> ## Documentation Index > Fetch the complete documentation index at: https://docs.nebius.com/llms.txt > Use this file to discover all available pages before exploring further. # Get ## OpenAPI ````yaml https://api.nebius.cloud/openapi.json get /mk8s/v1/node-groups/{id} openapi: 3.0.3 info: title: Nebius API version: version not set servers: - url: https://api.nebius.cloud security: - bearerAuth: [] tags: - name: nebius.ai.v1.EndpointService - name: nebius.ai.v1.JobService - name: nebius.applications.v1alpha1.K8sReleaseService - name: nebius.audit.v2.AuditEventExportService - name: nebius.audit.v2.AuditEventService - name: nebius.billing.v1.CalculatorService - name: nebius.billing.v1alpha1.CalculatorService - name: nebius.billing.v1alpha1.OneTimeExportService - name: nebius.capacity.v1.CapacityAllowanceService - name: nebius.capacity.v1.CapacityBlockGroupService - name: nebius.capacity.v1.CapacityIntervalService - name: nebius.capacity.v1.ResourceAdviceService - name: nebius.compute.v1.DiskService - name: nebius.compute.v1.DiskSnapshotService - name: nebius.compute.v1.FilesystemService - name: nebius.compute.v1.GpuClusterService - name: nebius.compute.v1.ImageService - name: nebius.compute.v1.InstanceService - name: nebius.compute.v1.MaintenanceService - name: nebius.compute.v1.NVLInstanceGroupService - name: nebius.compute.v1.NodeService - name: nebius.compute.v1.PlatformService - name: nebius.dns.v1.RecordService - name: nebius.dns.v1.ZoneService - name: nebius.iam.v1.AccessKeyService - name: nebius.iam.v1.AccessPermitService - name: nebius.iam.v1.AuthPublicKeyService - name: nebius.iam.v1.FederatedCredentialsService - name: nebius.iam.v1.FederationCertificateService - name: nebius.iam.v1.FederationService - name: nebius.iam.v1.GroupMembershipService - name: nebius.iam.v1.GroupService - name: nebius.iam.v1.InvitationService - name: nebius.iam.v1.ProfileService - name: nebius.iam.v1.ProjectService - name: nebius.iam.v1.ServiceAccountService - name: nebius.iam.v1.SessionManagementService - name: nebius.iam.v1.StaticKeyService - name: nebius.iam.v1.TenantService - name: nebius.iam.v1.TenantUserAccountService - name: nebius.iam.v1.TenantUserAccountWithAttributesService - name: nebius.iam.v2.AccessKeyService - name: nebius.iam.v2.ProjectService - name: nebius.iam.v2.TenantService - name: nebius.kms.v1.AsymmetricCryptoService - name: nebius.kms.v1.AsymmetricKeyService - name: nebius.kms.v1.SymmetricCryptoService - name: nebius.kms.v1.SymmetricKeyService - name: nebius.logging.agentmanager.v1.VersionService - name: nebius.logging.v1.LogExportService - name: nebius.maintenance.v1alpha1.MaintenanceService - name: nebius.mk8s.v1.ClusterService - name: nebius.mk8s.v1.NodeGroupService - name: nebius.mk8s.v1alpha1.ClusterService - name: nebius.mk8s.v1alpha1.NodeGroupService - name: nebius.msp.mlflow.v1alpha1.ClusterService - name: nebius.msp.postgresql.v1alpha1.BackupService - name: nebius.msp.postgresql.v1alpha1.ClusterService - name: nebius.mysterybox.v1.PayloadService - name: nebius.mysterybox.v1.SecretService - name: nebius.mysterybox.v1.SecretVersionService - name: nebius.quotas.v1.QuotaAllowanceService - name: nebius.registry.v1.ArtifactService - name: nebius.registry.v1.RegistryService - name: nebius.storage.v1.BucketService - name: nebius.storage.v1.TransferService - name: nebius.storage.v1alpha1.TransferService - name: nebius.tunnel.v1.TunnelService - name: nebius.vpc.v1.AllocationService - name: nebius.vpc.v1.NetworkService - name: nebius.vpc.v1.PoolService - name: nebius.vpc.v1.RouteService - name: nebius.vpc.v1.RouteTableService - name: nebius.vpc.v1.SecurityGroupService - name: nebius.vpc.v1.SecurityRuleService - name: nebius.vpc.v1.SubnetService - name: nebius.vpc.v1.TargetGroupService - name: nebius.vpc.v1alpha1.AllocationService - name: nebius.vpc.v1alpha1.NetworkService - name: nebius.vpc.v1alpha1.PoolService - name: nebius.vpc.v1alpha1.ScopeService - name: nebius.vpc.v1alpha1.SubnetService paths: /mk8s/v1/node-groups/{id}: get: tags: - nebius.mk8s.v1.NodeGroupService summary: Get operationId: NodeGroupService_Get parameters: - in: path name: id required: true schema: type: string - in: query name: resourceVersion schema: type: string responses: '200': content: application/json: schema: $ref: '#/components/schemas/mk8sV1NodeGroup' description: A successful response. components: schemas: mk8sV1NodeGroup: description: >- NodeGroup represents Kubernetes node pool - set of worker machines having the same configuration. A Node is a Nebius Compute Instance created in Cluster.metadata.parent_id container, running kubelet that registers in Kubernetes API and a Node object created. properties: metadata: $ref: '#/components/schemas/commonV1ResourceMetadata' spec: $ref: '#/components/schemas/mk8sV1NodeGroupSpec' status: $ref: '#/components/schemas/mk8sV1NodeGroupStatus' type: object commonV1ResourceMetadata: description: Common resource metadata. properties: createdAt: description: Timestamp indicating when the resource was created. format: date-time readOnly: true type: string id: description: Identifier for the resource, unique for its resource type. type: string labels: additionalProperties: type: string description: Labels associated with the resource. type: object name: description: Human readable name for the resource. type: string parentId: description: Identifier of the parent resource to which the resource belongs. type: string resourceVersion: description: >- Version of the resource for safe concurrent modifications and consistent reads. Positive and monotonically increases on each resource spec change (but *not* on each change of the resource's container(s) or status). Service allows zero value or current. format: int64 type: string updatedAt: description: Timestamp indicating when the resource was last updated. format: date-time readOnly: true type: string required: - parentId type: object mk8sV1NodeGroupSpec: properties: autoRepair: $ref: '#/components/schemas/mk8sV1NodeGroupAutoRepairSpec' autoscaling: $ref: '#/components/schemas/mk8sV1NodeGroupAutoscalingSpec' fixedNodeCount: description: >- Number of nodes in the group. Can be changed manually at any time, except for a node group with NVLink. format: int64 type: string strategy: $ref: '#/components/schemas/mk8sV1NodeGroupDeploymentStrategy' template: $ref: '#/components/schemas/mk8sV1NodeTemplate' version: description: >- Version is desired Kubernetes version of the cluster. For now only acceptable format is `.` like "1.31". Option for patch version update will be added later. By default the cluster control plane `.` version will be used. type: string required: - template type: object mk8sV1NodeGroupStatus: properties: events: items: $ref: '#/components/schemas/v1RecurrentResourceEvent' type: array nodeCount: description: |- Total number of nodes that are currently in the node group. Both ready and not ready nodes are counted. format: int64 type: string outdatedNodeCount: description: >- Total number of nodes that has outdated node configuration. These nodes will be replaced by new nodes with up-to-date configuration. format: int64 type: string readyNodeCount: description: >- Total number of nodes that successfully joined the cluster and are ready to serve workloads. Both outdated and up-to-date nodes are counted. format: int64 type: string reconciling: description: Show that there are changes are in flight. type: boolean state: $ref: '#/components/schemas/mk8sV1NodeGroupStatusState' strategy: $ref: '#/components/schemas/mk8sV1NodeGroupDeploymentStrategy' targetNodeCount: description: >- Desired total number of nodes that should be in the node group. It is either `NodeGroupSpec.fixed_node_count` or arbitrary number between `NodeGroupAutoscalingSpec.min_node_count` and `NodeGroupAutoscalingSpec.max_node_count` decided by autoscaler. format: int64 type: string version: description: >- Actual version of NodeGroup. Have format `..-nebius-node.` like "1.30.0-nebius-node.10". Where `..