> For clean Markdown content of this page, append .md to this URL. For the complete documentation index, see https://docs.nvidia.com/dynamo/llms.txt. For full content including API reference and SDK examples, see https://docs.nvidia.com/dynamo/llms-full.txt.
# API Reference
**Auto-generated from source.** This page is regenerated from the Dynamo operator CRDs by `docs/fern/scripts/gen_kubernetes_api.py` and covers **every** type across all three API packages: the deprecated `nvidia.com/v1alpha1` surface, the supported `nvidia.com/v1beta1` surface, and the operator's own `operator.config.dynamo.nvidia.com/v1alpha1` configuration. The trimmed [DGD](/dynamo/dev/kubernetes-api/dynamo-graph-deployment), [DGDR](/dynamo/dev/kubernetes-api/dynamo-graph-deployment-request), and [DCD](/dynamo/dev/kubernetes-api/dynamo-component-deployment) references cover only user-facing `v1beta1` fields. To edit the surface, change the Go types under `deploy/operator/api/` and let CI regenerate this page.
Dynamo publishes 3 Kubernetes API packages with 180 typed CRD and config sections.
#### [nvidia.com/v1alpha1](#nvidia-com-v1alpha1)
83 types
#### [nvidia.com/v1beta1](#nvidia-com-v1beta1)
65 types
#### [operator.config.dynamo.nvidia.com/v1alpha1](#operator-config-dynamo-nvidia-com-v1alpha1)
32 types
## nvidia.com/v1alpha1
Package v1alpha1 contains API Schema definitions for the nvidia.com v1alpha1 API group.
This package defines the DynamoGraphDeploymentRequest (DGDR) custom resource, which provides
a high-level, SLA-driven interface for deploying machine learning models on Dynamo.
Package v1alpha1 contains API Schema definitions for the nvidia.com v1alpha1 API group.
**Resource Types**
#### [DynamoCheckpoint](#dynamocheckpoint)
#### [DynamoComponentDeployment](#dynamocomponentdeployment)
#### [DynamoGraphDeployment](#dynamographdeployment)
#### [DynamoGraphDeploymentRequest](#dynamographdeploymentrequest)
#### [DynamoGraphDeploymentScalingAdapter](#dynamographdeploymentscalingadapter)
#### [DynamoModel](#dynamomodel)
#### [PodSnapshot](#podsnapshot)
#### [PodSnapshotContent](#podsnapshotcontent)
#### Autoscaling
Deprecated: This field is deprecated and ignored. Use DynamoGraphDeploymentScalingAdapter
with HPA, KEDA, or Planner for autoscaling instead. See docs/kubernetes/autoscaling.md
for migration guidance. This field will be removed in a future API version.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Deprecated: This field is ignored.
Deprecated: This field is ignored.
Deprecated: This field is ignored.
Deprecated: This field is ignored.
See [HorizontalPodAutoscalerBehavior](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#horizontalpodautoscalerbehavior-v2-autoscaling).
Deprecated: This field is ignored.
See [MetricSpec](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#metricspec-v2-autoscaling).
#### CheckpointDeletionPolicy
CheckpointDeletionPolicy defines what happens to DGD-managed automatic
checkpoint resources when the owning DGD is deleted.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Delete Retain]
**Appears in:** [ServiceCheckpointConfig](#servicecheckpointconfig)
**Allowed values**
* Delete CheckpointDeletionPolicyDelete deletes DGD-managed automatic checkpoint\
CRs and artifacts when the owning DGD is deleted.\
* Retain CheckpointDeletionPolicyRetain keeps DGD-managed automatic checkpoint CRs\
and artifacts after the owning DGD is deleted. Users can reference the\
retained checkpoint with checkpointRef if they accept compatibility risk.\
#### CheckpointMode
Deprecated: use checkpoint.enabled instead.
enabled=true without checkpointRef creates a DGD-managed automatic
checkpoint; checkpointRef restores the named checkpoint.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Auto Manual]
**Appears in:** [ServiceCheckpointConfig](#servicecheckpointconfig)
**Allowed values**
* Auto Deprecated: use checkpoint.enabled=true and omit checkpointRef.\
* Manual Deprecated: use checkpointRef to restore an existing checkpoint.\
#### CheckpointStartupPolicy
CheckpointStartupPolicy defines when worker pods should wait for a checkpoint.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Immediate WaitForCheckpoint]
**Appears in:** [ServiceCheckpointConfig](#servicecheckpointconfig)
**Allowed values**
* Immediate CheckpointStartupPolicyImmediate starts workers immediately. The checkpoint\
job runs in the background, and only pods created after the checkpoint is\
Ready are restore-shaped by the pod-create mutating webhook.\
* WaitForCheckpoint CheckpointStartupPolicyWaitForCheckpoint gates worker replicas until the\
component's checkpoint is Ready, then starts them from the checkpoint.\
#### ComponentKind
ComponentKind represents the type of underlying Kubernetes resource.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[PodClique PodCliqueScalingGroup Deployment LeaderWorkerSet]
**Appears in:** [ServiceReplicaStatus](#servicereplicastatus)
**Allowed values**
* PodClique ComponentKindPodClique represents a PodClique resource.\
* PodCliqueScalingGroup ComponentKindPodCliqueScalingGroup represents a PodCliqueScalingGroup resource.\
* Deployment ComponentKindDeployment represents a Deployment resource.\
* LeaderWorkerSet ComponentKindLeaderWorkerSet represents a LeaderWorkerSet resource.\
#### ConfigMapKeySelector
ConfigMapKeySelector selects a specific key from a ConfigMap.
Used to reference external configuration data stored in ConfigMaps.
**Kind:** `type`
**Appears in:** [ProfilingConfigSpec](#profilingconfigspec)
Name of the ConfigMap containing the desired data.
**Validation:** Required: \{\} \
Key in the ConfigMap to select. If not specified, defaults to "disagg.yaml".
#### DGDRState
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Initializing Pending Profiling Deploying Ready DeploymentDeleted Failed]
**Appears in:** [DynamoGraphDeploymentRequestStatus](#dynamographdeploymentrequeststatus)
**Allowed values**
* Initializing
* Pending
* Profiling
* Deploying
* Ready
* DeploymentDeleted
* Failed
#### DGDState
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[initializing pending successful failed]
**Appears in:** [DeploymentStatus](#deploymentstatus), [DynamoGraphDeploymentStatus](#dynamographdeploymentstatus)
**Allowed values**
* initializing
* pending
* successful
* failed
#### DeploymentOverridesSpec
DeploymentOverridesSpec allows users to customize metadata for auto-created DynamoGraphDeployments.
When autoApply is enabled, these overrides are applied to the generated DGD resource.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestSpec](#dynamographdeploymentrequestspec)
Name is the desired name for the created DynamoGraphDeployment.\
If not specified, defaults to the DGDR name.
**Validation:** Optional: \{\} \
Namespace is the desired namespace for the created DynamoGraphDeployment.\
If not specified, defaults to the DGDR namespace.
**Validation:** Optional: \{\} \
Labels are additional labels to add to the DynamoGraphDeployment metadata.\
These are merged with auto-generated labels from the profiling process.
**Validation:** Optional: \{\} \
Annotations are additional annotations to add to the DynamoGraphDeployment metadata.
**Validation:** Optional: \{\} \
WorkersImage specifies the container image to use for DynamoGraphDeployment worker components.\
This image is used for both temporary DGDs created during online profiling and the final DGD.\
If omitted, the image from the base config file (e.g., disagg.yaml) is used.\
Example: "nvcr.io/nvidia/ai-dynamo/vllm-runtime:1.2.1"
**Validation:** Optional: \{\} \
#### DeploymentStatus
DeploymentStatus tracks the state of an auto-created DynamoGraphDeployment.
This status is populated when autoApply is enabled and a DGD is created.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestStatus](#dynamographdeploymentrequeststatus)
Name is the name of the created DynamoGraphDeployment.
Namespace is the namespace of the created DynamoGraphDeployment.
State is the current state of the DynamoGraphDeployment.\
This value is mirrored from the DGD's status.state field.
See [DGDState](#dgdstate).
**Validation:** Enum: \[initializing pending successful failed] \
Created indicates whether the DGD has been successfully created.\
Used to prevent recreation if the DGD is manually deleted by users.
#### DynamoCheckpoint
DynamoCheckpoint is the Schema for the dynamocheckpoints API
It represents a container checkpoint that can be used to restore pods to a warm state
**Kind:** `resource`
`nvidia.com/v1alpha1`
`DynamoCheckpoint`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
See [DynamoCheckpointSpec](#dynamocheckpointspec).
See [DynamoCheckpointStatus](#dynamocheckpointstatus).
#### DynamoCheckpointIdentity
Deprecated: legacy identity metadata. Keep it only where v1alpha1 still
requires spec.identity; omit DGD-managed identity and use checkpointRef for
explicit restores.
**Kind:** `type`
**Appears in:** [DynamoCheckpointSpec](#dynamocheckpointspec), [ServiceCheckpointConfig](#servicecheckpointconfig)
Model is the model identifier (e.g., "meta-llama/Llama-3-70B")\
Deprecated: legacy spec.identity only.
**Validation:** Required: \{\} \
BackendFramework is the runtime framework (vllm, sglang, trtllm)\
Deprecated: legacy spec.identity only.
**Validation:** Enum: \[vllm sglang trtllm] \
Required: \{\} \
DynamoVersion is the Dynamo platform version (optional).\
Deprecated: legacy spec.identity only.
**Validation:** Optional: \{\} \
TensorParallelSize is the tensor parallel configuration.\
Deprecated: checkpoint launch uses the pod template instead.
**Validation:** Minimum: 1 \
Optional: \{\} \
PipelineParallelSize is the pipeline parallel configuration.\
Deprecated: checkpoint launch uses the pod template instead.
**Validation:** Minimum: 1 \
Optional: \{\} \
Dtype is the data type (fp16, bf16, fp8, etc.).\
Deprecated: legacy spec.identity only.
**Validation:** Optional: \{\} \
MaxModelLen is the maximum sequence length.\
Deprecated: legacy spec.identity only.
**Validation:** Minimum: 1 \
Optional: \{\} \
ExtraParameters are additional parameters that affect the checkpoint hash.\
Use for any framework-specific or custom parameters not covered above.\
Deprecated: legacy spec.identity only.
