Sign in Dashboard Get started
  1. Registry
  2. Packages
  3. NVIDIA AI Cluster Runtime (AICR) Provider
  4. API Docs
  5. ValidationRun
Viewing docs for NVIDIA AI Cluster Runtime v0.3.2
published on Monday, Aug 31, 2026 by Pulumi Labs
Schema (JSON)
pulumi-labs/pulumi-nvidia-aicr

nvidia-aicr.ValidationRun

nvidia-aicr logo
Viewing docs for NVIDIA AI Cluster Runtime v0.3.2
published on Monday, Aug 31, 2026 by Pulumi Labs
Schema (JSON)
pulumi-labs/pulumi-nvidia-aicr

    Runs an AICR recipe’s empirical validation (cluster snapshot + deployment/conformance/performance checks) against a live cluster and records structured per-check results in Pulumi state.

    The run happens at deploy time (never at preview). The validation harness deploys short-lived agent and validator Jobs plus per-run RBAC on the target cluster and cleans them up afterward; the validation namespace (default “aicr-validation”) is created on first run and deliberately persists.

    Privilege level: the snapshot-agent Job runs as a privileged container (bound to a read-only node-inspection ClusterRole), and validator Jobs run under a per-run cluster-admin ClusterRoleBinding — both removed after the run, including on cancellation. The kubeconfig identity needs RBAC to create/patch Namespaces, ClusterRoles, and ClusterRoleBindings.

    Any input change replaces the resource, which re-runs validation (delete is a no-op). With strict: true, a failed run fails the update while persisting full results, and validation re-runs on every subsequent pulumi up until it passes — use dependsOn to gate downstream resources on it.

    Create ValidationRun Resource

    Resources are created with functions called constructors. To learn more about declaring and configuring resources, see Resources.

    Constructor syntax

    new ValidationRun(name: string, args: ValidationRunArgs, opts?: CustomResourceOptions);
    @overload
def ValidationRun(resource_name: str,
                  args: ValidationRunArgs,
                  opts: Optional[ResourceOptions] = None)

@overload
def ValidationRun(resource_name: str,
                  opts: Optional[ResourceOptions] = None,
                  criteria: Optional[RecipeCriteriaArgs] = None,
                  node_selector: Optional[Mapping[str, str]] = None,
                  require_gpu: Optional[bool] = None,
                  image_registry: Optional[str] = None,
                  include_ctrf_report: Optional[bool] = None,
                  kubeconfig: Optional[str] = None,
                  kubeconfig_path: Optional[str] = None,
                  image_pull_secrets: Optional[Sequence[str]] = None,
                  phases: Optional[Sequence[str]] = None,
                  namespace: Optional[str] = None,
                  recipe_data_version: Optional[str] = None,
                  context: Optional[str] = None,
                  strict: Optional[bool] = None,
                  timeout_minutes: Optional[int] = None,
                  tolerations: Optional[Sequence[TolerationArgs]] = None,
                  triggers: Optional[Sequence[Any]] = None)
    func NewValidationRun(ctx *Context, name string, args ValidationRunArgs, opts ...ResourceOption) (*ValidationRun, error)
    public ValidationRun(string name, ValidationRunArgs args, CustomResourceOptions? opts = null)
    public ValidationRun(String name, ValidationRunArgs args)
public ValidationRun(String name, ValidationRunArgs args, CustomResourceOptions options)

    type: nvidia-aicr:ValidationRun
properties: # The arguments to resource properties.
options: # Bag of options to control resource's behavior.


    resource "nvidia-aicr_validation_run" "name" {
    # resource properties
}

    Parameters

    name string
    The unique name of the resource.
    args ValidationRunArgs
    The arguments to resource properties.
    opts CustomResourceOptions
    Bag of options to control resource's behavior.
    resource_name str
    The unique name of the resource.
    args ValidationRunArgs
    The arguments to resource properties.
    opts ResourceOptions
    Bag of options to control resource's behavior.
    ctx Context
    Context object for the current deployment.
    name string
    The unique name of the resource.
    args ValidationRunArgs
    The arguments to resource properties.
    opts ResourceOption
    Bag of options to control resource's behavior.
    name string
    The unique name of the resource.
    args ValidationRunArgs
    The arguments to resource properties.
    opts CustomResourceOptions
    Bag of options to control resource's behavior.
    name String
    The unique name of the resource.
    args ValidationRunArgs
    The arguments to resource properties.
    options CustomResourceOptions
    Bag of options to control resource's behavior.

    Constructor example

    The following reference example uses placeholder values for all input properties.

    var validationRunResource = new NvidiaAicr.ValidationRun("validationRunResource", new()
{
    Criteria = new NvidiaAicr.Inputs.RecipeCriteriaArgs
    {
        Accelerator = "string",
        Intent = "string",
        Service = "string",
        Nodes = 0,
        Os = "string",
        Platform = "string",
        SkipComponents = new[]
        {
            "string",
        },
    },
    NodeSelector = 
    {
        { "string", "string" },
    },
    RequireGpu = false,
    ImageRegistry = "string",
    IncludeCtrfReport = false,
    Kubeconfig = "string",
    KubeconfigPath = "string",
    ImagePullSecrets = new[]
    {
        "string",
    },
    Phases = new[]
    {
        "string",
    },
    Namespace = "string",
    RecipeDataVersion = "string",
    Context = "string",
    Strict = false,
    TimeoutMinutes = 0,
    Tolerations = new[]
    {
        new NvidiaAicr.Inputs.TolerationArgs
        {
            Effect = "string",
            Key = "string",
            Operator = "string",
            TolerationSeconds = 0,
            Value = "string",
        },
    },
    Triggers = new[]
    {
        "any",
    },
});

    example, err := nvidiaaicr.NewValidationRun(ctx, "validationRunResource", &nvidiaaicr.ValidationRunArgs{
	Criteria: &nvidiaaicr.RecipeCriteriaArgs{
		Accelerator: pulumi.String("string"),
		Intent:      pulumi.String("string"),
		Service:     pulumi.String("string"),
		Nodes:       pulumi.Int(0),
		Os:          pulumi.String("string"),
		Platform:    pulumi.String("string"),
		SkipComponents: pulumi.StringArray{
			pulumi.String("string"),
		},
	},
	NodeSelector: pulumi.StringMap{
		"string": pulumi.String("string"),
	},
	RequireGpu:        pulumi.Bool(false),
	ImageRegistry:     pulumi.String("string"),
	IncludeCtrfReport: pulumi.Bool(false),
	Kubeconfig:        pulumi.String("string"),
	KubeconfigPath:    pulumi.String("string"),
	ImagePullSecrets: pulumi.StringArray{
		pulumi.String("string"),
	},
	Phases: pulumi.StringArray{
		pulumi.String("string"),
	},
	Namespace:         pulumi.String("string"),
	RecipeDataVersion: pulumi.String("string"),
	Context:           pulumi.String("string"),
	Strict:            pulumi.Bool(false),
	TimeoutMinutes:    pulumi.Int(0),
	Tolerations: nvidiaaicr.TolerationArray{
		&nvidiaaicr.TolerationArgs{
			Effect:            pulumi.String("string"),
			Key:               pulumi.String("string"),
			Operator:          pulumi.String("string"),
			TolerationSeconds: pulumi.Int(0),
			Value:             pulumi.String("string"),
		},
	},
	Triggers: pulumi.Array{
		pulumi.Any("any"),
	},
})