` is Kubernetes version and `` is version of Node infrastructure and configuration, which update may include bug fixes, security updates and new features depending on worker node configuration. type: string type: object mk8sV1NodeGroupAutoRepairSpec: properties: conditions: description: Conditions that determine whether a node should be auto repaired. items: $ref: '#/components/schemas/mk8sV1NodeAutoRepairCondition' type: array type: object mk8sV1NodeGroupAutoscalingSpec: properties: maxNodeCount: format: int64 type: string minNodeCount: format: int64 type: string type: object mk8sV1NodeGroupDeploymentStrategy: properties: drainTimeout: description: >- Maximum amount of time that the service will spend attempting to gracefully drain a node (evicting its pods) before falling back to pod deletion. A value of 0 means no timeout: the node can be drained for an unlimited time. Important consequence of that is if PodDisruptionBudget doesn't allow evicting a pod, then NodeGroup update with node re-creation will hang on that pod eviction. Note that this is different from `kubectl drain --timeout`, which gives up and returns an error. On 2026-08-01, defaults to 0. IMPORTANT: starting from Q3 2026 new default is 10m; for new clusters it will default to 10m, node groups in existing clusters will be gradually migrated during Q3 to the default of 10m as well. To get the actual value for your node group, please see 'strategy' in its status. type: string maxSurge: $ref: '#/components/schemas/mk8sV1PercentOrCount' maxUnavailable: $ref: '#/components/schemas/mk8sV1PercentOrCount' type: object mk8sV1NodeTemplate: properties: bootDisk: $ref: '#/components/schemas/nebiusMk8sV1DiskSpec' cloudInitUserData: description: |- cloud-init user-data Should contain at least one SSH key. type: string filesystems: description: >- Static attachments of Compute Filesystem. Can be used as a workaround, until CSI for Compute Disk and Filesystem will be available. items: $ref: '#/components/schemas/nebiusMk8sV1AttachedFilesystemSpec' type: array gpuCluster: $ref: '#/components/schemas/nebiusMk8sV1GpuClusterSpec' gpuSettings: $ref: '#/components/schemas/mk8sV1GpuSettings' localDisks: $ref: '#/components/schemas/nebiusMk8sV1LocalDisksSpec' maxPods: description: >- The maximum number of Pods per node for your cluster. If omitted, MK8S assigns the default value of 110. When you configure the maximum number of Pods per node for the cluster, MK8S uses this value to allocate a CIDR range for every node in group. The node CIDR prefix is calculated as `32 - ceil(log2(2 * max_pods))`, i.e. the smallest IPv4 subnet whose total address count is at least `2 * max_pods`. Not all IPs are usable for workload Pods because some of them are consumed by system Pods. format: int64 type: string metadata: $ref: '#/components/schemas/mk8sV1NodeMetadataTemplate' networkInterfaces: items: $ref: '#/components/schemas/mk8sV1NetworkInterfaceTemplate' type: array nvlink: $ref: '#/components/schemas/v1NVLinkSpec' os: title: >- OS version that will be used to create the boot disk of Compute Instances in the NodeGroup. Supported platform / Kubernetes version / OS / driver presets combinations * `gpu-l40s-a`, `gpu-l40s-d`, `gpu-h100-sxm`, `gpu-h200-sxm`, `cpu-e1`, `cpu-e2`, `cpu-d3`: * `drivers_preset`: `""` * `version`: 1.30 → `"ubuntu22.04"` * `version`: 1.31 → `"ubuntu22.04"` (default), `"ubuntu24.04"` * `gpu-l40s-a`, `gpu-l40s-d`, `gpu-h100-sxm`, `gpu-h200-sxm`: * `drivers_preset`: `"cuda12"` (CUDA 12.4) * `version`: 1.30, 1.31 → `"ubuntu22.04"` * `drivers_preset`: `"cuda12.4"` * `version`: 1.31 → `"ubuntu22.04"` * `drivers_preset`: `"cuda12.8"` * `version`: 1.31 → `"ubuntu24.04"` * `gpu-b200-sxm`: * `drivers_preset`: `""` * `version`: 1.30, 1.31 → `"ubuntu24.04"` * `drivers_preset`: `"cuda12"` (CUDA 12.8) * `version`: 1.30, 1.31 → `"ubuntu24.04"` * `drivers_preset`: `"cuda12.8"` * `version`: 1.31 → `"ubuntu24.04"` * `gpu-b200-sxm-a`: * `drivers_preset`: `""` * `version`: 1.31 → `"ubuntu24.04"` * `drivers_preset`: `"cuda12.8"` * `version`: 1.31 → `"ubuntu24.04"` type: string preemptible: $ref: '#/components/schemas/nebiusMk8sV1PreemptibleSpec' reservationPolicy: $ref: '#/components/schemas/nebiusMk8sV1ReservationPolicy' resources: $ref: '#/components/schemas/nebiusMk8sV1ResourcesSpec' serviceAccountId: description: >- the