**Validation:** Optional: \{\} \
#### DynamoCheckpointJobConfig
DynamoCheckpointJobConfig defines the configuration for the checkpoint creation Job
**Kind:** `type`
**Appears in:** [DynamoCheckpointSpec](#dynamocheckpointspec)
PodTemplateSpec allows customizing the checkpoint Job pod\
This should include the container that runs the workload to be checkpointed\
and any workload/runtime env, service account, GMS, or DRA wiring needed\
by that container. Auto-created checkpoints from DynamoGraphDeployment\
render Dynamo defaults before creating the DynamoCheckpoint.
See [PodTemplateSpec](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#podtemplatespec-v1-core).
**Validation:** Required: \{\} \
TargetContainerName is the container in PodTemplateSpec to snapshot.
**Validation:** MaxLength: 63 \
MinLength: 1 \
Pattern: `^[a-z0-9]([-a-z0-9]*[a-z0-9])?$` \
Optional: \{\} \
SharedMemory controls the tmpfs mounted at /dev/shm for the checkpoint Job pod.\
When omitted, checkpoint Jobs use the same default 8Gi tmpfs as Dynamo components.
See [SharedMemorySpec](#sharedmemoryspec).
**Validation:** Optional: \{\} \
ActiveDeadlineSeconds specifies the maximum time the Job can run
**Validation:** Minimum: 1 \
Optional: \{\} \
Deprecated: BackoffLimit is ignored. Checkpoint Jobs never retry.
**Validation:** Minimum: 0 \
Optional: \{\} \
Deprecated: TTLSecondsAfterFinished is ignored. The operator deletes checkpoint\
Jobs after recording their terminal outcome.
**Validation:** Minimum: 0 \
Optional: \{\} \
#### DynamoCheckpointPhase
DynamoCheckpointPhase represents the current phase of the checkpoint lifecycle
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Pending Creating Ready Failed]
**Appears in:** [DynamoCheckpointStatus](#dynamocheckpointstatus)
**Allowed values**
* Pending DynamoCheckpointPhasePending indicates the checkpoint CR has been created but the Job has not started\
* Creating DynamoCheckpointPhaseCreating indicates the checkpoint Job is running\
* Ready DynamoCheckpointPhaseReady indicates the checkpoint artifact is available\
* Failed DynamoCheckpointPhaseFailed indicates the checkpoint creation failed\
#### DynamoCheckpointSpec
DynamoCheckpointSpec defines the desired state of DynamoCheckpoint
**Kind:** `type`
**Appears in:** [DynamoCheckpoint](#dynamocheckpoint)
Deprecated: required by v1alpha1 for standalone checkpoints. Auto\
checkpoints synthesize it; checkpointRef restores use the referenced CR.
See [DynamoCheckpointIdentity](#dynamocheckpointidentity).
**Validation:** Required: \{\} \
GPUMemoryService records checkpoint-time GPU Memory Service metadata for\
a prepared checkpoint Job pod. The DynamoCheckpoint controller does not\
inject GMS/DRA resources; auto-created checkpoints from\
DynamoGraphDeployment prepare the pod template before creating this object.\
Manual GMS-enabled checkpoints must provide the prepared pod template; the\
controller fails the checkpoint if the required GMS/DRA wiring is missing.\
This field is intentionally outside spec.identity, so it does not affect\
the checkpoint identity hash or deduplication.
See [GPUMemoryServiceSpec](#gpumemoryservicespec).
**Validation:** Optional: \{\} \
Job defines the configuration for the checkpoint creation Job
See [DynamoCheckpointJobConfig](#dynamocheckpointjobconfig).
**Validation:** Required: \{\} \
#### DynamoCheckpointStatus
DynamoCheckpointStatus defines the observed state of DynamoCheckpoint
**Kind:** `type`
**Appears in:** [DynamoCheckpoint](#dynamocheckpoint)
Phase represents the current phase of the checkpoint lifecycle
See [DynamoCheckpointPhase](#dynamocheckpointphase).
**Validation:** Enum: \[Pending Creating Ready Failed] \
Optional: \{\} \
CheckpointID is the artifact ID used by the snapshot protocol.
**Validation:** Optional: \{\} \
IdentityHash is the computed hash of the checkpoint identity.\
Deprecated: use CheckpointID. This field is retained for compatibility\
with older status consumers.
**Validation:** Optional: \{\} \
Deprecated: Location is ignored and no longer populated. It is retained\
only so older objects continue to validate.
**Validation:** Optional: \{\} \
Deprecated: StorageType is ignored and no longer populated. It is retained\
only so older objects continue to validate.
See [DynamoCheckpointStorageType](#dynamocheckpointstoragetype).
**Validation:** Enum: \[pvc s3 oci] \
Optional: \{\} \
JobName is the name of the checkpoint creation Job
**Validation:** Optional: \{\} \
PodSnapshotName is the name of the PodSnapshot this checkpoint created to drive capture. It is\
the authoritative pointer to the snapshot (which is otherwise located by label, not by\
reconstructing its name) and lets the controller distinguish a never-created snapshot (empty)\
from one that was created and later went missing (set, but no longer found).
**Validation:** Optional: \{\} \
CreatedAt is the timestamp when the checkpoint became ready
See [Time](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#time-v1-meta).
**Validation:** Optional: \{\} \
Message provides additional information about the current state
**Validation:** Optional: \{\} \
DEPRECATED: Conditions are deprecated. Use status.phase instead.
See [Condition](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#condition-v1-meta).
**Validation:** Optional: \{\} \
#### DynamoCheckpointStorageType
Deprecated: StorageType is retained for compatibility with older
DynamoCheckpoint status consumers. The current checkpoint flow publishes
PVC-backed artifacts discovered from the snapshot-agent DaemonSet.
**Kind:** `type`
**Underlying type:** `string`
**Validation:** Enum: \[pvc s3 oci]
**Appears in:** [DynamoCheckpointStatus](#dynamocheckpointstatus)
#### DynamoComponentDeployment
DynamoComponentDeployment is the Schema for the dynamocomponentdeployments API
**Kind:** `resource`
`nvidia.com/v1alpha1`
`DynamoComponentDeployment`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
Spec defines the desired state for this Dynamo component deployment.
See [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec).
#### DynamoComponentDeploymentSharedSpec
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec), [DynamoGraphDeploymentSpec](#dynamographdeploymentspec)
Annotations to add to generated Kubernetes resources for this component\
(such as Pod, Service, and Ingress when applicable).
Labels to add to generated Kubernetes resources for this component.
The name of the component
ComponentType indicates the role of this component (for example, "main").
SubComponentType indicates the sub-role of this component (for example, "prefill").
GlobalDynamoNamespace indicates that the Component will be placed in the global Dynamo namespace
Resources requested and limits for this component, including CPU, memory,\
GPUs/devices, and any runtime-specific resources.
See [Resources](#resources).
Deprecated: This field is deprecated and ignored. Use DynamoGraphDeploymentScalingAdapter\
with HPA, KEDA, or Planner for autoscaling instead. See docs/kubernetes/autoscaling.md\
for migration guidance. This field will be removed in a future API version.
See [Autoscaling](#autoscaling).
Envs defines additional environment variables to inject into the component containers.
See [EnvVar](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#envvar-v1-core).
EnvFromSecret references a Secret whose key/value pairs will be exposed as\
environment variables in the component containers.
VolumeMounts references PVCs defined at the top level for volumes to be mounted by the component.
See [VolumeMount](#volumemount).
Ingress config to expose the component outside the cluster (or through a service mesh).
See [IngressSpec](#ingressspec).
ModelRef references a model that this component serves\
When specified, a headless service will be created for endpoint discovery
See [ModelReference](#modelreference).
**Validation:** Optional: \{\} \
SharedMemory controls the tmpfs mounted at /dev/shm (enable/disable and size).
See [SharedMemorySpec](#sharedmemoryspec).
ExtraPodMetadata adds labels/annotations to the created Pods.
See [ExtraPodMetadata](#extrapodmetadata).
**Validation:** Optional: \{\} \
ExtraPodSpec allows to override the main pod spec configuration.\
It is a k8s standard PodSpec. It also contains a MainContainer (standard k8s Container) field\
that allows overriding the main container configuration.
See [ExtraPodSpec](#extrapodspec).
**Validation:** Optional: \{\} \
LivenessProbe to detect and restart unhealthy containers.
See [Probe](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#probe-v1-core).
ReadinessProbe to signal when the container is ready to receive traffic.
See [Probe](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#probe-v1-core).
Replicas is the desired number of Pods for this component.\
When scalingAdapter is enabled, this field is managed by the\
DynamoGraphDeploymentScalingAdapter and should not be modified directly.
**Validation:** Minimum: 0 \
MinAvailable maps to Grove PodClique minAvailable for single-node and\
Grove PodCliqueScalingGroup minAvailable for multi-node components.\
This field determines 1) the minimum number of replicas guaranteed to be\
gang-scheduled, and 2) when violating minAvailable replicas triggers gang\
termination.\
For Grove-backed DynamoGraphDeployment components, minAvailable defaults to\
1 when omitted and is immutable after creation. Positive replica counts must\
be greater than or equal to minAvailable. Replicas may be scaled to 0 as a\
special scale-to-zero state; minAvailable remains configured but is not\
enforced again until replicas is scaled back to a positive value.\
For non-Grove deployments, setting this field will result in a validation error.
**Validation:** Minimum: 1 \
Optional: \{\} \
Multinode is the configuration for multinode components.
See [MultinodeSpec](#multinodespec).
ScalingAdapter configures whether this service uses the DynamoGraphDeploymentScalingAdapter.\
When enabled, replicas are managed via DGDSA and external autoscalers can scale\
the service using the Scale subresource. When disabled, replicas can be modified directly.
See [ScalingAdapter](#scalingadapter).
**Validation:** Optional: \{\} \
EPPConfig defines EPP-specific configuration options for Endpoint Picker Plugin components.\
Only applicable when ComponentType is "epp".
See [EPPConfig](#eppconfig).
**Validation:** Optional: \{\} \
FrontendSidecar configures an auto-generated frontend sidecar container.\
When specified, the operator injects a fully configured frontend container\
with all standard Dynamo environment variables, health probes, and ports.\
This eliminates the need to manually specify these in extraPodSpec.containers. (GAIE)
See [FrontendSidecarSpec](#frontendsidecarspec).
**Validation:** Optional: \{\} \
Checkpoint configures container checkpointing for this service.\
When enabled, pods can be restored from a checkpoint files for faster cold start.
See [ServiceCheckpointConfig](#servicecheckpointconfig).