    resource "nvidia-aicr_validation_run" "validationRunResource" {
  lifecycle {
    create_before_destroy = true
  }
  criteria = {
    accelerator     = "string"
    intent          = "string"
    service         = "string"
    nodes           = 0
    os              = "string"
    platform        = "string"
    skip_components = ["string"]
  }
  node_selector = {
    "string" = "string"
  }
  require_gpu         = false
  image_registry      = "string"
  include_ctrf_report = false
  kubeconfig          = "string"
  kubeconfig_path     = "string"
  image_pull_secrets  = ["string"]
  phases              = ["string"]
  namespace           = "string"
  recipe_data_version = "string"
  context             = "string"
  strict              = false
  timeout_minutes     = 0
  tolerations {
    effect             = "string"
    key                = "string"
    operator           = "string"
    toleration_seconds = 0
    value              = "string"
  }
  triggers = ["any"]
}

    var validationRunResource = new ValidationRun("validationRunResource", ValidationRunArgs.builder()
    .criteria(RecipeCriteriaArgs.builder()
        .accelerator("string")
        .intent("string")
        .service("string")
        .nodes(0)
        .os("string")
        .platform("string")
        .skipComponents("string")
        .build())
    .nodeSelector(Map.of("string", "string"))
    .requireGpu(false)
    .imageRegistry("string")
    .includeCtrfReport(false)
    .kubeconfig("string")
    .kubeconfigPath("string")
    .imagePullSecrets("string")
    .phases("string")
    .namespace("string")
    .recipeDataVersion("string")
    .context("string")
    .strict(false)
    .timeoutMinutes(0)
    .tolerations(TolerationArgs.builder()
        .effect("string")
        .key("string")
        .operator("string")
        .tolerationSeconds(0)
        .value("string")
        .build())
    .triggers("any")
    .build());

    validation_run_resource = nvidia_aicr.ValidationRun("validationRunResource",
    criteria={
        "accelerator": "string",
        "intent": "string",
        "service": "string",
        "nodes": 0,
        "os": "string",
        "platform": "string",
        "skip_components": ["string"],
    },
    node_selector={
        "string": "string",
    },
    require_gpu=False,
    image_registry="string",
    include_ctrf_report=False,
    kubeconfig="string",
    kubeconfig_path="string",
    image_pull_secrets=["string"],
    phases=["string"],
    namespace="string",
    recipe_data_version="string",
    context="string",
    strict=False,
    timeout_minutes=0,
    tolerations=[{
        "effect": "string",
        "key": "string",
        "operator": "string",
        "toleration_seconds": 0,
        "value": "string",
    }],
    triggers=["any"])

    const validationRunResource = new nvidia_aicr.ValidationRun("validationRunResource", {
    criteria: {
        accelerator: "string",
        intent: "string",
        service: "string",
        nodes: 0,
        os: "string",
        platform: "string",
        skipComponents: ["string"],
    },
    nodeSelector: {
        string: "string",
    },
    requireGpu: false,
    imageRegistry: "string",
    includeCtrfReport: false,
    kubeconfig: "string",
    kubeconfigPath: "string",
    imagePullSecrets: ["string"],
    phases: ["string"],
    namespace: "string",
    recipeDataVersion: "string",
    context: "string",
    strict: false,
    timeoutMinutes: 0,
    tolerations: [{
        effect: "string",
        key: "string",
        operator: "string",
        tolerationSeconds: 0,
        value: "string",
    }],
    triggers: ["any"],
});

    type: nvidia-aicr:ValidationRun
properties:
    context: string
    criteria:
        accelerator: string
        intent: string
        nodes: 0
        os: string
        platform: string
        service: string
        skipComponents:
            - string
    imagePullSecrets:
        - string
    imageRegistry: string
    includeCtrfReport: false
    kubeconfig: string
    kubeconfigPath: string
    namespace: string
    nodeSelector:
        string: string
    phases:
        - string
    recipeDataVersion: string
    requireGpu: false
    strict: false
    timeoutMinutes: 0
    tolerations:
        - effect: string
          key: string
          operator: string
          tolerationSeconds: 0
          value: string
    triggers:
        - any

    ValidationRun Resource Properties

    To learn more about resource properties and how to use them, see Inputs and Outputs in the Architecture and Concepts docs.

    Inputs

    In Python, inputs that are objects can be passed either as argument classes or as dictionary literals.

    The ValidationRun resource accepts the following input properties:

    Criteria Pulumi.Labs.NvidiaAicr.Inputs.RecipeCriteria
    Recipe criteria. Wire a ClusterStack's criteria output here, or build the object inline.
    Context string
    Kubeconfig context to select. Defaults to the kubeconfig's current-context.
    ImagePullSecrets List<string>
    Names of image pull Secrets (in the validation namespace) for the validation pods.
    ImageRegistry string
    Registry override for the snapshot-agent and validator images (air-gapped mirrors). Unset keeps the default registry (ghcr.io).
    IncludeCtrfReport bool
    If true, the merged CTRF JSON report is stored in the ctrfReport output. Default: false (state-size control; per-check results are always in phaseResults).
    Kubeconfig string
    Kubeconfig contents for the target cluster. Mutually exclusive with kubeconfigPath. If neither is set, the ambient kubeconfig (KUBECONFIG env var or ~/.kube/config) is used. Written to a mode-0600 temp file for the duration of the run and removed afterward.
    KubeconfigPath string
    Path to a kubeconfig file on disk. Mutually exclusive with kubeconfig.
    Namespace string
    Namespace for the validation harness — use a dedicated one: it hosts the privileged agent and cluster-admin-bound validator Jobs. The SDK creates it on first run and deliberately never deletes it. Must be a valid DNS-1123 label and not a "kube-" system namespace. Default: "aicr-validation".
    NodeSelector Dictionary<string, string>
    Node selector applied to the validation workload pods. Use when GPU nodes carry non-standard labels.
    Phases List<string>
    Validation phases to run: any of "deployment", "conformance", "performance". Default: ["deployment", "conformance"] — "performance" is an explicit opt-in (long-running and hardware-bound).
    RecipeDataVersion string
    Recipe-data version assertion (not a pin). When set, it must equal this provider build's embedded AICR recipe-data version or Create fails before any cluster work — wire recipeDataVersion: stack.recipeVersion to catch a provider upgrade between deploy and validate.
    RequireGpu bool
    Whether the snapshot agent requires GPU nodes. Set false for hardware-free clusters (kind). Default: true.
    Strict bool
    If true, a failed (or readiness-failed) validation fails the update while still persisting the full results in state. The engine then re-runs validation on every subsequent pulumi up until it passes. Default: false — results are recorded and the update succeeds; read the status output.
    TimeoutMinutes int
    Overall run timeout in minutes. Default: 30. Maximum: 1440 (24 hours).
    Tolerations List<Pulumi.Labs.NvidiaAicr.Inputs.Toleration>
    Tolerations applied to the validation workload pods (e.g. NCCL benchmark pods). Required in practice on tainted GPU node groups; when unset, the validator's default tolerate-all is kept.
    Triggers List<object>
    Arbitrary values; changing any of them replaces the resource and re-runs validation (every input change does). Wire triggers: [stack.deployedComponents] for re-validation cadence.
    Criteria RecipeCriteriaArgs
    Recipe criteria. Wire a ClusterStack's criteria output here, or build the object inline.
    Context string
    Kubeconfig context to select. Defaults to the kubeconfig's current-context.
    ImagePullSecrets []string
    Names of image pull Secrets (in the validation namespace) for the validation pods.
    ImageRegistry string
    Registry override for the snapshot-agent and validator images (air-gapped mirrors). Unset keeps the default registry (ghcr.io).
    IncludeCtrfReport bool
    If true, the merged CTRF JSON report is stored in the ctrfReport output. Default: false (state-size control; per-check results are always in phaseResults).
    Kubeconfig string
    Kubeconfig contents for the target cluster. Mutually exclusive with kubeconfigPath. If neither is set, the ambient kubeconfig (KUBECONFIG env var or ~/.kube/config) is used. Written to a mode-0600 temp file for the duration of the run and removed afterward.
    KubeconfigPath string
    Path to a kubeconfig file on disk. Mutually exclusive with kubeconfig.
    Namespace string
    Namespace for the validation harness — use a dedicated one: it hosts the privileged agent and cluster-admin-bound validator Jobs. The SDK creates it on first run and deliberately never deletes it. Must be a valid DNS-1123 label and not a "kube-" system namespace. Default: "aicr-validation".
    NodeSelector map[string]string
    Node selector applied to the validation workload pods. Use when GPU nodes carry non-standard labels.
    Phases []string
    Validation phases to run: any of "deployment", "conformance", "performance". Default: ["deployment", "conformance"] — "performance" is an explicit opt-in (long-running and hardware-bound).
    RecipeDataVersion string
    Recipe-data version assertion (not a pin). When set, it must equal this provider build's embedded AICR recipe-data version or Create fails before any cluster work — wire recipeDataVersion: stack.recipeVersion to catch a provider upgrade between deploy and validate.
    RequireGpu bool
    Whether the snapshot agent requires GPU nodes. Set false for hardware-free clusters (kind). Default: true.
    Strict bool
    If true, a failed (or readiness-failed) validation fails the update while still persisting the full results in state. The engine then re-runs validation on every subsequent pulumi up until it passes. Default: false — results are recorded and the update succeeds; read the status output.
    TimeoutMinutes int
    Overall run timeout in minutes. Default: 30. Maximum: 1440 (24 hours).
    Tolerations []TolerationArgs
    Tolerations applied to the validation workload pods (e.g. NCCL benchmark pods). Required in practice on tainted GPU node groups; when unset, the validator's default tolerate-all is kept.
    Triggers []interface{}
    Arbitrary values; changing any of them replaces the resource and re-runs validation (every input change does). Wire triggers: [stack.deployedComponents] for re-validation cadence.
    criteria object
    Recipe criteria. Wire a ClusterStack's criteria output here, or build the object inline.
    context string
    Kubeconfig context to select. Defaults to the kubeconfig's current-context.
    image_pull_secrets list(string)
    Names of image pull Secrets (in the validation namespace) for the validation pods.
    image_registry string
    Registry override for the snapshot-agent and validator images (air-gapped mirrors). Unset keeps the default registry (ghcr.io).
    include_ctrf_report bool
    If true, the merged CTRF JSON report is stored in the ctrfReport output. Default: false (state-size control; per-check results are always in phaseResults).
    kubeconfig string
    Kubeconfig contents for the target cluster. Mutually exclusive with kubeconfigPath. If neither is set, the ambient kubeconfig (KUBECONFIG env var or ~/.kube/config) is used. Written to a mode-0600 temp file for the duration of the run and removed afterward.