Nebius service account whose credentials will be available on the nodes of the group. With these credentials, it is possible to make `nebius` CLI or public API requests from the nodes without the need for extra authentication. This service account is also used to make requests to container registry. `resource.serviceaccount.issueAccessToken` permission is required to use this field. type: string taints: description: >- Kubernetes Node taints. For now change will not be propagated to existing nodes, so will be applied only to Kubernetes Nodes created after the field change. That behaviour may change later. So, for now you will need to manually set them to existing nodes, if that is needed. Field change will NOT trigger NodeGroup roll out. items: $ref: '#/components/schemas/mk8sV1NodeTaint' type: array required: - resources type: object v1RecurrentResourceEvent: properties: firstOccurredAt: format: date-time title: Time of the first occurrence of a recurrent event type: string lastOccurrence: $ref: '#/components/schemas/v1ResourceEvent' occurrenceCount: format: int64 title: >- The number of times this event has occurred between `first_occurred_at` and `last_occurrence.occurred_at`. Must be > 0 type: string title: >- A resource event that has occurred (more or less in the same way) multiple times across a service-defined aggregation interval type: object mk8sV1NodeGroupStatusState: default: STATE_UNSPECIFIED enum: - STATE_UNSPECIFIED - PROVISIONING - RUNNING - DELETING type: string mk8sV1NodeAutoRepairCondition: properties: disabled: description: When true, disables the default auto-repair condition rules. type: boolean status: $ref: '#/components/schemas/mk8sV1ConditionStatus' timeout: description: >- The duration after which the node is automatically repaired if the condition remains in the specified status. type: string type: description: Node condition type. type: string required: - type type: object mk8sV1PercentOrCount: properties: count: format: int64 type: string percent: format: int64 type: string type: object nebiusMk8sV1DiskSpec: properties: blockSizeBytes: format: int64 type: string sizeBytes: format: int64 type: string sizeGibibytes: format: int64 type: string sizeKibibytes: format: int64 type: string sizeMebibytes: format: int64 type: string type: $ref: '#/components/schemas/nebiusMk8sV1DiskSpecDiskType' type: object nebiusMk8sV1AttachedFilesystemSpec: properties: attachMode: $ref: '#/components/schemas/nebiusMk8sV1AttachedFilesystemSpecAttachMode' existingFilesystem: $ref: '#/components/schemas/nebiusMk8sV1ExistingFilesystem' mountTag: description: >- Specifies the user-defined identifier, allowing to use it as a device in mount command. type: string required: - attachMode - mountTag type: object nebiusMk8sV1GpuClusterSpec: properties: id: type: string type: object mk8sV1GpuSettings: description: GPU-related settings. properties: driversPreset: title: >- Identifier of the predefined set of drivers included in the ComputeImage deployed on ComputeInstances that are part of the NodeGroup. Supported presets for different platform / Kubernetes version combinations: * `gpu-l40s-a`, `gpu-l40s-d`, `gpu-h100-sxm`, `gpu-h200-sxm`: * `version`: 1.30 → `"cuda12"` (CUDA 12.4) * `version`: 1.31 → `"cuda12"` (CUDA 12.4), `"cuda12.4"`, `"cuda12.8"` * `gpu-b200-sxm`: * `version`: 1.31 → `"cuda12"` (CUDA 12.8), `"cuda12.8"` * `gpu-b200-sxm-a`: * `version`: 1.31 → `"cuda12.8"` type: string required: - driversPreset type: object nebiusMk8sV1LocalDisksSpec: properties: config: $ref: '#/components/schemas/v1LocalDisksSpecConfig' passthroughGroup: $ref: '#/components/schemas/nebiusMk8sV1PassthroughGroupRequest' type: object mk8sV1NodeMetadataTemplate: properties: labels: additionalProperties: type: string description: >- Kubernetes Node labels. Keys and values must follow Kubernetes label syntax: https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/ For now change will not be propagated to existing nodes, so will be applied only to Kubernetes Nodes created after the field change. That behavior may change later. So, for now you will need