**Validation:** Optional: \{\} \
TopologyConstraint for this service. packDomain is required.\
When both this and spec.topologyConstraint.packDomain are set, packDomain\
must be narrower than or equal to the spec-level packDomain.
See [TopologyConstraint](#topologyconstraint).
**Validation:** Optional: \{\} \
GPUMemoryService configures the GPU Memory Service (GMS) sidecar.\
When enabled, a GMS sidecar is injected and GPU access is managed via DRA.
See [GPUMemoryServiceSpec](#gpumemoryservicespec).
**Validation:** Optional: \{\} \
Failover configures GMS (GPU Memory Service) failover for this service.\
For intraPod mode: the main container is cloned into two engine containers (active + standby).\
For interPod mode: the operator creates a dedicated GMS weight server pod and\
multiple engine pods per rank that share GPUs via DRA resource claims.
See [FailoverSpec](#failoverspec).
**Validation:** Optional: \{\} \
#### DynamoComponentDeploymentSpec
DynamoComponentDeploymentSpec defines the desired state of DynamoComponentDeployment
**Kind:** `type`
**Appears in:** [DynamoComponentDeployment](#dynamocomponentdeployment)
BackendFramework specifies the backend framework (e.g., "sglang", "vllm", "trtllm")
**Validation:** Enum: \[sglang vllm trtllm] \
Annotations to add to generated Kubernetes resources for this component\
(such as Pod, Service, and Ingress when applicable).
Labels to add to generated Kubernetes resources for this component.
The name of the component
ComponentType indicates the role of this component (for example, "main").
SubComponentType indicates the sub-role of this component (for example, "prefill").
GlobalDynamoNamespace indicates that the Component will be placed in the global Dynamo namespace
Resources requested and limits for this component, including CPU, memory,\
GPUs/devices, and any runtime-specific resources.
See [Resources](#resources).
Deprecated: This field is deprecated and ignored. Use DynamoGraphDeploymentScalingAdapter\
with HPA, KEDA, or Planner for autoscaling instead. See docs/kubernetes/autoscaling.md\
for migration guidance. This field will be removed in a future API version.
See [Autoscaling](#autoscaling).
Envs defines additional environment variables to inject into the component containers.
See [EnvVar](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#envvar-v1-core).
EnvFromSecret references a Secret whose key/value pairs will be exposed as\
environment variables in the component containers.
VolumeMounts references PVCs defined at the top level for volumes to be mounted by the component.
See [VolumeMount](#volumemount).
Ingress config to expose the component outside the cluster (or through a service mesh).
See [IngressSpec](#ingressspec).
ModelRef references a model that this component serves\
When specified, a headless service will be created for endpoint discovery
See [ModelReference](#modelreference).
**Validation:** Optional: \{\} \
SharedMemory controls the tmpfs mounted at /dev/shm (enable/disable and size).
See [SharedMemorySpec](#sharedmemoryspec).
ExtraPodMetadata adds labels/annotations to the created Pods.
See [ExtraPodMetadata](#extrapodmetadata).
**Validation:** Optional: \{\} \
ExtraPodSpec allows to override the main pod spec configuration.\
It is a k8s standard PodSpec. It also contains a MainContainer (standard k8s Container) field\
that allows overriding the main container configuration.
See [ExtraPodSpec](#extrapodspec).
**Validation:** Optional: \{\} \
LivenessProbe to detect and restart unhealthy containers.
See [Probe](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#probe-v1-core).
ReadinessProbe to signal when the container is ready to receive traffic.
See [Probe](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#probe-v1-core).
Replicas is the desired number of Pods for this component.\
When scalingAdapter is enabled, this field is managed by the\
DynamoGraphDeploymentScalingAdapter and should not be modified directly.
**Validation:** Minimum: 0 \
MinAvailable maps to Grove PodClique minAvailable for single-node and\
Grove PodCliqueScalingGroup minAvailable for multi-node components.\
This field determines 1) the minimum number of replicas guaranteed to be\
gang-scheduled, and 2) when violating minAvailable replicas triggers gang\
termination.\
For Grove-backed DynamoGraphDeployment components, minAvailable defaults to\
1 when omitted and is immutable after creation. Positive replica counts must\
be greater than or equal to minAvailable. Replicas may be scaled to 0 as a\
special scale-to-zero state; minAvailable remains configured but is not\
enforced again until replicas is scaled back to a positive value.\
For non-Grove deployments, setting this field will result in a validation error.
**Validation:** Minimum: 1 \
Optional: \{\} \
Multinode is the configuration for multinode components.
See [MultinodeSpec](#multinodespec).
ScalingAdapter configures whether this service uses the DynamoGraphDeploymentScalingAdapter.\
When enabled, replicas are managed via DGDSA and external autoscalers can scale\
the service using the Scale subresource. When disabled, replicas can be modified directly.
See [ScalingAdapter](#scalingadapter).
**Validation:** Optional: \{\} \
EPPConfig defines EPP-specific configuration options for Endpoint Picker Plugin components.\
Only applicable when ComponentType is "epp".
See [EPPConfig](#eppconfig).
**Validation:** Optional: \{\} \
FrontendSidecar configures an auto-generated frontend sidecar container.\
When specified, the operator injects a fully configured frontend container\
with all standard Dynamo environment variables, health probes, and ports.\
This eliminates the need to manually specify these in extraPodSpec.containers. (GAIE)
See [FrontendSidecarSpec](#frontendsidecarspec).
**Validation:** Optional: \{\} \
Checkpoint configures container checkpointing for this service.\
When enabled, pods can be restored from a checkpoint files for faster cold start.
See [ServiceCheckpointConfig](#servicecheckpointconfig).
**Validation:** Optional: \{\} \
TopologyConstraint for this service. packDomain is required.\
When both this and spec.topologyConstraint.packDomain are set, packDomain\
must be narrower than or equal to the spec-level packDomain.
See [TopologyConstraint](#topologyconstraint).
**Validation:** Optional: \{\} \
GPUMemoryService configures the GPU Memory Service (GMS) sidecar.\
When enabled, a GMS sidecar is injected and GPU access is managed via DRA.
See [GPUMemoryServiceSpec](#gpumemoryservicespec).
**Validation:** Optional: \{\} \
Failover configures GMS (GPU Memory Service) failover for this service.\
For intraPod mode: the main container is cloned into two engine containers (active + standby).\
For interPod mode: the operator creates a dedicated GMS weight server pod and\
multiple engine pods per rank that share GPUs via DRA resource claims.
See [FailoverSpec](#failoverspec).
**Validation:** Optional: \{\} \
#### DynamoGraphDeployment
DynamoGraphDeployment is the Schema for the dynamographdeployments API.
**Kind:** `resource`
`nvidia.com/v1alpha1`
`DynamoGraphDeployment`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
Spec defines the desired state for this graph deployment.
See [DynamoGraphDeploymentSpec](#dynamographdeploymentspec).
Status reflects the current observed state of this graph deployment.
See [DynamoGraphDeploymentStatus](#dynamographdeploymentstatus).
#### DynamoGraphDeploymentExperimentalSpec
DynamoGraphDeploymentExperimentalSpec groups graph-level opt-in preview
features. Component-level experimental features are represented separately
on component specs.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentSpec](#dynamographdeploymentspec)
KvTransferPolicy configures topology-aware routing for KV-cache\
transfers between prefill and decode workers.
See [KvTransferPolicy](#kvtransferpolicy).
**Validation:** Optional: \{\} \
#### DynamoGraphDeploymentRequest
DynamoGraphDeploymentRequest is the Schema for the dynamographdeploymentrequests API.
It serves as the primary interface for users to request model deployments with
specific performance and resource constraints, enabling SLA-driven deployments.
Lifecycle:
1. Initializing → Pending: Validates spec and prepares for profiling
2. Pending → Profiling: Creates and runs profiling job (online or AIC)
3. Profiling → Ready/Deploying: Generates DGD spec after profiling completes
4. Deploying → Ready: When autoApply=true, monitors DGD until Ready
5. Ready: Terminal state when DGD is operational or spec is available
6. DeploymentDeleted: Terminal state when auto-created DGD is manually deleted
The spec becomes immutable once profiling starts. Users must delete and recreate
the DGDR to modify configuration after this point.
DEPRECATION NOTICE: v1alpha1 DynamoGraphDeploymentRequest is deprecated.
Please migrate to nvidia.com/v1beta1 DynamoGraphDeploymentRequest.
v1alpha1 will be removed in a future release.
**Kind:** `resource`
`nvidia.com/v1alpha1`
`DynamoGraphDeploymentRequest`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
Spec defines the desired state for this deployment request.
See [DynamoGraphDeploymentRequestSpec](#dynamographdeploymentrequestspec).
Status reflects the current observed state of this deployment request.
See [DynamoGraphDeploymentRequestStatus](#dynamographdeploymentrequeststatus).
#### DynamoGraphDeploymentRequestSpec
DynamoGraphDeploymentRequestSpec defines the desired state of a DynamoGraphDeploymentRequest.
This CRD serves as the primary interface for users to request model deployments with
specific performance constraints and resource requirements, enabling SLA-driven deployments.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequest](#dynamographdeploymentrequest)
Model specifies the model to deploy (e.g., "Qwen/Qwen3-0.6B", "meta-llama/Llama-3-70b").\
This is a high-level identifier for easy reference in kubectl output and logs.\
The controller automatically sets this value in profilingConfig.config.deployment.model.
**Validation:** Required: \{\} \
Backend specifies the inference backend for profiling.\
The controller automatically sets this value in profilingConfig.config.engine.backend.\
Profiling runs on real GPUs or via AIC simulation to collect performance data.
**Validation:** Enum: \[auto vllm sglang trtllm] \
Required: \{\} \
UseMocker indicates whether to deploy a mocker DynamoGraphDeployment instead of\
a real backend deployment. When true, the deployment uses simulated engines that\
don't require GPUs, using the profiling data to simulate realistic timing behavior.\
Mocker is available in all backend images and useful for large-scale experiments.\
Profiling still runs against the real backend (specified above) to collect performance data.