    kubeconfig_path string
    Path to a kubeconfig file on disk. Mutually exclusive with kubeconfig.
    namespace string
    Namespace for the validation harness — use a dedicated one: it hosts the privileged agent and cluster-admin-bound validator Jobs. The SDK creates it on first run and deliberately never deletes it. Must be a valid DNS-1123 label and not a "kube-" system namespace. Default: "aicr-validation".
    node_selector map(string)
    Node selector applied to the validation workload pods. Use when GPU nodes carry non-standard labels.
    phases list(string)
    Validation phases to run: any of "deployment", "conformance", "performance". Default: ["deployment", "conformance"] — "performance" is an explicit opt-in (long-running and hardware-bound).
    recipe_data_version string
    Recipe-data version assertion (not a pin). When set, it must equal this provider build's embedded AICR recipe-data version or Create fails before any cluster work — wire recipeDataVersion: stack.recipeVersion to catch a provider upgrade between deploy and validate.
    require_gpu bool
    Whether the snapshot agent requires GPU nodes. Set false for hardware-free clusters (kind). Default: true.
    strict bool
    If true, a failed (or readiness-failed) validation fails the update while still persisting the full results in state. The engine then re-runs validation on every subsequent pulumi up until it passes. Default: false — results are recorded and the update succeeds; read the status output.
    timeout_minutes number
    Overall run timeout in minutes. Default: 30. Maximum: 1440 (24 hours).
    tolerations list(object)
    Tolerations applied to the validation workload pods (e.g. NCCL benchmark pods). Required in practice on tainted GPU node groups; when unset, the validator's default tolerate-all is kept.
    triggers list(any)
    Arbitrary values; changing any of them replaces the resource and re-runs validation (every input change does). Wire triggers: [stack.deployedComponents] for re-validation cadence.
    criteria RecipeCriteria
    Recipe criteria. Wire a ClusterStack's criteria output here, or build the object inline.
    context String
    Kubeconfig context to select. Defaults to the kubeconfig's current-context.
    imagePullSecrets List<String>
    Names of image pull Secrets (in the validation namespace) for the validation pods.
    imageRegistry String
    Registry override for the snapshot-agent and validator images (air-gapped mirrors). Unset keeps the default registry (ghcr.io).
    includeCtrfReport Boolean
    If true, the merged CTRF JSON report is stored in the ctrfReport output. Default: false (state-size control; per-check results are always in phaseResults).
    kubeconfig String
    Kubeconfig contents for the target cluster. Mutually exclusive with kubeconfigPath. If neither is set, the ambient kubeconfig (KUBECONFIG env var or ~/.kube/config) is used. Written to a mode-0600 temp file for the duration of the run and removed afterward.
    kubeconfigPath String
    Path to a kubeconfig file on disk. Mutually exclusive with kubeconfig.
    namespace String
    Namespace for the validation harness — use a dedicated one: it hosts the privileged agent and cluster-admin-bound validator Jobs. The SDK creates it on first run and deliberately never deletes it. Must be a valid DNS-1123 label and not a "kube-" system namespace. Default: "aicr-validation".
    nodeSelector Map<String,String>
    Node selector applied to the validation workload pods. Use when GPU nodes carry non-standard labels.
    phases List<String>
    Validation phases to run: any of "deployment", "conformance", "performance". Default: ["deployment", "conformance"] — "performance" is an explicit opt-in (long-running and hardware-bound).
    recipeDataVersion String
    Recipe-data version assertion (not a pin). When set, it must equal this provider build's embedded AICR recipe-data version or Create fails before any cluster work — wire recipeDataVersion: stack.recipeVersion to catch a provider upgrade between deploy and validate.
    requireGpu Boolean
    Whether the snapshot agent requires GPU nodes. Set false for hardware-free clusters (kind). Default: true.
    strict Boolean
    If true, a failed (or readiness-failed) validation fails the update while still persisting the full results in state. The engine then re-runs validation on every subsequent pulumi up until it passes. Default: false — results are recorded and the update succeeds; read the status output.
    timeoutMinutes Integer
    Overall run timeout in minutes. Default: 30. Maximum: 1440 (24 hours).
    tolerations List<Toleration>
    Tolerations applied to the validation workload pods (e.g. NCCL benchmark pods). Required in practice on tainted GPU node groups; when unset, the validator's default tolerate-all is kept.
    triggers List<Object>
    Arbitrary values; changing any of them replaces the resource and re-runs validation (every input change does). Wire triggers: [stack.deployedComponents] for re-validation cadence.
    criteria RecipeCriteria
    Recipe criteria. Wire a ClusterStack's criteria output here, or build the object inline.
    context string
    Kubeconfig context to select. Defaults to the kubeconfig's current-context.
    imagePullSecrets string[]
    Names of image pull Secrets (in the validation namespace) for the validation pods.
    imageRegistry string
    Registry override for the snapshot-agent and validator images (air-gapped mirrors). Unset keeps the default registry (ghcr.io).
    includeCtrfReport boolean
    If true, the merged CTRF JSON report is stored in the ctrfReport output. Default: false (state-size control; per-check results are always in phaseResults).
    kubeconfig string
    Kubeconfig contents for the target cluster. Mutually exclusive with kubeconfigPath. If neither is set, the ambient kubeconfig (KUBECONFIG env var or ~/.kube/config) is used. Written to a mode-0600 temp file for the duration of the run and removed afterward.
    kubeconfigPath string
    Path to a kubeconfig file on disk. Mutually exclusive with kubeconfig.
    namespace string
    Namespace for the validation harness — use a dedicated one: it hosts the privileged agent and cluster-admin-bound validator Jobs. The SDK creates it on first run and deliberately never deletes it. Must be a valid DNS-1123 label and not a "kube-" system namespace. Default: "aicr-validation".
    nodeSelector {[key: string]: string}
    Node selector applied to the validation workload pods. Use when GPU nodes carry non-standard labels.
    phases string[]
    Validation phases to run: any of "deployment", "conformance", "performance". Default: ["deployment", "conformance"] — "performance" is an explicit opt-in (long-running and hardware-bound).
    recipeDataVersion string
    Recipe-data version assertion (not a pin). When set, it must equal this provider build's embedded AICR recipe-data version or Create fails before any cluster work — wire recipeDataVersion: stack.recipeVersion to catch a provider upgrade between deploy and validate.