to manually set them to existing nodes, if that is needed. System labels containing "kubernetes.io" and "k8s.io" will be ignored. Field change will NOT trigger NodeGroup roll out. type: object type: object mk8sV1NetworkInterfaceTemplate: properties: publicIpAddress: $ref: '#/components/schemas/nebiusMk8sV1PublicIPAddress' securityGroups: description: >- User provided VPC Security Groups which will be assigned to all nodes of this NodeGroup. items: $ref: '#/components/schemas/nebiusMk8sV1SecurityGroup' type: array subnetId: description: >- Nebius VPC Subnet ID that will be attached to a node cloud instance network interface. By default Cluster control plane subnet_id used. Subnet should be located in the same network with control plane. type: string type: object v1NVLinkSpec: properties: nvlInstanceGroupId: description: Existing NVLInstanceGroup ID to use. type: string type: object nebiusMk8sV1PreemptibleSpec: type: object nebiusMk8sV1ReservationPolicy: description: >- ReservationPolicy is copied as-is from NebiusAPI `compute/v1/instance.proto`. properties: policy: $ref: '#/components/schemas/nebiusMk8sV1ReservationPolicyPolicy' reservationIds: items: type: string title: Capacity block groups, order matters type: array type: object nebiusMk8sV1ResourcesSpec: properties: platform: type: string preset: type: string required: - platform type: object mk8sV1NodeTaint: properties: effect: $ref: '#/components/schemas/mk8sV1NodeTaintEffect' key: type: string value: type: string required: - key - value - effect title: >- See https://kubernetes.io/docs/concepts/scheduling-eviction/taint-and-toleration/ type: object v1ResourceEvent: properties: code: title: >- Event code (unique within the API service), in UpperCamelCase, e.g. `"DiskAttached"` type: string error: $ref: '#/components/schemas/googleRpcStatus' level: $ref: '#/components/schemas/ResourceEventLevel' message: title: >- A human-readable message describing what has happened (and suggested actions for the user, if this is a `WARN` or `ERROR` level event) type: string occurredAt: format: date-time title: Time at which the event has occurred type: string title: >- Represents an API Resource-related event which is potentially important to the end-user. What exactly constitutes an *event* to be reported is service-dependent type: object mk8sV1ConditionStatus: default: CONDITION_STATUS_UNSPECIFIED enum: - CONDITION_STATUS_UNSPECIFIED - 'TRUE' - 'FALSE' - UNKNOWN type: string nebiusMk8sV1DiskSpecDiskType: default: UNSPECIFIED enum: - UNSPECIFIED - NETWORK_SSD - NETWORK_HDD - NETWORK_SSD_IO_M3 - NETWORK_SSD_NON_REPLICATED type: string nebiusMk8sV1AttachedFilesystemSpecAttachMode: default: UNSPECIFIED enum: - UNSPECIFIED - READ_ONLY - READ_WRITE type: string nebiusMk8sV1ExistingFilesystem: properties: id: type: string required: - id type: object v1LocalDisksSpecConfig: description: >- LocalDisksSpecConfig defines actions that managed Kubernetes service performs on mounted local disks to provide them inside Kubernetes cluster with a convenient interface. properties: kubeletEphemeral: description: The default when LocalDisksSpecConfig is not set. title: >- kubelet_ephemeral: combine all local disks into a single storage volume and use it as kubelet's local ephemeral storage on the node See also https://kubernetes.io/docs/concepts/storage/ephemeral-storage/ type: boolean none: description: >- none: "do nothing" - local disks will be provisioned as on a regular compute instance. type: boolean type: object nebiusMk8sV1PassthroughGroupRequest: properties: requested: description: >- Passthrough local disks from the underlying host. Devices are expected to appear in the guest as NVMe devices (nvme0, nvme1, ...), but the exact number depends on the preset. Enabled only when this field is explicitly set. type: boolean type: object nebiusMk8sV1PublicIPAddress: description: Describes a public IP address. type: object nebiusMk8sV1SecurityGroup: properties: id: type: string type: object nebiusMk8sV1ReservationPolicyPolicy: default: AUTO description: >2- - AUTO: 1) Will try to launch instance in any reservation_ids if provided. 