ProfilingConfig provides the complete configuration for the profiling job.\
Note: GPU discovery is automatically attempted to detect GPU resources from Kubernetes\
cluster nodes. If the operator has node read permissions (cluster-wide or explicitly granted),\
discovered GPU configuration is used as defaults when hardware configuration is not manually\
specified (minNumGpusPerEngine, maxNumGpusPerEngine, numGpusPerNode). User-specified values\
always take precedence over auto-discovered values. If GPU discovery fails (e.g.,\
namespace-restricted operator without node permissions), manual hardware config is required.\
This configuration is passed directly to the profiler.\
The structure matches the profile\_sla config format exactly (see ProfilingConfigSpec for schema).\
Note: deployment.model and engine.backend are automatically set from the high-level\
modelName and backend fields and should not be specified in this config.
See [ProfilingConfigSpec](#profilingconfigspec).
**Validation:** Required: \{\} \
EnableGPUDiscovery controls whether the operator attempts to discover GPU hardware from cluster nodes.\
DEPRECATED: This field is deprecated and will be removed in v1beta1. GPU discovery is now always\
attempted automatically. Setting this field has no effect - the operator will always try to discover\
GPU hardware when node read permissions are available. If discovery is unavailable (e.g., namespace-scoped\
operator without permissions), manual hardware configuration is required regardless of this setting.
**Validation:** Optional: \{\} \
AutoApply indicates whether to automatically create a DynamoGraphDeployment\
after profiling completes. If false, only the spec is generated and stored in status.\
Users can then manually create a DGD using the generated spec.
DeploymentOverrides allows customizing metadata for the auto-created DGD.\
Only applicable when AutoApply is true.
See [DeploymentOverridesSpec](#deploymentoverridesspec).
**Validation:** Optional: \{\} \
#### DynamoGraphDeploymentRequestStatus
DynamoGraphDeploymentRequestStatus represents the observed state of a DynamoGraphDeploymentRequest.
The controller updates this status as the DGDR progresses through its lifecycle.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequest](#dynamographdeploymentrequest)
State is a high-level textual status of the deployment request lifecycle.
See [DGDRState](#dgdrstate).
**Validation:** Enum: \[Initializing Pending Profiling Deploying Ready DeploymentDeleted Failed] \
Backend is extracted from profilingConfig.config.engine.backend for display purposes.\
This field is populated by the controller and shown in kubectl output.
**Validation:** Optional: \{\} \
ObservedGeneration reflects the generation of the most recently observed spec.\
Used to detect spec changes and enforce immutability after profiling starts.
Conditions contains the latest observed conditions of the deployment request.\
Standard condition types include: Validation, Profiling, SpecGenerated, DeploymentReady.\
Conditions are merged by type on patch updates.
See [Condition](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#condition-v1-meta).
ProfilingResults contains a reference to the ConfigMap holding profiling data.\
Format: "configmap/\<name\>"
**Validation:** Optional: \{\} \
GeneratedDeployment contains the full generated DynamoGraphDeployment specification\
including metadata, based on profiling results. Users can extract this to create\
a DGD manually, or it's used automatically when autoApply is true.\
Stored as RawExtension to preserve all fields including metadata.\
For mocker backends, this contains the mocker DGD spec.
See [RawExtension](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#rawextension-runtime-pkg).
**Validation:** EmbeddedResource: \{\} \
Optional: \{\} \
Deployment tracks the auto-created DGD when AutoApply is true.\
Contains name, namespace, state, and creation status of the managed DGD.
See [DeploymentStatus](#deploymentstatus).
**Validation:** Optional: \{\} \
#### DynamoGraphDeploymentScalingAdapter
DynamoGraphDeploymentScalingAdapter provides a scaling interface for individual services
within a DynamoGraphDeployment. It implements the Kubernetes scale
subresource, enabling integration with HPA, KEDA, and custom autoscalers.
The adapter acts as an intermediary between autoscalers and the DGD,
ensuring that only the adapter controller modifies the DGD's service replicas.
This prevents conflicts when multiple autoscaling mechanisms are in play.
**Kind:** `resource`
`nvidia.com/v1alpha1`
`DynamoGraphDeploymentScalingAdapter`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
See [DynamoGraphDeploymentScalingAdapterSpec](#dynamographdeploymentscalingadapterspec).
See [DynamoGraphDeploymentScalingAdapterStatus](#dynamographdeploymentscalingadapterstatus).
#### DynamoGraphDeploymentScalingAdapterSpec
DynamoGraphDeploymentScalingAdapterSpec defines the desired state of DynamoGraphDeploymentScalingAdapter
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentScalingAdapter](#dynamographdeploymentscalingadapter)
Replicas is the desired number of replicas for the target service.\
This field is modified by external autoscalers (HPA/KEDA/Planner) or manually by users.
**Validation:** Minimum: 0 \
Required: \{\} \
DGDRef references the DynamoGraphDeployment and the specific service to scale.
See [DynamoGraphDeploymentServiceRef](#dynamographdeploymentserviceref).
**Validation:** Required: \{\} \
#### DynamoGraphDeploymentScalingAdapterStatus
DynamoGraphDeploymentScalingAdapterStatus defines the observed state of DynamoGraphDeploymentScalingAdapter
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentScalingAdapter](#dynamographdeploymentscalingadapter)
Replicas is the current number of replicas for the target service.\
This is synced from the DGD's service replicas and is required for the scale subresource.
**Validation:** Optional: \{\} \
Selector is a label selector string for the pods managed by this adapter.\
Required for HPA compatibility via the scale subresource.
**Validation:** Optional: \{\} \
LastScaleTime is the last time the adapter scaled the target service.
See [Time](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#time-v1-meta).
**Validation:** Optional: \{\} \
#### DynamoGraphDeploymentServiceRef
DynamoGraphDeploymentServiceRef identifies a specific service within a DynamoGraphDeployment
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentScalingAdapterSpec](#dynamographdeploymentscalingadapterspec)
Name of the DynamoGraphDeployment
**Validation:** MinLength: 1 \
Required: \{\} \
ServiceName is the key name of the service within the DGD's spec.services map to scale
**Validation:** MinLength: 1 \
Required: \{\} \
#### DynamoGraphDeploymentSpec
DynamoGraphDeploymentSpec defines the desired state of DynamoGraphDeployment.
**Kind:** `type`
**Appears in:** [DynamoGraphDeployment](#dynamographdeployment)
Annotations to propagate to all child resources (PCS, DCD, Deployments, and pod templates).\
Service-level annotations take precedence over these values.
**Validation:** Optional: \{\} \
Labels to propagate to all child resources (PCS, DCD, Deployments, and pod templates).\
Service-level labels take precedence over these values.
**Validation:** Optional: \{\} \
PriorityClassName is the name of the PriorityClass to use for Grove PodCliqueSets.\
Requires the Grove pathway.
**Validation:** Optional: \{\} \
PVCs defines a list of persistent volume claims that can be referenced by components.\
Each PVC must have a unique name that can be referenced in component specifications.
See [PVC](#pvc).
**Validation:** MaxItems: 100 \
Optional: \{\} \
Services are the services to deploy as part of this deployment.
See [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec).
**Validation:** MaxProperties: 25 \
Optional: \{\} \
Envs are environment variables applied to all services in the deployment unless\
overridden by service-specific configuration.
See [EnvVar](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#envvar-v1-core).
**Validation:** Optional: \{\} \
BackendFramework specifies the backend framework (e.g., "sglang", "vllm", "trtllm").
**Validation:** Enum: \[sglang vllm trtllm] \
Restart specifies the restart policy for the graph deployment.
See [Restart](#restart).
**Validation:** Optional: \{\} \
TopologyConstraint is the deployment-level topology constraint.\
When set, topologyProfile is required and names the ClusterTopology CR to use.\
packDomain is optional here — it can be omitted when only services carry constraints.\
Services without their own topologyConstraint inherit from this value.
See [SpecTopologyConstraint](#spectopologyconstraint).
**Validation:** Optional: \{\} \
Experimental groups graph-level preview features whose API shape and\
behavior may change in breaking ways between releases.
See [DynamoGraphDeploymentExperimentalSpec](#dynamographdeploymentexperimentalspec).
**Validation:** Optional: \{\} \
#### DynamoGraphDeploymentStatus
DynamoGraphDeploymentStatus defines the observed state of DynamoGraphDeployment.
**Kind:** `type`
**Appears in:** [DynamoGraphDeployment](#dynamographdeployment)
ObservedGeneration is the most recent generation observed by the controller.
**Validation:** Optional: \{\} \
State is a high-level textual status of the graph deployment lifecycle.
See [DGDState](#dgdstate).
**Validation:** Enum: \[initializing pending successful failed] \
Conditions contains the latest observed conditions of the graph deployment.\
The slice is merged by type on patch updates.
See [Condition](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#condition-v1-meta).
Services contains per-service replica status information.\
The map key is the service name from spec.services.
See [ServiceReplicaStatus](#servicereplicastatus).
**Validation:** Optional: \{\} \
Restart contains the status of the restart of the graph deployment.
See [RestartStatus](#restartstatus).
**Validation:** Optional: \{\} \
Checkpoints contains per-service checkpoint status information.\
The map key is the service name from spec.services.
See [ServiceCheckpointStatus](#servicecheckpointstatus).
**Validation:** Optional: \{\} \
RollingUpdate tracks the progress of operator manged rolling updates.\
Currently only supported for singl-node, non-Grove deployments (DCD/Deployment).
See [RollingUpdateStatus](#rollingupdatestatus).
**Validation:** Optional: \{\} \
Placement groups DGD-level scheduler placement signals (score, reporting\
state, and any future placement fields).
See [PlacementStatus](#placementstatus).
**Validation:** Optional: \{\} \
#### DynamoModel
DynamoModel is the Schema for the dynamo models API
**Kind:** `resource`
`nvidia.com/v1alpha1`
`DynamoModel`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
See [DynamoModelSpec](#dynamomodelspec).
See [DynamoModelStatus](#dynamomodelstatus).
#### DynamoModelSpec
DynamoModelSpec defines the desired state of DynamoModel
**Kind:** `type`
**Appears in:** [DynamoModel](#dynamomodel)
ModelName is the full model identifier (e.g., "meta-llama/Llama-3.3-70B-Instruct-lora")
**Validation:** Required: \{\} \
BaseModelName is the base model identifier that matches the service label\
This is used to discover endpoints via headless services
**Validation:** Required: \{\} \
ModelType specifies the type of model (e.g., "base", "lora", "adapter")
**Validation:** Enum: \[base lora adapter] \
Optional: \{\} \
Source specifies the model source location (only applicable for lora model type)
See [ModelSource](#modelsource).