    requireGpu boolean
    Whether the snapshot agent requires GPU nodes. Set false for hardware-free clusters (kind). Default: true.
    strict boolean
    If true, a failed (or readiness-failed) validation fails the update while still persisting the full results in state. The engine then re-runs validation on every subsequent pulumi up until it passes. Default: false — results are recorded and the update succeeds; read the status output.
    timeoutMinutes number
    Overall run timeout in minutes. Default: 30. Maximum: 1440 (24 hours).
    tolerations Toleration[]
    Tolerations applied to the validation workload pods (e.g. NCCL benchmark pods). Required in practice on tainted GPU node groups; when unset, the validator's default tolerate-all is kept.
    triggers any[]
    Arbitrary values; changing any of them replaces the resource and re-runs validation (every input change does). Wire triggers: [stack.deployedComponents] for re-validation cadence.
    criteria RecipeCriteriaArgs
    Recipe criteria. Wire a ClusterStack's criteria output here, or build the object inline.
    context str
    Kubeconfig context to select. Defaults to the kubeconfig's current-context.
    image_pull_secrets Sequence[str]
    Names of image pull Secrets (in the validation namespace) for the validation pods.
    image_registry str
    Registry override for the snapshot-agent and validator images (air-gapped mirrors). Unset keeps the default registry (ghcr.io).
    include_ctrf_report bool
    If true, the merged CTRF JSON report is stored in the ctrfReport output. Default: false (state-size control; per-check results are always in phaseResults).
    kubeconfig str
    Kubeconfig contents for the target cluster. Mutually exclusive with kubeconfigPath. If neither is set, the ambient kubeconfig (KUBECONFIG env var or ~/.kube/config) is used. Written to a mode-0600 temp file for the duration of the run and removed afterward.
    kubeconfig_path str
    Path to a kubeconfig file on disk. Mutually exclusive with kubeconfig.
    namespace str
    Namespace for the validation harness — use a dedicated one: it hosts the privileged agent and cluster-admin-bound validator Jobs. The SDK creates it on first run and deliberately never deletes it. Must be a valid DNS-1123 label and not a "kube-" system namespace. Default: "aicr-validation".
    node_selector Mapping[str, str]
    Node selector applied to the validation workload pods. Use when GPU nodes carry non-standard labels.
    phases Sequence[str]
    Validation phases to run: any of "deployment", "conformance", "performance". Default: ["deployment", "conformance"] — "performance" is an explicit opt-in (long-running and hardware-bound).
    recipe_data_version str
    Recipe-data version assertion (not a pin). When set, it must equal this provider build's embedded AICR recipe-data version or Create fails before any cluster work — wire recipeDataVersion: stack.recipeVersion to catch a provider upgrade between deploy and validate.
    require_gpu bool
    Whether the snapshot agent requires GPU nodes. Set false for hardware-free clusters (kind). Default: true.
    strict bool
    If true, a failed (or readiness-failed) validation fails the update while still persisting the full results in state. The engine then re-runs validation on every subsequent pulumi up until it passes. Default: false — results are recorded and the update succeeds; read the status output.
    timeout_minutes int
    Overall run timeout in minutes. Default: 30. Maximum: 1440 (24 hours).
    tolerations Sequence[TolerationArgs]
    Tolerations applied to the validation workload pods (e.g. NCCL benchmark pods). Required in practice on tainted GPU node groups; when unset, the validator's default tolerate-all is kept.
    triggers Sequence[Any]
    Arbitrary values; changing any of them replaces the resource and re-runs validation (every input change does). Wire triggers: [stack.deployedComponents] for re-validation cadence.
    criteria Property Map
    Recipe criteria. Wire a ClusterStack's criteria output here, or build the object inline.
    context String
    Kubeconfig context to select. Defaults to the kubeconfig's current-context.
    imagePullSecrets List<String>
    Names of image pull Secrets (in the validation namespace) for the validation pods.
    imageRegistry String
    Registry override for the snapshot-agent and validator images (air-gapped mirrors). Unset keeps the default registry (ghcr.io).
    includeCtrfReport Boolean
    If true, the merged CTRF JSON report is stored in the ctrfReport output. Default: false (state-size control; per-check results are always in phaseResults).
    kubeconfig String
    Kubeconfig contents for the target cluster. Mutually exclusive with kubeconfigPath. If neither is set, the ambient kubeconfig (KUBECONFIG env var or ~/.kube/config) is used. Written to a mode-0600 temp file for the duration of the run and removed afterward.
    kubeconfigPath String
    Path to a kubeconfig file on disk. Mutually exclusive with kubeconfig.
    namespace String
    Namespace for the validation harness — use a dedicated one: it hosts the privileged agent and cluster-admin-bound validator Jobs. The SDK creates it on first run and deliberately never deletes it. Must be a valid DNS-1123 label and not a "kube-" system namespace. Default: "aicr-validation".
    nodeSelector Map<String>
    Node selector applied to the validation workload pods. Use when GPU nodes carry non-standard labels.
    phases List<String>
    Validation phases to run: any of "deployment", "conformance", "performance". Default: ["deployment", "conformance"] — "performance" is an explicit opt-in (long-running and hardware-bound).
    recipeDataVersion String
    Recipe-data version assertion (not a pin). When set, it must equal this provider build's embedded AICR recipe-data version or Create fails before any cluster work — wire recipeDataVersion: stack.recipeVersion to catch a provider upgrade between deploy and validate.
    requireGpu Boolean
    Whether the snapshot agent requires GPU nodes. Set false for hardware-free clusters (kind). Default: true.
    strict Boolean
    If true, a failed (or readiness-failed) validation fails the update while still persisting the full results in state. The engine then re-runs validation on every subsequent pulumi up until it passes. Default: false — results are recorded and the update succeeds; read the status output.
    timeoutMinutes Number
    Overall run timeout in minutes. Default: 30. Maximum: 1440 (24 hours).
    tolerations List<Property Map>
    Tolerations applied to the validation workload pods (e.g. NCCL benchmark pods). Required in practice on tainted GPU node groups; when unset, the validator's default tolerate-all is kept.
    triggers List<Any>
    Arbitrary values; changing any of them replaces the resource and re-runs validation (every input change does). Wire triggers: [stack.deployedComponents] for re-validation cadence.