2) Will try to launch instance in any of the available capacity block. 3) Will try to launch instance in PAYG if 1 & 2 are not satisfied. - FORBID: The instance is launched only using on-demand (PAYG) capacity. No attempt is made to find or use a Capacity Block. It's an error to provide reservation_ids with policy = FORBID - STRICT: 1) Will try to launch the instance in Capacity Blocks from reservation_ids if provided. 2) If reservation_ids are not provided will try to launch instance in suitable & available Capacity Block. 3) Fail otherwise. enum: - AUTO - FORBID - STRICT type: string mk8sV1NodeTaintEffect: default: EFFECT_UNSPECIFIED enum: - EFFECT_UNSPECIFIED - NO_EXECUTE - NO_SCHEDULE - PREFER_NO_SCHEDULE type: string googleRpcStatus: description: >- The `Status` type defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by [gRPC](https://github.com/grpc). Each `Status` message contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the [API Design Guide](https://cloud.google.com/apis/design/errors). properties: code: description: |- The status code, which should be an enum value of [google.rpc.Code][google.rpc.Code]. format: int32 type: integer details: description: >- A list of messages that carry the error details. There is a common set of message types for APIs to use. items: $ref: '#/components/schemas/protobufAny' type: array message: description: >- A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the [google.rpc.Status.details][google.rpc.Status.details] field, or localized by the client. type: string type: object ResourceEventLevel: default: UNSPECIFIED enum: - UNSPECIFIED - DEBUG - INFO - WARN - ERROR title: |- - UNSPECIFIED: Unspecified event severity level - DEBUG: A debug event providing detailed insight. Such events are used to debug problems with specific resource(s) and process(es) - INFO: A normal event or state change. Informs what is happening with the API resource. Does not require user attention or interaction - WARN: Warning event. Indicates a potential or minor problem with the API resource and/or the corresponding processes. Needs user attention, but requires no immediate action (yet) - ERROR: Error event. Indicates a serious problem with the API resource and/or the corresponding processes. Requires immediate user action type: string protobufAny: additionalProperties: {} description: >- `Any` contains an arbitrary serialized protocol buffer message along with a URL that describes the type of the serialized message properties: '@type': description: >- A URL/resource name that uniquely identifies the type of the serialized protocol buffer message. This string must contain at least one "/" character. The last segment of the URL's path must represent the fully qualified name of the type (as in `path/google.protobuf.Duration`). The name should be in a canonical form (e.g., leading "." is not accepted). In practice, teams usually precompile into the binary all types that they expect it to use in the context of Any. However, for URLs which use the scheme `http`, `https`, or no scheme, one can optionally set up a type server that maps type URLs to message definitions as follows: * If no scheme is provided, `https` is assumed. * An HTTP GET on the URL must yield a [google.protobuf.Type][] value in binary format, or produce an error. * Applications are allowed to cache lookup results based on the URL, or have them precompiled into a binary to avoid any lookup. Therefore, binary compatibility needs to be preserved on changes to types. (Use versioned type names to manage breaking changes.) Note: this functionality is not currently available in the official protobuf release, and it is not used for type URLs beginning with type.googleapis.com. As of May 2023, there are no widely used type server implementations and no plans to implement one. Schemes other than `http`, `https` (or the empty scheme) might be used with implementation specific semantics. type: string type: object securitySchemes: bearerAuth: description: The Authorization header expects a Bearer token. scheme: bearer type: http ````