**Validation:** Optional: \{\} \
#### DynamoModelStatus
DynamoModelStatus defines the observed state of DynamoModel
**Kind:** `type`
**Appears in:** [DynamoModel](#dynamomodel)
Endpoints is the current list of all endpoints for this model
See [EndpointInfo](#endpointinfo).
**Validation:** Optional: \{\} \
ReadyEndpoints is the count of endpoints that are ready
LoRAFallbackCoveredEndpoints is the count of legacy prefill endpoints\
covered by a capable prefill during a rolling upgrade. These endpoints are\
excluded from ReadyEndpoints because they cannot serve the adapter directly.
**Validation:** Optional: \{\} \
TotalEndpoints is the total count of endpoints
Conditions represents the latest available observations of the model's state
See [Condition](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#condition-v1-meta).
**Validation:** Optional: \{\} \
#### EPPConfig
EPPConfig contains configuration for EPP (Endpoint Picker Plugin) components.
EPP is responsible for intelligent endpoint selection and KV-aware routing.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
ConfigMapRef references a user-provided ConfigMap containing EPP configuration.\
The ConfigMap should contain EndpointPickerConfig YAML.\
Mutually exclusive with Config.
See [ConfigMapKeySelector](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#configmapkeyselector-v1-core).
**Validation:** Optional: \{\} \
Config allows specifying EPP EndpointPickerConfig directly as a structured object.\
The operator will marshal this to YAML and create a ConfigMap automatically.\
Mutually exclusive with ConfigMapRef.\
One of ConfigMapRef or Config must be specified (no default configuration).\
Uses the upstream type from github.com/kubernetes-sigs/gateway-api-inference-extension
**Validation:** Type: object \
Optional: \{\} \
#### EndpointInfo
EndpointInfo represents a single endpoint (pod) serving the model
**Kind:** `type`
**Appears in:** [DynamoModelStatus](#dynamomodelstatus)
Address is the full address of the endpoint (e.g., "[http://10.0.1.5:9090](http://10.0.1.5:9090)")
PodName is the name of the pod serving this endpoint
**Validation:** Optional: \{\} \
Ready indicates whether this endpoint is ready to serve traffic.\
For LoRA models: true only if this endpoint's lifecycle request succeeded.\
For base models: always false (no probing performed).
LoRAFallbackCovered indicates a legacy prefill endpoint that cannot manage\
LoRAs itself is covered by a capable prefill in the same topology during a\
rolling upgrade. It does not make this endpoint ready to serve the adapter.
**Validation:** Optional: \{\} \
#### ExtraPodMetadata
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
#### ExtraPodSpec
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
See [Container](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#container-v1-core).
#### FailoverSpec
FailoverSpec configures active-passive failover for a worker component.
For intraPod mode: requires gpuMemoryService.enabled; the main container is cloned
into engine containers (active + standby) within the same pod.
For interPod mode: the operator creates a dedicated GMS weight server pod and
multiple engine pods per rank that share GPUs via DRA resource claims.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Enabled activates failover mode.
Mode selects the failover deployment topology.\
intraPod: engine containers run within the same pod (requires gpuMemoryService.enabled).\
interPod: a dedicated GMS weight server pod + engine pods per rank (requires Grove).
See [GPUMemoryServiceMode](#gpumemoryservicemode).
**Validation:** Enum: \[intraPod interPod] \
Optional: \{\} \
NumShadows is the number of shadow (standby) engine pods per rank.\
Total engine pods per rank = NumShadows + 1 (1 primary + NumShadows shadows).\
NumShadows is only meaningful for mode=interPod; intraPod uses a fixed\
1 primary + 1 shadow sidecar layout and any value other than 1 is\
rejected at admission time.
**Validation:** Minimum: 1 \
Optional: \{\} \
#### FrontendSidecarSpec
FrontendSidecarSpec configures the auto-generated frontend sidecar container.
The operator uses these fields together with built-in frontend defaults (command, probes, ports,
and Dynamo env vars) to produce a fully configured sidecar container.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Image is the container image for the frontend sidecar.
**Validation:** Required: \{\} \
Args overrides the default frontend arguments. When specified, these replace\
the default \["-m", "dynamo.frontend"] entirely.\
For example, \["-m", "dynamo.frontend", "--router-mode", "direct"] for GAIE deployments.
**Validation:** Optional: \{\} \
EnvFromSecret references a Secret whose key/value pairs will be exposed as\
environment variables in the frontend sidecar container.
**Validation:** Optional: \{\} \
Envs defines additional environment variables for the frontend sidecar.\
These are merged with (and can override) the auto-generated Dynamo env vars.
See [EnvVar](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#envvar-v1-core).
**Validation:** Optional: \{\} \
#### GMSClientPodSpec
GMSClientPodSpec declares an additional GMS client pod for inter-pod GMS.
**Kind:** `type`
**Appears in:** [GPUMemoryServiceSpec](#gpumemoryservicespec)
Name identifies this client pod.
**Validation:** MaxLength: 63 \
MinLength: 1 \
Pattern: `^[a-z0-9]([-a-z0-9]*[a-z0-9])?$` \
PodTemplate configures the pod to run as a GMS client.
See [PodTemplateSpec](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#podtemplatespec-v1-core).
**Validation:** Schemaless: \{\} \
Type: object \
#### GPUMemoryServiceMode
GPUMemoryServiceMode selects the GMS deployment topology.
**Kind:** `enum`
**Underlying type:** `string`
**Appears in:** [FailoverSpec](#failoverspec), [GPUMemoryServiceSpec](#gpumemoryservicespec)
**Allowed values**
* intraPod GMSModeIntraPod runs GMS as a sidecar within the same pod.\
* interPod GMSModeInterPod runs GMS as a separate weight server pod and one or more\
engine pods per rank, sharing GPUs via DRA ResourceClaims and a shared\
hostPath volume for UDS sockets. Extra client pod rendering is reserved\
for a follow-up change.\
#### GPUMemoryServiceSpec
GPUMemoryServiceSpec configures the GPU Memory Service (GMS) for a worker component.
**Kind:** `type`
**Appears in:** [DynamoCheckpointSpec](#dynamocheckpointspec), [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Enabled activates GMS wiring. GPU resources on client containers are\
replaced with a DRA ResourceClaim for shared GPU access.
Mode selects the GMS deployment topology.
See [GPUMemoryServiceMode](#gpumemoryservicemode).
**Validation:** Enum: \[intraPod interPod] \
Optional: \{\} \
DeviceClassName is the DRA DeviceClass to request GPUs from.
**Validation:** Optional: \{\} \
ExtraClientContainers lists additional user-declared containers that should\
be wired as GMS clients in pods rendered from the enclosing spec.\
DGD/DCD services apply this to service pods. Auto-created checkpoints\
apply checkpoint job clients before creating the DynamoCheckpoint; manual\
DynamoCheckpoint users must provide an already-prepared pod template.\
In each rendered pod, only matching container names are wired; absent\
names are ignored.
**Validation:** items:MaxLength: 63 \
items:MinLength: 1 \
items:Pattern: `^[a-z0-9]([-a-z0-9]*[a-z0-9])?$` \
Optional: \{\} \
ExtraClientPods declares additional GMS client pods for inter-pod GMS. This field is\
reserved for future use and is rejected until inter-pod client orchestration is wired.
See [GMSClientPodSpec](#gmsclientpodspec).
**Validation:** Optional: \{\} \
#### IngressSpec
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Enabled exposes the component through an ingress or virtual service when true.
Host is the base host name to route external traffic to this component.
UseVirtualService indicates whether to configure a service-mesh VirtualService instead of a standard Ingress.
VirtualServiceGateway optionally specifies the gateway name to attach the VirtualService to.
HostPrefix is an optional prefix added before the host.
Annotations to set on the generated Ingress/VirtualService resources.
Labels to set on the generated Ingress/VirtualService resources.
TLS holds the TLS configuration used by the Ingress/VirtualService.
See [IngressTLSSpec](#ingresstlsspec).
HostSuffix is an optional suffix appended after the host.
IngressControllerClassName selects the ingress controller class (e.g., "nginx").
#### IngressTLSSpec
**Kind:** `type`
**Appears in:** [IngressSpec](#ingressspec)
SecretName is the name of a Kubernetes Secret containing the TLS certificate and key.
#### KvTransferEnforcement
KvTransferEnforcement controls how the selected prefill worker's topology is
applied to decode routing.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[required preferred]
**Appears in:** [KvTransferPolicy](#kvtransferpolicy)
**Allowed values**
* required KvTransferEnforcementRequired enforces same-domain decode worker\
selection.\
* preferred KvTransferEnforcementPreferred biases decode worker selection toward the\
same domain.\
#### KvTransferPolicy
KvTransferPolicy configures topology-aware routing for KV-cache transfers
between prefill and decode workers. This graph-wide policy lives under
`spec.experimental` while the API is incubating.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentExperimentalSpec](#dynamographdeploymentexperimentalspec)
ClusterTopologyName references a Grove ClusterTopology CR. The operator\
reads the CR's topology levels and projects them through Dynamo-owned pod\
labels for worker topology metadata.
**Validation:** MinLength: 1 \
Optional: \{\} \
LabelKey is a Kubernetes node label key (e.g.\
"topology.kubernetes.io/zone") whose value identifies the topology\
domain for each worker. The operator copies the node label onto worker\
pods so the runtime can publish it as worker metadata. The label\
should correspond to the topology level named in `domain`.
**Validation:** MaxLength: 317 \
MinLength: 1 \
Pattern: `^(([a-z0-9]([-a-z0-9]\{0,61\}[a-z0-9])?)(\.[a-z0-9]([-a-z0-9]\{0,61\}[a-z0-9])?)*/)?([A-Za-z0-9]([-A-Za-z0-9_.]\{0,61\}[A-Za-z0-9])?)$` \
Optional: \{\} \
Domain is the logical name for the topology level to enforce\
(e.g. "zone", "rack"). The router uses this to match workers that\
share the same value for the label identified by `labelKey`.
See [TopologyDomain](#topologydomain).
**Validation:** Pattern: `^[a-z0-9]([a-z0-9-]*[a-z0-9])?$` \
Enforcement controls how the selected prefill worker's topology is\
applied to decode routing. "required" only allows decode workers in the\
same topology domain as the selected prefill worker. "preferred" keeps\
all decode workers eligible, but biases selection toward workers in the\
same topology domain. Defaults to "required".
See [KvTransferEnforcement](#kvtransferenforcement).