    Outputs

    All input properties are implicitly available as output properties. Additionally, the ValidationRun resource produces the following output properties:

    CompletedAt string
    When the run completed, RFC 3339.
    Failed int
    Number of checks that failed.
    Id string
    The provider-assigned unique ID for this managed resource.
    Other int
    Number of checks with an inconclusive outcome (crash, OOM, timeout; includes CTRF "pending").
    Passed int
    Number of checks that passed.
    PhaseResults List<Pulumi.Labs.NvidiaAicr.Outputs.CheckResult>
    Per-check results across all phases run.
    ReadinessMessage string
    The readiness pre-flight failure message; empty unless status is "readiness-failed".
    RecipeName string
    The resolved AICR recipe name that was validated.
    RecipeVersion string
    The AICR recipe data version that was validated.
    RunId string
    The run identifier labeling this run's cluster-side artifacts.
    Skipped int
    Number of checks that were skipped.
    Status string
    Rollup verdict: "passed", "failed", or "readiness-failed". Checks with status "other" never flip the rollup (they are counted and listed). Infrastructure errors have no status — they fail the update outright.
    CtrfReport string
    The merged CTRF JSON report. Set only when includeCtrfReport is true.
    CompletedAt string
    When the run completed, RFC 3339.
    Failed int
    Number of checks that failed.
    Id string
    The provider-assigned unique ID for this managed resource.
    Other int
    Number of checks with an inconclusive outcome (crash, OOM, timeout; includes CTRF "pending").
    Passed int
    Number of checks that passed.
    PhaseResults []CheckResult
    Per-check results across all phases run.
    ReadinessMessage string
    The readiness pre-flight failure message; empty unless status is "readiness-failed".
    RecipeName string
    The resolved AICR recipe name that was validated.
    RecipeVersion string
    The AICR recipe data version that was validated.
    RunId string
    The run identifier labeling this run's cluster-side artifacts.
    Skipped int
    Number of checks that were skipped.
    Status string
    Rollup verdict: "passed", "failed", or "readiness-failed". Checks with status "other" never flip the rollup (they are counted and listed). Infrastructure errors have no status — they fail the update outright.
    CtrfReport string
    The merged CTRF JSON report. Set only when includeCtrfReport is true.
    completed_at string
    When the run completed, RFC 3339.
    failed number
    Number of checks that failed.
    id string
    The provider-assigned unique ID for this managed resource.
    other number
    Number of checks with an inconclusive outcome (crash, OOM, timeout; includes CTRF "pending").
    passed number
    Number of checks that passed.
    phase_results list(object)
    Per-check results across all phases run.
    readiness_message string
    The readiness pre-flight failure message; empty unless status is "readiness-failed".
    recipe_name string
    The resolved AICR recipe name that was validated.
    recipe_version string
    The AICR recipe data version that was validated.
    run_id string
    The run identifier labeling this run's cluster-side artifacts.
    skipped number
    Number of checks that were skipped.
    status string
    Rollup verdict: "passed", "failed", or "readiness-failed". Checks with status "other" never flip the rollup (they are counted and listed). Infrastructure errors have no status — they fail the update outright.
    ctrf_report string
    The merged CTRF JSON report. Set only when includeCtrfReport is true.
    completedAt String
    When the run completed, RFC 3339.
    failed Integer
    Number of checks that failed.
    id String
    The provider-assigned unique ID for this managed resource.
    other Integer
    Number of checks with an inconclusive outcome (crash, OOM, timeout; includes CTRF "pending").
    passed Integer
    Number of checks that passed.
    phaseResults List<CheckResult>
    Per-check results across all phases run.
    readinessMessage String
    The readiness pre-flight failure message; empty unless status is "readiness-failed".
    recipeName String
    The resolved AICR recipe name that was validated.
    recipeVersion String
    The AICR recipe data version that was validated.
    runId String
    The run identifier labeling this run's cluster-side artifacts.
    skipped Integer
    Number of checks that were skipped.
    status String
    Rollup verdict: "passed", "failed", or "readiness-failed". Checks with status "other" never flip the rollup (they are counted and listed). Infrastructure errors have no status — they fail the update outright.
    ctrfReport String
    The merged CTRF JSON report. Set only when includeCtrfReport is true.
    completedAt string
    When the run completed, RFC 3339.
    failed number
    Number of checks that failed.
    id string
    The provider-assigned unique ID for this managed resource.
    other number
    Number of checks with an inconclusive outcome (crash, OOM, timeout; includes CTRF "pending").
    passed number
    Number of checks that passed.
    phaseResults CheckResult[]
    Per-check results across all phases run.
    readinessMessage string
    The readiness pre-flight failure message; empty unless status is "readiness-failed".
    recipeName string
    The resolved AICR recipe name that was validated.
    recipeVersion string
    The AICR recipe data version that was validated.
    runId string
    The run identifier labeling this run's cluster-side artifacts.
    skipped number
    Number of checks that were skipped.
    status string
    Rollup verdict: "passed", "failed", or "readiness-failed". Checks with status "other" never flip the rollup (they are counted and listed). Infrastructure errors have no status — they fail the update outright.
    ctrfReport string
    The merged CTRF JSON report. Set only when includeCtrfReport is true.
    completed_at str
    When the run completed, RFC 3339.
    failed int
    Number of checks that failed.
    id str
    The provider-assigned unique ID for this managed resource.
    other int
    Number of checks with an inconclusive outcome (crash, OOM, timeout; includes CTRF "pending").
    passed int
    Number of checks that passed.
    phase_results Sequence[CheckResult]
    Per-check results across all phases run.
    readiness_message str
    The readiness pre-flight failure message; empty unless status is "readiness-failed".
    recipe_name str
    The resolved AICR recipe name that was validated.
    recipe_version str
    The AICR recipe data version that was validated.
    run_id str
    The run identifier labeling this run's cluster-side artifacts.
    skipped int
    Number of checks that were skipped.
    status str
    Rollup verdict: "passed", "failed", or "readiness-failed". Checks with status "other" never flip the rollup (they are counted and listed). Infrastructure errors have no status — they fail the update outright.
    ctrf_report str
    The merged CTRF JSON report. Set only when includeCtrfReport is true.
    completedAt String
    When the run completed, RFC 3339.
    failed Number
    Number of checks that failed.
    id String
    The provider-assigned unique ID for this managed resource.
    other Number
    Number of checks with an inconclusive outcome (crash, OOM, timeout; includes CTRF "pending").
    passed Number
    Number of checks that passed.
    phaseResults List<Property Map>
    Per-check results across all phases run.
    readinessMessage String
    The readiness pre-flight failure message; empty unless status is "readiness-failed".
    recipeName String
    The resolved AICR recipe name that was validated.
    recipeVersion String
    The AICR recipe data version that was validated.
    runId String
    The run identifier labeling this run's cluster-side artifacts.
    skipped Number
    Number of checks that were skipped.
    status String
    Rollup verdict: "passed", "failed", or "readiness-failed". Checks with status "other" never flip the rollup (they are counted and listed). Infrastructure errors have no status — they fail the update outright.
    ctrfReport String
    The merged CTRF JSON report. Set only when includeCtrfReport is true.

    Supporting Types

    CheckResult, CheckResultArgs

    The outcome of one validator check.
    Message string
    The check's failure or diagnostic message, if any.
    Name string
    The check's name (e.g. "gpu-operator-health").
    Phase string
    The validation phase the check ran in.
    Status string
    The check's status: "passed", "failed", "skipped", "pending", or "other".
    Message string
    The check's failure or diagnostic message, if any.
    Name string
    The check's name (e.g. "gpu-operator-health").
    Phase string
    The validation phase the check ran in.
    Status string
    The check's status: "passed", "failed", "skipped", "pending", or "other".
    message string
    The check's failure or diagnostic message, if any.
    name string
    The check's name (e.g. "gpu-operator-health").
    phase string
    The validation phase the check ran in.
    status string
    The check's status: "passed", "failed", "skipped", "pending", or "other".
    message String
    The check's failure or diagnostic message, if any.
    name String
    The check's name (e.g. "gpu-operator-health").
    phase String
    The validation phase the check ran in.
    status String
    The check's status: "passed", "failed", "skipped", "pending", or "other".
    message string
    The check's failure or diagnostic message, if any.
    name string
    The check's name (e.g. "gpu-operator-health").
    phase string
    The validation phase the check ran in.
    status string
    The check's status: "passed", "failed", "skipped", "pending", or "other".
    message str
    The check's failure or diagnostic message, if any.
    name str
    The check's name (e.g. "gpu-operator-health").
    phase str
    The validation phase the check ran in.
    status str
    The check's status: "passed", "failed", "skipped", "pending", or "other".
    message String
    The check's failure or diagnostic message, if any.
    name String
    The check's name (e.g. "gpu-operator-health").
    phase String
    The validation phase the check ran in.
    status String
    The check's status: "passed", "failed", "skipped", "pending", or "other".

