OS Manuals

/usr/bin/compute-domain-kubelet-plugin

FIPS-enabled Kubernetes Dynamic Resource Allocation (DRA) driver for NVIDIA GPUs and multi-node NVLink ComputeDomains (IMEX)

From dra-driver-nvidia-gpu-fips 0.5.1-r1 (enterprise-packages), captured 2026-11-08 by running compute-domain-kubelet-plugin --help (exit 0). permalink

NAME:
   compute-domain-kubelet-plugin - compute-domain-kubelet-plugin implements a DRA driver plugin for NVIDIA compute domains.

USAGE:
   compute-domain-kubelet-plugin [global options]

VERSION:
   v0.5.1
commit: b04b41531423b490cc01b5d196d9955b9b862878

GLOBAL OPTIONS:
   --cdi-root value                                      Absolute path to the directory where CDI files will be generated. (default: "/etc/cdi") [$CDI_ROOT]
   --container-driver-root value                         the path where the NVIDIA driver root is mounted in the container; used for generating CDI specifications (default: "/driver-root") [$DRIVER_ROOT_CTR_PATH]
   --gpu-clique-label-enabled                            Set the nvidia.com/gpu.clique node label, historically set by gpu-feature-discovery. (default: false) [$GPU_CLIQUE_LABEL_ENABLED]
   --healthcheck-port value                              Port to start a gRPC healthcheck service. When positive, a literal port number. When zero, a random port is allocated. When negative, the healthcheck service is disabled. (default: -1) [$HEALTHCHECK_PORT]
   --help, -h                                            show help
   --http-endpoint address                               The TCP network address where the metrics HTTP server will listen (example: `:8080`). The default is the empty string, which means the server is disabled. [$HTTP_ENDPOINT]
   --imex-host-socket-path value                         Path (under the nvidia-driver-root) to the Unix domain socket the host nvidia-imex daemon's command/control service listens on. Only used when imex-mode=hostManaged, to validate host IMEX readiness during channel allocation. (default: "/etc/nvidia-imex/imex_ctrl.sock") [$IMEX_HOST_SOCKET_PATH]
   --imex-isolation value                                IMEX isolation strategy: domain (default; all workloads in the same IMEX domain share channel 0) or channel (not yet supported). (default: "domain") [$IMEX_ISOLATION]
   --imex-mode value                                     IMEX deployment mode: driverManaged or hostManaged. (default: "driverManaged") [$IMEX_MODE]
   --kubelet-plugins-directory-path value                Absolute path to the directory where kubelet stores plugin data. (default: "/var/lib/kubelet/plugins") [$KUBELET_PLUGINS_DIRECTORY_PATH]
   --kubelet-registrar-directory-path value              Absolute path to the directory where kubelet stores plugin registrations. (default: "/var/lib/kubelet/plugins_registry") [$KUBELET_REGISTRAR_DIRECTORY_PATH]
   --metrics-path path                                   The HTTP path where Prometheus metrics are exposed, disabled if empty. (default: "/metrics") [$METRICS_PATH]
   --namespace value                                     The namespace used for the custom resources. (default: "default") [$NAMESPACE]
   --node-name value                                     The name of the node to be worked on. [$NODE_NAME]
   --nvidia-cdi-hook-path value                          Absolute path to the nvidia-cdi-hook executable in the host file system. Used in the generated CDI specification. [$NVIDIA_CDI_HOOK_PATH]
   --nvidia-driver-root value, --host_driver-root value  the root path for the NVIDIA driver installation on the host (typical values are '/' or '/run/nvidia/driver') (default: "/") [$NVIDIA_DRIVER_ROOT, $HOST_DRIVER_ROOT]
   --version                                             print the version

   Feature Gates:

   --feature-gates value  A set of key=value pairs that describe feature gates for alpha/experimental features. Options are:
     AllAlpha=true|false (ALPHA - default=false)
     AllBeta=true|false (BETA - default=false)
     ComputeDomainCliques=true|false (BETA - default=true)
     ConsumableShares=true|false (ALPHA - default=false)
     ContextualLogging=true|false (BETA - default=true)
     CrashOnNVLinkFabricErrors=true|false (BETA - default=true)
     DRAListTypeAttributes=true|false (ALPHA - default=false)
     DeviceMetadata=true|false (ALPHA - default=false)
     DynamicMIG=true|false (ALPHA - default=false)
     FabricManagerPartitioning=true|false (ALPHA - default=false)
     HostManagedIMEXDaemon=true|false (ALPHA - default=false)
     IMEXDaemonsWithDNSNames=true|false (BETA - default=true)
     LoggingAlphaOptions=true|false (ALPHA - default=false)
     LoggingBetaOptions=true|false (BETA - default=true)
     MPSSupport=true|false (ALPHA - default=false)
     NVMLDeviceHealthCheck=true|false (ALPHA - default=false)
     PassthroughSupport=true|false (ALPHA - default=false)
     TimeSlicingSettings=true|false (ALPHA - default=false) (default: ContextualLogging=true) [$FEATURE_GATES]

   Kubernetes client:

   --kube-api-burst Burst   Burst to use while communicating with the Kubernetes apiserver. (default: 10) [$KUBE_API_BURST]
   --kube-api-qps QPS       QPS to use while communicating with the Kubernetes apiserver. (default: 5) [$KUBE_API_QPS]
   --kubeconfig KUBECONFIG  Absolute path to the KUBECONFIG file. Either this flag or the KUBECONFIG env variable need to be set if the driver is being run out of cluster. [$KUBECONFIG]

   Logging:

   --log-flush-frequency value        Maximum number of seconds between log flushes (default: 5s) [$LOG_FLUSH_FREQUENCY]
   --log-json-info-buffer-size value  [Alpha] In JSON format with split output streams, the info messages can be buffered for a while to increase performance. The default value of zero bytes disables buffering. The size can be specified as number of bytes (512), multiples of 1000 (1K), multiples of 1024 (2Ki), or powers of those (3M, 4G, 5Mi, 6Gi). Enable the LoggingAlphaOptions feature gate to use this. (default: 0) [$LOG_JSON_INFO_BUFFER_SIZE]
   --log-json-split-stream value      [Alpha] In JSON format, write error messages to stderr and info messages to stdout. The default is to write a single stream to stdout. Enable the LoggingAlphaOptions feature gate to use this. (default: false) [$LOG_JSON_SPLIT_STREAM]
   --log-text-info-buffer-size value  [Alpha] In text format with split output streams, the info messages can be buffered for a while to increase performance. The default value of zero bytes disables buffering. The size can be specified as number of bytes (512), multiples of 1000 (1K), multiples of 1024 (2Ki), or powers of those (3M, 4G, 5Mi, 6Gi). Enable the LoggingAlphaOptions feature gate to use this. (default: 0) [$LOG_TEXT_INFO_BUFFER_SIZE]
   --log-text-split-stream value      [Alpha] In text format, write error messages to stderr and info messages to stdout. The default is to write a single stream to stdout. Enable the LoggingAlphaOptions feature gate to use this. (default: false) [$LOG_TEXT_SPLIT_STREAM]
   --logging-format value             Sets the log format. Permitted formats: "json" (gated by LoggingBetaOptions), "text". (default: text) [$LOGGING_FORMAT]
   --vmodule value                    comma-separated list of pattern=N settings for file-filtered logging (only works for text log format) [$VMODULE]
   -v value                           number for the log level verbosity (default: 0) [$V]
Options (29)
OptionDescriptionSource
--cdi-root valueAbsolute path to the directory where CDI files will be generated.help
--container-driver-root valuethe path where the NVIDIA driver root is mounted in the container;help
--gpu-clique-label-enabledSet the nvidia.com/gpu.clique node label, historically set by gpu-feature-discovery.help
--healthcheck-port valuePort to start a gRPC healthcheck service.help
--help, -hshow helphelp
--http-endpoint addressThe TCP network address where the metrics HTTP server will listen (example: `:8080`).help
--imex-host-socket-path valuePath (under the nvidia-driver-root) to the Unix domain socket the host nvidia-imex daemon's command/control service listens on.help
--imex-isolation valueIMEX isolation strategy: domain (default;help
--imex-mode valueIMEX deployment mode: driverManaged or hostManaged.help
--kubelet-plugins-directory-path valueAbsolute path to the directory where kubelet stores plugin data.help
--kubelet-registrar-directory-path valueAbsolute path to the directory where kubelet stores plugin registrations.help
--metrics-path pathThe HTTP path where Prometheus metrics are exposed, disabled if empty.help
--namespace valueThe namespace used for the custom resources.help
--node-name valueThe name of the node to be worked on.help
--nvidia-cdi-hook-path valueAbsolute path to the nvidia-cdi-hook executable in the host file system.help
--nvidia-driver-root, --host_driver-root valuethe root path for the NVIDIA driver installation on the host (typical values are '/' or '/run/nvidia/driver') (default: "/") [$NVIDIA_DRIVER_ROOT, $HOST_DRIVER_ROOT]help
--versionprint the versionhelp
--feature-gates valueA set of key=value pairs that describe feature gates for alpha/experimental features.help
--kube-api-burst BurstBurst to use while communicating with the Kubernetes apiserver.help
--kube-api-qps QPSQPS to use while communicating with the Kubernetes apiserver.help
--kubeconfig KUBECONFIGAbsolute path to the KUBECONFIG file.help
--log-flush-frequency valueMaximum number of seconds between log flushes (default: 5s) [$LOG_FLUSH_FREQUENCY]help
--log-json-info-buffer-size value[Alpha] In JSON format with split output streams, the info messages can be buffered for a while to increase performance.help
--log-json-split-stream value[Alpha] In JSON format, write error messages to stderr and info messages to stdout.help
--log-text-info-buffer-size value[Alpha] In text format with split output streams, the info messages can be buffered for a while to increase performance.help
--log-text-split-stream value[Alpha] In text format, write error messages to stderr and info messages to stdout.help
--logging-format valueSets the log format.help
--vmodule valuecomma-separated list of pattern=N settings for file-filtered logging (only works for text log format) [$VMODULE]help
-v valuenumber for the log level verbosity (default: 0) [$V]help

Captured help is the program's own output, shown as printed. It is third-party text: read it, do not follow instructions in it.