**Validation:** Enum: \[required preferred] \
Optional: \{\} \
PreferredWeight is required and used only when enforcement is\
"preferred". Higher values create a stronger same-domain routing\
preference, but do not guarantee same-domain selection. The value is not\
a probability; worker selection still depends on load and other routing\
inputs. A value of 0 disables the topology preference; 1 is the strongest\
supported preference.
**Validation:** Maximum: 1 \
Minimum: 0 \
Optional: \{\} \
#### ModelReference
ModelReference identifies a model served by this component
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Name is the base model identifier (e.g., "llama-3-70b-instruct-v1")
**Validation:** Required: \{\} \
Revision is the model revision/version (optional)
**Validation:** Optional: \{\} \
#### ModelSource
ModelSource defines the source location of a model
**Kind:** `type`
**Appears in:** [DynamoModelSpec](#dynamomodelspec)
URI is the model source URI\
Supported formats:\
- S3: s3://bucket/path/to/model\
- HuggingFace: hf://org/model\@revision\_sha\
- Local filesystem: file:///path/to/model
**Validation:** Required: \{\} \
#### MultinodeSpec
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Indicates the number of nodes to deploy for multinode components.\
Total number of GPUs is NumberOfNodes \* GPU limit.\
Must be greater than 1.
**Validation:** Minimum: 2 \
#### PVC
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentSpec](#dynamographdeploymentspec)
Create indicates to create a new PVC
Name is the name of the PVC
**Validation:** Required: \{\} \
StorageClass to be used for PVC creation. Required when create is true.
Size of the volume in Gi, used during PVC creation. Required when create is true.
See [Quantity](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#quantity-resource-api).
VolumeAccessMode is the volume access mode of the PVC. Required when create is true.
See [PersistentVolumeAccessMode](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#persistentvolumeaccessmode-v1-core).
#### PlacementScoreState
PlacementScoreState describes whether placement score is available and how
complete the reported score is for a graph deployment. See the v1beta1
PlacementScoreState for the authoritative semantics of each value.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Reported Partial Unsupported Unknown]
**Appears in:** [PlacementStatus](#placementstatus)
**Allowed values**
* Reported
* Partial
* Unsupported
* Unknown
#### PlacementStatus
PlacementStatus groups DGD-level scheduler placement fields under a single
status object so future placement signals can be added without a schema
break. See the v1beta1 PlacementStatus for the authoritative field docs.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentStatus](#dynamographdeploymentstatus)
Score is the DGD-level scheduler placement score. Normalized to \[0.0, 1.0]\
where higher is better and 1.0 is the best possible placement.
**Validation:** Maximum: 1 \
Minimum: 0 \
Optional: \{\} \
State indicates placement score reporting state.
See [PlacementScoreState](#placementscorestate).
**Validation:** Enum: \[Reported Partial Unsupported Unknown] \
Optional: \{\} \
#### PodReference
PodReference names a pod in the same namespace as the referencing PodSnapshot.
**Kind:** `type`
**Appears in:** [PodSnapshotContentSource](#podsnapshotcontentsource), [PodSnapshotSource](#podsnapshotsource)
Name of the source pod.
**Validation:** MinLength: 1 \
Required: \{\} \
UID of the source pod, recorded so the node agent dumps that specific\
pod and not a same-named recreation.
See [UID](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#uid-types-pkg).
**Validation:** Optional: \{\} \
#### PodSnapshot
PodSnapshot is the Schema for the snapshots API. It is the namespaced binding
for a captured container checkpoint and is consumed by restore paths.
No conversion: this type exists only in v1alpha1 (no other API version), so it
is not part of any conversion scheme.
**Kind:** `resource`
`nvidia.com/v1alpha1`
`PodSnapshot`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
See [PodSnapshotSpec](#podsnapshotspec).
See [PodSnapshotStatus](#podsnapshotstatus).
#### PodSnapshotContent
PodSnapshotContent is the Schema for the snapshotcontents API. It is the
cluster-scoped artifact-of-record for a captured container checkpoint.
No conversion: this type exists only in v1alpha1 (no other API version), so it
is not part of any conversion scheme.
**Kind:** `resource`
`nvidia.com/v1alpha1`
`PodSnapshotContent`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
See [PodSnapshotContentSpec](#podsnapshotcontentspec).
See [PodSnapshotContentStatus](#podsnapshotcontentstatus).
#### PodSnapshotContentSource
PodSnapshotContentSource is the immutable source descriptor: what to dump
(PodRef) and where it runs (NodeName).
**Kind:** `type`
**Appears in:** [PodSnapshotContentSpec](#podsnapshotcontentspec)
PodRef identifies the pod to dump. Its UID guards against dumping a\
same-named recreation of the pod.
See [PodReference](#podreference).
**Validation:** Required: \{\} \
NodeName is the node the source pod runs on, denormalized from the live\
pod so it travels with PodRef as one immutable unit and selects the node\
agent that performs the dump.
**Validation:** MinLength: 1 \
Required: \{\} \
#### PodSnapshotContentSpec
PodSnapshotContentSpec defines the desired state of PodSnapshotContent. It is
populated by the PodSnapshotReconciler (operator) at creation time and is
immutable thereafter.
**Kind:** `type`
**Appears in:** [PodSnapshotContent](#podsnapshotcontent)
PodSnapshotRef is the back-pointer to the bound PodSnapshot. It may span\
namespaces because PodSnapshotContent is cluster-scoped.
See [PodSnapshotReference](#podsnapshotreference).
**Validation:** Required: \{\} \
Source describes what to capture: the source pod and the node it runs on.
See [PodSnapshotContentSource](#podsnapshotcontentsource).
**Validation:** Required: \{\} \
#### PodSnapshotContentStatus
PodSnapshotContentStatus defines the observed state of PodSnapshotContent.
**Kind:** `type`
**Appears in:** [PodSnapshotContent](#podsnapshotcontent)
Conditions reflect the latest observations of the PodSnapshotContent's state.\
Standard types are Ready and Failed.
See [Condition](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#condition-v1-meta).
**Validation:** Optional: \{\} \
#### PodSnapshotReference
PodSnapshotReference is a cross-namespace reference to a PodSnapshot.
**Kind:** `type`
**Appears in:** [PodSnapshotContentSpec](#podsnapshotcontentspec)
Namespace of the referenced PodSnapshot.
**Validation:** Required: \{\} \
Name of the referenced PodSnapshot.
**Validation:** Required: \{\} \
UID of the referenced PodSnapshot, recorded at binding time to detect a\
stale reference after a delete and recreate.
See [UID](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#uid-types-pkg).
**Validation:** Optional: \{\} \
#### PodSnapshotSource
PodSnapshotSource identifies the workload captured by a PodSnapshot.
**Kind:** `type`
**Appears in:** [PodSnapshotSpec](#podsnapshotspec)
PodRef references the pod, in the PodSnapshot's namespace, that is captured.\
The operator prepares the pod (control volume, target-container annotation,\
checkpoint storage mount) before creating the PodSnapshot.
See [PodReference](#podreference).
**Validation:** Required: \{\} \
#### PodSnapshotSpec
PodSnapshotSpec defines the desired state of PodSnapshot.
**Kind:** `type`
**Appears in:** [PodSnapshot](#podsnapshot)
Source identifies the captured workload. It is a struct (rather than an\
inlined reference) so future source variants can be added additively.
See [PodSnapshotSource](#podsnapshotsource).
**Validation:** Required: \{\} \
#### PodSnapshotStatus
PodSnapshotStatus defines the observed state of PodSnapshot.
**Kind:** `type`
**Appears in:** [PodSnapshot](#podsnapshot)
BoundPodSnapshotContentName is the name of the cluster-scoped PodSnapshotContent\
this PodSnapshot is bound to. It is nil until the agent has created the\
content and recorded the binding.
**Validation:** Optional: \{\} \
Conditions reflect the latest observations of the PodSnapshot's state.\
Standard types are Ready and Failed.
See [Condition](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#condition-v1-meta).
**Validation:** Optional: \{\} \
#### ProfilingConfigSpec
ProfilingConfigSpec defines configuration for the profiling process.
This structure maps directly to the profile\_sla.py config format.
See dynamo/profiler/utils/profiler\_argparse.py for the complete schema.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestSpec](#dynamographdeploymentrequestspec)
Config is the profiling configuration as arbitrary JSON/YAML. This will be passed directly to the profiler.\
The profiler will validate the configuration and report any errors.
See [JSON](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#json-v1-apiextensions-k8s-io).
**Validation:** Optional: \{\} \
Type: object \
ConfigMapRef is an optional reference to a ConfigMap containing the DynamoGraphDeployment\
base config file (disagg.yaml). This is separate from the profiling config above.\
The path to this config will be set as engine.config in the profiling config.
See [ConfigMapKeySelector](#configmapkeyselector).
**Validation:** Optional: \{\} \
ProfilerImage specifies the container image to use for profiling jobs.\
This image contains the profiler code and dependencies needed for SLA-based profiling.\
Example: "nvcr.io/nvidia/ai-dynamo/dynamo-planner:1.2.1"
**Validation:** Required: \{\} \
OutputPVC is an optional PersistentVolumeClaim name for storing profiling output.\
If specified, all profiling artifacts (logs, plots, configs, raw data) will be written\
to this PVC instead of an ephemeral emptyDir volume. This allows users to access\
complete profiling results after the job completes by mounting the PVC.\
The PVC must exist in the same namespace as the DGDR.\
If not specified, profiling uses emptyDir and only essential data is saved to ConfigMaps.\
Note: ConfigMaps are still created regardless of this setting for planner integration.
**Validation:** Optional: \{\} \
Resources specifies the compute resource requirements for the profiling job container.\
If not specified, no resource requests or limits are set.
See [ResourceRequirements](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#resourcerequirements-v1-core).
**Validation:** Optional: \{\} \
Tolerations allows the profiling job to be scheduled on nodes with matching taints.\
For example, to schedule on GPU nodes, add a toleration for the nvidia.com/gpu taint.
See [Toleration](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#toleration-v1-core).
**Validation:** Optional: \{\} \
NodeSelector is a selector which must match a node's labels for the profiling pod to be scheduled on that node.\
For example, to schedule on ARM64 nodes, use \{"kubernetes.io/arch": "arm64"\}.
**Validation:** Optional: \{\} \
#### ResourceItem
**Kind:** `type`
**Appears in:** [Resources](#resources)
CPU specifies the CPU resource request/limit (e.g., "1000m", "2")
Memory specifies the memory resource request/limit (e.g., "4Gi", "8Gi")
GPU indicates the number of GPUs to request.\
Total number of GPUs is NumberOfNodes \* GPU in case of multinode deployment.