    RecipeCriteria, RecipeCriteriaArgs

    Recipe-selection criteria, mirroring ClusterStack's accelerator / service / intent / os / platform / nodes inputs, plus the skipComponents the stack deployed without. Wire a ClusterStack's criteria output here so deployment and validation resolve the identical recipe and agree on which of its components are in scope.
    Accelerator string
    GPU accelerator type. Supported values: "h100", "gb200", "b200", "rtx-pro-6000" (eks/lke only).
    Intent string
    Workload intent. Supported values: "training", "inference".
    Service string
    Kubernetes service. Supported values: "aks", "bcm", "eks", "gke", "kind", "lke", "oke".
    Nodes int
    Worker-node count hint used to size the recipe.
    Os string
    Operating system flavor of the worker nodes. Leave unset for OS-agnostic resolution. Supported values: "ubuntu", "cos", "ol".
    Platform string
    ML platform/framework. Supported values: "kubeflow" (training), "dynamo" (inference), "nim" (inference, eks with h100 or rtx-pro-6000 only). kubeflow and dynamo have no recipes on lke/bcm.
    SkipComponents List<string>
    Recipe components the stack intentionally did not deploy (ClusterStack's skipComponents). ValidationRun treats them as out of scope rather than missing: checks that presuppose one of them (e.g. the gpu-operator health, DCGM metrics, and GPU-HPA checks when "gpu-operator" is skipped) are reported "skipped" with a reason instead of failing, and the SDK's component-aware checks see the components as disabled. A ClusterStack's criteria output carries its own skipComponents, so wiring it keeps validation aligned with the deployed subset automatically. Skipping does not verify a replacement you run yourself — those checks are simply not made.
    Accelerator string
    GPU accelerator type. Supported values: "h100", "gb200", "b200", "rtx-pro-6000" (eks/lke only).
    Intent string
    Workload intent. Supported values: "training", "inference".
    Service string
    Kubernetes service. Supported values: "aks", "bcm", "eks", "gke", "kind", "lke", "oke".
    Nodes int
    Worker-node count hint used to size the recipe.
    Os string
    Operating system flavor of the worker nodes. Leave unset for OS-agnostic resolution. Supported values: "ubuntu", "cos", "ol".
    Platform string
    ML platform/framework. Supported values: "kubeflow" (training), "dynamo" (inference), "nim" (inference, eks with h100 or rtx-pro-6000 only). kubeflow and dynamo have no recipes on lke/bcm.
    SkipComponents []string
    Recipe components the stack intentionally did not deploy (ClusterStack's skipComponents). ValidationRun treats them as out of scope rather than missing: checks that presuppose one of them (e.g. the gpu-operator health, DCGM metrics, and GPU-HPA checks when "gpu-operator" is skipped) are reported "skipped" with a reason instead of failing, and the SDK's component-aware checks see the components as disabled. A ClusterStack's criteria output carries its own skipComponents, so wiring it keeps validation aligned with the deployed subset automatically. Skipping does not verify a replacement you run yourself — those checks are simply not made.
    accelerator string
    GPU accelerator type. Supported values: "h100", "gb200", "b200", "rtx-pro-6000" (eks/lke only).
    intent string
    Workload intent. Supported values: "training", "inference".
    service string
    Kubernetes service. Supported values: "aks", "bcm", "eks", "gke", "kind", "lke", "oke".
    nodes number
    Worker-node count hint used to size the recipe.
    os string
    Operating system flavor of the worker nodes. Leave unset for OS-agnostic resolution. Supported values: "ubuntu", "cos", "ol".
    platform string
    ML platform/framework. Supported values: "kubeflow" (training), "dynamo" (inference), "nim" (inference, eks with h100 or rtx-pro-6000 only). kubeflow and dynamo have no recipes on lke/bcm.
    skip_components list(string)
    Recipe components the stack intentionally did not deploy (ClusterStack's skipComponents). ValidationRun treats them as out of scope rather than missing: checks that presuppose one of them (e.g. the gpu-operator health, DCGM metrics, and GPU-HPA checks when "gpu-operator" is skipped) are reported "skipped" with a reason instead of failing, and the SDK's component-aware checks see the components as disabled. A ClusterStack's criteria output carries its own skipComponents, so wiring it keeps validation aligned with the deployed subset automatically. Skipping does not verify a replacement you run yourself — those checks are simply not made.
    accelerator String
    GPU accelerator type. Supported values: "h100", "gb200", "b200", "rtx-pro-6000" (eks/lke only).
    intent String
    Workload intent. Supported values: "training", "inference".
    service String
    Kubernetes service. Supported values: "aks", "bcm", "eks", "gke", "kind", "lke", "oke".
    nodes Integer
    Worker-node count hint used to size the recipe.
    os String
    Operating system flavor of the worker nodes. Leave unset for OS-agnostic resolution. Supported values: "ubuntu", "cos", "ol".
    platform String
    ML platform/framework. Supported values: "kubeflow" (training), "dynamo" (inference), "nim" (inference, eks with h100 or rtx-pro-6000 only). kubeflow and dynamo have no recipes on lke/bcm.
    skipComponents List<String>
    Recipe components the stack intentionally did not deploy (ClusterStack's skipComponents). ValidationRun treats them as out of scope rather than missing: checks that presuppose one of them (e.g. the gpu-operator health, DCGM metrics, and GPU-HPA checks when "gpu-operator" is skipped) are reported "skipped" with a reason instead of failing, and the SDK's component-aware checks see the components as disabled. A ClusterStack's criteria output carries its own skipComponents, so wiring it keeps validation aligned with the deployed subset automatically. Skipping does not verify a replacement you run yourself — those checks are simply not made.
    accelerator string
    GPU accelerator type. Supported values: "h100", "gb200", "b200", "rtx-pro-6000" (eks/lke only).
    intent string
    Workload intent. Supported values: "training", "inference".
    service string
    Kubernetes service. Supported values: "aks", "bcm", "eks", "gke", "kind", "lke", "oke".
    nodes number
    Worker-node count hint used to size the recipe.
    os string
    Operating system flavor of the worker nodes. Leave unset for OS-agnostic resolution. Supported values: "ubuntu", "cos", "ol".
    platform string
    ML platform/framework. Supported values: "kubeflow" (training), "dynamo" (inference), "nim" (inference, eks with h100 or rtx-pro-6000 only). kubeflow and dynamo have no recipes on lke/bcm.
    skipComponents string[]
    Recipe components the stack intentionally did not deploy (ClusterStack's skipComponents). ValidationRun treats them as out of scope rather than missing: checks that presuppose one of them (e.g. the gpu-operator health, DCGM metrics, and GPU-HPA checks when "gpu-operator" is skipped) are reported "skipped" with a reason instead of failing, and the SDK's component-aware checks see the components as disabled. A ClusterStack's criteria output carries its own skipComponents, so wiring it keeps validation aligned with the deployed subset automatically. Skipping does not verify a replacement you run yourself — those checks are simply not made.
    accelerator str
    GPU accelerator type. Supported values: "h100", "gb200", "b200", "rtx-pro-6000" (eks/lke only).
    intent str
    Workload intent. Supported values: "training", "inference".
    service str
    Kubernetes service. Supported values: "aks", "bcm", "eks", "gke", "kind", "lke", "oke".
    nodes int
    Worker-node count hint used to size the recipe.
    os str
    Operating system flavor of the worker nodes. Leave unset for OS-agnostic resolution. Supported values: "ubuntu", "cos", "ol".
    platform str
    ML platform/framework. Supported values: "kubeflow" (training), "dynamo" (inference), "nim" (inference, eks with h100 or rtx-pro-6000 only). kubeflow and dynamo have no recipes on lke/bcm.
    skip_components Sequence[str]
    Recipe components the stack intentionally did not deploy (ClusterStack's skipComponents). ValidationRun treats them as out of scope rather than missing: checks that presuppose one of them (e.g. the gpu-operator health, DCGM metrics, and GPU-HPA checks when "gpu-operator" is skipped) are reported "skipped" with a reason instead of failing, and the SDK's component-aware checks see the components as disabled. A ClusterStack's criteria output carries its own skipComponents, so wiring it keeps validation aligned with the deployed subset automatically. Skipping does not verify a replacement you run yourself — those checks are simply not made.
    accelerator String
    GPU accelerator type. Supported values: "h100", "gb200", "b200", "rtx-pro-6000" (eks/lke only).
    intent String
    Workload intent. Supported values: "training", "inference".
    service String
    Kubernetes service. Supported values: "aks", "bcm", "eks", "gke", "kind", "lke", "oke".
    nodes Number
    Worker-node count hint used to size the recipe.
    os String
    Operating system flavor of the worker nodes. Leave unset for OS-agnostic resolution. Supported values: "ubuntu", "cos", "ol".
    platform String
    ML platform/framework. Supported values: "kubeflow" (training), "dynamo" (inference), "nim" (inference, eks with h100 or rtx-pro-6000 only). kubeflow and dynamo have no recipes on lke/bcm.
    skipComponents List<String>
    Recipe components the stack intentionally did not deploy (ClusterStack's skipComponents). ValidationRun treats them as out of scope rather than missing: checks that presuppose one of them (e.g. the gpu-operator health, DCGM metrics, and GPU-HPA checks when "gpu-operator" is skipped) are reported "skipped" with a reason instead of failing, and the SDK's component-aware checks see the components as disabled. A ClusterStack's criteria output carries its own skipComponents, so wiring it keeps validation aligned with the deployed subset automatically. Skipping does not verify a replacement you run yourself — those checks are simply not made.

