specify a build tool different from Docker:

make IMAGE_BUILD_CMD="buildah bud"

Testing

Unit tests are automatically run as part of the container image build. You can also run them manually in the source code tree by simply running:

make test

End-to-end tests are built on top of the e2e test framework of Kubernetes, and, they required a cluster to run them on. For running the tests on your test cluster you need to specify the kubeconfig to be used:

make e2e-test KUBECONFIG=$HOME/.kube/config

Targeting Nodes with Specific Features

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.

Node Annotations

NFD annotates nodes it is running on:

Annotation Description
nfd.node.kubernetes.io/master.version Version of the nfd-master instance running on the node. Informative use only.
nfd.node.kubernetes.io/worker.version Version of the nfd-worker instance running on the node. Informative use only.
nfd.node.kubernetes.io/feature-labels Comma-separated list of node labels managed by NFD. NFD uses this internally so must not be edited by users.
nfd.node.kubernetes.io/extended-resources Comma-separated list of node extended resources managed by NFD. NFD uses this internally so must not be edited by users.

Unapplicable annotations are not created, i.e. for example master.version is only created on nodes running nfd-master.

References

Github issues

Design proposal

Governance

This is a SIG-node subproject, hosted under the Kubernetes SIGs organization in Github. The project was established in 2016 as a Kubernetes Incubator project and migrated to Kubernetes SIGs in 2018.

License

This is open source software released under the Apache 2.0 License.

Demo

A demo on the benefits of using node feature discovery can be found in demo.