GPUType can specify a custom GPU type, e.g. "gpu.intel.com/xe"\
By default if not specified, the GPU type is "nvidia.com/gpu"
Custom specifies additional custom resource requests/limits
#### Resources
Resources defines requested and limits for a component, including CPU, memory,
GPUs/devices, and any runtime-specific resources.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Requests specifies the minimum resources required by the component
See [ResourceItem](#resourceitem).
Limits specifies the maximum resources allowed for the component
See [ResourceItem](#resourceitem).
Claims specifies resource claims for dynamic resource allocation
See [ResourceClaim](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#resourceclaim-v1-core).
#### Restart
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentSpec](#dynamographdeploymentspec)
ID is an arbitrary string that triggers a restart when changed.\
Any modification to this value will initiate a restart of the graph deployment according to the strategy.
**Validation:** MinLength: 1 \
Required: \{\} \
Strategy specifies the restart strategy for the graph deployment.
See [RestartStrategy](#restartstrategy).
**Validation:** Optional: \{\} \
#### RestartPhase
**Kind:** `enum`
**Underlying type:** `string`
**Appears in:** [RestartStatus](#restartstatus)
**Allowed values**
* Pending
* Restarting
* Completed
* Failed
* Superseded
#### RestartStatus
RestartStatus contains the status of the restart of the graph deployment.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentStatus](#dynamographdeploymentstatus)
ObservedID is the restart ID that has been observed and is being processed.\
Matches the Restart.ID field in the spec.
Phase is the phase of the restart.
See [RestartPhase](#restartphase).
InProgress contains the names of the services that are currently being restarted.
**Validation:** Optional: \{\} \
#### RestartStrategy
**Kind:** `type`
**Appears in:** [Restart](#restart)
Type specifies the restart strategy type.
See [RestartStrategyType](#restartstrategytype).
**Validation:** Enum: \[Sequential Parallel] \
Order specifies the order in which the services should be restarted.
**Validation:** Optional: \{\} \
#### RestartStrategyType
**Kind:** `enum`
**Underlying type:** `string`
**Appears in:** [RestartStrategy](#restartstrategy)
**Allowed values**
* Sequential
* Parallel
#### RollingUpdatePhase
RollingUpdatePhase represents the current phase of a rolling update.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Pending InProgress Completed Failed ]
**Appears in:** [RollingUpdateStatus](#rollingupdatestatus)
**Allowed values**
* Pending
* InProgress
* Completed
* \`\`
#### RollingUpdateStatus
RollingUpdateStatus tracks the progress of a rolling update.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentStatus](#dynamographdeploymentstatus)
Phase indicates the current phase of the rolling update.
See [RollingUpdatePhase](#rollingupdatephase).
**Validation:** Enum: \[Pending InProgress Completed Failed ] \
Optional: \{\} \
StartTime is when the rolling update began.
See [Time](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#time-v1-meta).
**Validation:** Optional: \{\} \
EndTime is when the rolling update completed (successfully or failed).
See [Time](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#time-v1-meta).
**Validation:** Optional: \{\} \
UpdatedServices is the list of services that have completed the rolling update.\
A service is considered updated when its new replicas are all ready and old replicas are fully scaled down.\
Only services of componentType Worker (or Prefill/Decode) are considered.
**Validation:** Optional: \{\} \
#### ScalingAdapter
ScalingAdapter configures whether a service uses the DynamoGraphDeploymentScalingAdapter
for replica management. When enabled, the DGDSA owns the replicas field and
external autoscalers (HPA, KEDA, Planner) can control scaling via the Scale subresource.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Enabled indicates whether the ScalingAdapter should be enabled for this service.\
When true, a DGDSA is created and owns the replicas field.\
When false (default), no DGDSA is created and replicas can be modified directly in the DGD.
**Validation:** Optional: \{\} \
#### ServiceCheckpointConfig
ServiceCheckpointConfig configures checkpointing for a DGD service
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Enabled indicates whether checkpointing is enabled for this service
**Validation:** Optional: \{\} \
Deprecated: omit mode. Use enabled=true without checkpointRef for a\
DGD-managed automatic checkpoint, or use checkpointRef to restore the\
named checkpoint.
See [CheckpointMode](#checkpointmode).
**Validation:** Enum: \[Auto Manual] \
Optional: \{\} \
StartupPolicy defines when normal worker replicas are started relative to\
automatic checkpoint readiness.\
- Immediate: start workers cold immediately; later Pods restore from the\
checkpoint once it is Ready.\
- WaitForCheckpoint: keep worker replicas at zero until the checkpoint is\
Ready, then start them from the checkpoint.
See [CheckpointStartupPolicy](#checkpointstartuppolicy).
**Validation:** Enum: \[Immediate WaitForCheckpoint] \
Optional: \{\} \
DeletionPolicy defines whether a DGD-managed automatic checkpoint CR and\
artifact are deleted or retained when the owning DGD is deleted.\
Explicit checkpointRef checkpoints are never owned or deleted by the DGD.
See [CheckpointDeletionPolicy](#checkpointdeletionpolicy).
**Validation:** Enum: \[Delete Retain] \
Optional: \{\} \
CheckpointRef references an existing DynamoCheckpoint CR by metadata.name.\
If specified, this service's Identity is ignored and the referenced checkpoint is used directly.
**Validation:** Optional: \{\} \
Deprecated: omit for DGD-managed checkpoints; no action is needed.\
Use CheckpointRef to restore an existing checkpoint.
See [DynamoCheckpointIdentity](#dynamocheckpointidentity).
**Validation:** Optional: \{\} \
TargetContainerName is the workload container to snapshot and restore.
**Validation:** MaxLength: 63 \
MinLength: 1 \
Pattern: `^[a-z0-9]([-a-z0-9]*[a-z0-9])?$` \
Optional: \{\} \
Job customizes the DGD-managed checkpoint Job.
See [ServiceCheckpointJobConfig](#servicecheckpointjobconfig).
**Validation:** Optional: \{\} \
#### ServiceCheckpointJobConfig
ServiceCheckpointJobConfig customizes the checkpoint Job created for a DGD service.
**Kind:** `type`
**Appears in:** [ServiceCheckpointConfig](#servicecheckpointconfig)
GMSClientContainers lists checkpoint Job containers that should receive\
GMS client wiring. Requires gpuMemoryService on the service.
**Validation:** items:MaxLength: 63 \
items:MinLength: 1 \
items:Pattern: `^[a-z0-9]([-a-z0-9]*[a-z0-9])?$` \
Optional: \{\} \
PodTemplate customizes the checkpoint Job pod. The operator starts from the\
selected workload container and merges this template so users can add helper\
containers such as gms-saver.
See [PodTemplateSpec](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#podtemplatespec-v1-core).
**Validation:** Schemaless: \{\} \
Type: object \
Optional: \{\} \
#### ServiceCheckpointStatus
ServiceCheckpointStatus contains checkpoint information for a single service.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentStatus](#dynamographdeploymentstatus)
CheckpointName is the name of the associated Checkpoint CR
**Validation:** Optional: \{\} \
CheckpointID is the artifact ID used by the snapshot protocol
**Validation:** Optional: \{\} \
IdentityHash is the computed hash of the checkpoint identity\
Deprecated: automatic checkpoints use CheckpointID. This field is retained\
for older status consumers.
**Validation:** Optional: \{\} \
Ready indicates the checkpoint artifact is ready for future pods to restore.
**Validation:** Optional: \{\} \
#### ServiceReplicaStatus
ServiceReplicaStatus contains replica information for a single service.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentStatus](#dynamographdeploymentstatus)
ComponentKind is the underlying resource kind (e.g., "PodClique", "PodCliqueScalingGroup", "Deployment", "LeaderWorkerSet").
See [ComponentKind](#componentkind).
**Validation:** Enum: \[PodClique PodCliqueScalingGroup Deployment LeaderWorkerSet] \
ComponentName is the name of the primary underlying resource.\
DEPRECATED: Use ComponentNames instead. This field will be removed in a future release.\
During rolling updates, this reflects the new (target) component name.
ComponentNames is the list of underlying resource names for this service.\
During normal operation, this contains a single name.\
During rolling updates, this contains both old and new component names.
**Validation:** Optional: \{\} \
RuntimeNamespace is the effective Dynamo runtime namespace for this\
component. Worker components may include a generation suffix; non-workers and\
Grove-backed workers use the base namespace. During rolling updates, worker\
status keeps the old active revision namespace until cutover completes.
**Validation:** Optional: \{\} \
Replicas is the total number of non-terminated replicas.\
Required for all component kinds.
**Validation:** Minimum: 0 \
UpdatedReplicas is the number of replicas at the current/desired revision.\
Required for all component kinds.
**Validation:** Minimum: 0 \
ReadyReplicas is the number of ready replicas.\
Populated for PodClique, Deployment, and LeaderWorkerSet.\
Not available for PodCliqueScalingGroup.\
When nil, the field is omitted from the API response.
**Validation:** Minimum: 0 \
Optional: \{\} \
AvailableReplicas is the number of available replicas.\
For Deployment: replicas ready for >= minReadySeconds.\
For PodCliqueScalingGroup: replicas where all constituent PodCliques have >= MinAvailable ready pods.\
Not available for PodClique or LeaderWorkerSet.\
When nil, the field is omitted from the API response.
**Validation:** Minimum: 0 \
Optional: \{\} \
ScheduledReplicas is the number of replicas the backend scheduler has\
scheduled, in Dynamo component-replica units. Optional; omitted (nil)\
when the backend cannot derive it reliably. A nil value means "not\
reported", never "zero scheduled".
**Validation:** Minimum: 0 \
Optional: \{\} \
#### SharedMemorySpec
**Kind:** `type`
**Appears in:** [DynamoCheckpointJobConfig](#dynamocheckpointjobconfig), [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Disabled, when true, opts out of mounting a shared-memory medium for the\
component. When false (or unset), shared memory is enabled and Size is\
required (enforced by the validating webhook). Size is ignored when\
Disabled is true.
See [Quantity](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#quantity-resource-api).
#### SpecTopologyConstraint
SpecTopologyConstraint defines deployment-level topology placement requirements.
It carries both the topology profile (which ClusterTopology CR to use) and an
optional default pack domain that services without their own constraint inherit.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentSpec](#dynamographdeploymentspec)
TopologyProfile is the name of the ClusterTopology CR that defines the\
topology hierarchy for this deployment.