    Toleration, TolerationArgs

    A Kubernetes pod toleration applied to validation workload pods.
    Effect string
    The taint effect to match: "NoSchedule", "PreferNoSchedule", or "NoExecute". Empty matches all effects.
    Key string
    The taint key the toleration applies to. Empty means match all keys (with operator "Exists").
    Operator string
    Key-value relationship: "Exists" or "Equal". Default: "Equal".
    TolerationSeconds int
    How long the pod tolerates a "NoExecute" taint, in seconds.
    Value string
    The taint value to match (with operator "Equal").
    Effect string
    The taint effect to match: "NoSchedule", "PreferNoSchedule", or "NoExecute". Empty matches all effects.
    Key string
    The taint key the toleration applies to. Empty means match all keys (with operator "Exists").
    Operator string
    Key-value relationship: "Exists" or "Equal". Default: "Equal".
    TolerationSeconds int
    How long the pod tolerates a "NoExecute" taint, in seconds.
    Value string
    The taint value to match (with operator "Equal").
    effect string
    The taint effect to match: "NoSchedule", "PreferNoSchedule", or "NoExecute". Empty matches all effects.
    key string
    The taint key the toleration applies to. Empty means match all keys (with operator "Exists").
    operator string
    Key-value relationship: "Exists" or "Equal". Default: "Equal".
    toleration_seconds number
    How long the pod tolerates a "NoExecute" taint, in seconds.
    value string
    The taint value to match (with operator "Equal").
    effect String
    The taint effect to match: "NoSchedule", "PreferNoSchedule", or "NoExecute". Empty matches all effects.
    key String
    The taint key the toleration applies to. Empty means match all keys (with operator "Exists").
    operator String
    Key-value relationship: "Exists" or "Equal". Default: "Equal".
    tolerationSeconds Integer
    How long the pod tolerates a "NoExecute" taint, in seconds.
    value String
    The taint value to match (with operator "Equal").
    effect string
    The taint effect to match: "NoSchedule", "PreferNoSchedule", or "NoExecute". Empty matches all effects.
    key string
    The taint key the toleration applies to. Empty means match all keys (with operator "Exists").
    operator string
    Key-value relationship: "Exists" or "Equal". Default: "Equal".
    tolerationSeconds number
    How long the pod tolerates a "NoExecute" taint, in seconds.
    value string
    The taint value to match (with operator "Equal").
    effect str
    The taint effect to match: "NoSchedule", "PreferNoSchedule", or "NoExecute". Empty matches all effects.
    key str
    The taint key the toleration applies to. Empty means match all keys (with operator "Exists").
    operator str
    Key-value relationship: "Exists" or "Equal". Default: "Equal".
    toleration_seconds int
    How long the pod tolerates a "NoExecute" taint, in seconds.
    value str
    The taint value to match (with operator "Equal").
    effect String
    The taint effect to match: "NoSchedule", "PreferNoSchedule", or "NoExecute". Empty matches all effects.
    key String
    The taint key the toleration applies to. Empty means match all keys (with operator "Exists").
    operator String
    Key-value relationship: "Exists" or "Equal". Default: "Equal".
    tolerationSeconds Number
    How long the pod tolerates a "NoExecute" taint, in seconds.
    value String
    The taint value to match (with operator "Equal").

    Package Details

    Repository
    nvidia-aicr pulumi-labs/pulumi-nvidia-aicr
    License
    Apache-2.0
    nvidia-aicr logo
    Viewing docs for NVIDIA AI Cluster Runtime v0.3.2
    published on Monday, Aug 31, 2026 by Pulumi Labs
    Schema (JSON)
    pulumi-labs/pulumi-nvidia-aicr

      Try Pulumi Cloud free.
      Your team will thank you.

      Start free trial