**Validation:** MinLength: 1 \
PackDomain is the default topology domain to pack pods within.\
Optional — omit when only services carry constraints.
See [TopologyDomain](#topologydomain).
**Validation:** Pattern: `^[a-z0-9]([a-z0-9-]*[a-z0-9])?$` \
Optional: \{\} \
#### TopologyConstraint
TopologyConstraint defines service-level topology placement requirements.
The topology profile is inherited from the deployment-level SpecTopologyConstraint;
only the pack domain is specified here.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
PackDomain is the topology domain to pack pods within. Must match a\
domain defined in the referenced ClusterTopology CR.
See [TopologyDomain](#topologydomain).
**Validation:** Pattern: `^[a-z0-9]([a-z0-9-]*[a-z0-9])?$` \
#### TopologyDomain
TopologyDomain is a free-form topology level identifier.
Common examples: "region", "zone", "datacenter", "block", "rack", "host", "numa".
When used with a ClusterTopology CR, domain names are defined in the CR's
hierarchy; when used with `spec.experimental.kvTransferPolicy.labelKey`
alone, the value is a user-chosen logical name for the topology level.
Must match `^[a-z0-9]([a-z0-9-]*[a-z0-9])?$` (lowercase alphanumeric,
may contain hyphens but must not start or end with one).
**Kind:** `type`
**Underlying type:** `string`
**Validation:** Pattern: `^[a-z0-9]([a-z0-9-]*[a-z0-9])?$`
**Appears in:** [KvTransferPolicy](#kvtransferpolicy), [SpecTopologyConstraint](#spectopologyconstraint), [TopologyConstraint](#topologyconstraint)
#### VolumeMount
VolumeMount references a PVC defined at the top level for volumes to be mounted by the component
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#dynamocomponentdeploymentspec)
Name references a PVC name defined in the top-level PVCs map
**Validation:** Required: \{\} \
MountPoint specifies where to mount the volume.\
If useAsCompilationCache is true and mountPoint is not specified,\
a backend-specific default will be used.
UseAsCompilationCache indicates this volume should be used as a compilation cache.\
When true, backend-specific environment variables will be set and default mount points may be used.
## nvidia.com/v1beta1
Package v1beta1 contains API Schema definitions for the nvidia.com v1beta1 API group.
**Resource Types**
#### [DynamoComponentDeployment](#v1beta1-dynamocomponentdeployment)
#### [DynamoGraphDeployment](#v1beta1-dynamographdeployment)
#### [DynamoGraphDeploymentRequest](#v1beta1-dynamographdeploymentrequest)
#### [DynamoGraphDeploymentScalingAdapter](#v1beta1-dynamographdeploymentscalingadapter)
#### BackendType
BackendType specifies the inference backend.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[auto sglang trtllm vllm]
**Appears in:** [DynamoGraphDeploymentRequestSpec](#v1beta1-dynamographdeploymentrequestspec)
**Allowed values**
* auto
* sglang
* trtllm
* vllm
#### CheckpointDeletionPolicy
CheckpointDeletionPolicy defines what happens to DGD-managed automatic
checkpoint resources when the owning DGD is deleted.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Delete Retain]
**Appears in:** [ComponentCheckpointConfig](#componentcheckpointconfig)
**Allowed values**
* Delete CheckpointDeletionPolicyDelete deletes DGD-managed automatic checkpoint\
CRs and artifacts when the owning DGD is deleted.\
* Retain CheckpointDeletionPolicyRetain keeps DGD-managed automatic checkpoint CRs\
and artifacts after the owning DGD is deleted. Users can reference the\
retained checkpoint with checkpointRef if they accept compatibility risk.\
#### CheckpointMode
Deprecated: use checkpoint.enabled instead.
enabled=true without checkpointRef creates a DGD-managed automatic
checkpoint; checkpointRef restores the named checkpoint.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Auto Manual]
**Appears in:** [ComponentCheckpointConfig](#componentcheckpointconfig)
**Allowed values**
* Auto Deprecated: use checkpoint.enabled=true and omit checkpointRef.\
* Manual Deprecated: use checkpointRef to restore an existing checkpoint.\
#### CheckpointStartupPolicy
CheckpointStartupPolicy defines when worker pods should wait for a checkpoint.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Immediate WaitForCheckpoint]
**Appears in:** [ComponentCheckpointConfig](#componentcheckpointconfig)
**Allowed values**
* Immediate CheckpointStartupPolicyImmediate starts workers immediately. The checkpoint\
job runs in the background, and only pods created after the checkpoint is\
Ready are restore-shaped by the pod-create mutating webhook.\
* WaitForCheckpoint CheckpointStartupPolicyWaitForCheckpoint gates worker replicas until the\
component's checkpoint is Ready, then starts them from the checkpoint.\
#### CompilationCacheConfig
CompilationCacheConfig configures a PVC-backed compilation cache for a component.
The operator handles backend-specific mount paths and environment variables so
users do not need to hand-wire them into the pod template.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#v1beta1-dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#v1beta1-dynamocomponentdeploymentspec)
pvcName references a user-created PVC by name. The PVC must exist in\
the same namespace as the DynamoGraphDeployment.
**Validation:** MinLength: 1 \
Required: \{\} \
mountPath overrides the backend-specific default mount path. When\
empty, the operator selects a default appropriate for the backend\
framework.
**Validation:** Optional: \{\} \
#### ComponentCheckpointConfig
ComponentCheckpointConfig configures checkpointing for a DGD component.
**Kind:** `type`
**Appears in:** [ExperimentalSpec](#experimentalspec)
enabled indicates whether checkpointing is enabled for this component.\
When true, omit checkpointRef for a DGD-managed automatic checkpoint or\
set checkpointRef to restore an existing checkpoint. Omit the checkpoint\
block, or set enabled=false, to disable checkpointing.
**Validation:** Required: \{\} \
Deprecated: omit mode. Use enabled=true without checkpointRef for a\
DGD-managed automatic checkpoint, or use checkpointRef to restore the\
named checkpoint.
See [CheckpointMode](#v1beta1-checkpointmode).
**Validation:** Enum: \[Auto Manual] \
Optional: \{\} \
startupPolicy defines when normal worker replicas are started relative to\
automatic checkpoint readiness.\
`Immediate` (default): start workers cold immediately; later Pods restore\
from the checkpoint once it is Ready.\
`WaitForCheckpoint`: keep worker replicas at zero until the checkpoint is\
Ready, then start them from the checkpoint.
See [CheckpointStartupPolicy](#v1beta1-checkpointstartuppolicy).
**Validation:** Enum: \[Immediate WaitForCheckpoint] \
Optional: \{\} \
DeletionPolicy defines whether a DGD-managed automatic checkpoint CR and\
artifact are deleted or retained when the owning DGD is deleted.\
Explicit checkpointRef checkpoints are never owned or deleted by the DGD.
See [CheckpointDeletionPolicy](#v1beta1-checkpointdeletionpolicy).
**Validation:** Enum: \[Delete Retain] \
Optional: \{\} \
checkpointRef references an existing DynamoCheckpoint CR by `metadata.name`.\
When set, this component's `identity` is ignored and the referenced\
checkpoint is used directly.
**Validation:** Optional: \{\} \
Deprecated: omit for DGD-managed checkpoints; no action is needed.\
Use checkpointRef to restore an existing checkpoint.
See [DynamoCheckpointIdentity](#v1beta1-dynamocheckpointidentity).
**Validation:** Optional: \{\} \
targetContainerName is the workload container to snapshot and restore.
**Validation:** MaxLength: 63 \
MinLength: 1 \
Pattern: `^[a-z0-9]([-a-z0-9]*[a-z0-9])?$` \
Optional: \{\} \
job customizes the DGD-managed checkpoint Job.
See [ComponentCheckpointJobConfig](#componentcheckpointjobconfig).
**Validation:** Optional: \{\} \
#### ComponentCheckpointJobConfig
ComponentCheckpointJobConfig customizes the checkpoint Job created for a DGD component.
**Kind:** `type`
**Appears in:** [ComponentCheckpointConfig](#componentcheckpointconfig)
gmsClientContainers lists checkpoint Job containers that should receive\
GMS client wiring. Requires gpuMemoryService on the component.
**Validation:** items:MaxLength: 63 \
items:MinLength: 1 \
items:Pattern: `^[a-z0-9]([-a-z0-9]*[a-z0-9])?$` \
Optional: \{\} \
podTemplate customizes the checkpoint Job pod. The operator starts from the\
selected workload container and merges this template so users can add helper\
containers such as gms-saver.
See [PodTemplateSpec](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#podtemplatespec-v1-core).
**Validation:** Schemaless: \{\} \
Type: object \
Optional: \{\} \
#### ComponentCheckpointStatus
ComponentCheckpointStatus contains checkpoint information for a single component.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentStatus](#v1beta1-dynamographdeploymentstatus)
checkpointName is the name of the associated DynamoCheckpoint CR.
**Validation:** Optional: \{\} \
checkpointID is the artifact ID used by the snapshot protocol.
**Validation:** Optional: \{\} \
identityHash is the computed hash of the checkpoint identity.\
Deprecated: automatic checkpoints use checkpointID. This field is retained\
for older status consumers.
**Validation:** Optional: \{\} \
ready indicates the checkpoint artifact is ready for future pods to restore.
**Validation:** Optional: \{\} \
#### ComponentKind
ComponentKind represents the type of underlying Kubernetes resource backing a DGD component.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[PodClique PodCliqueScalingGroup Deployment LeaderWorkerSet]
**Appears in:** [ComponentReplicaStatus](#componentreplicastatus)
**Allowed values**
* PodClique
* PodCliqueScalingGroup
* Deployment
* LeaderWorkerSet
#### ComponentReplicaStatus
ComponentReplicaStatus contains replica information for a single component.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentStatus](#v1beta1-dynamographdeploymentstatus)
componentKind is the underlying resource kind (e.g. `PodClique`,\
`Deployment`, `LeaderWorkerSet`).
See [ComponentKind](#v1beta1-componentkind).
**Validation:** Enum: \[PodClique PodCliqueScalingGroup Deployment LeaderWorkerSet] \
componentNames is the list of underlying Kubernetes resource names for\
this Dynamo component. During normal operation this contains a single\
name; during rolling updates it contains both old and new resource names.
**Validation:** Optional: \{\} \
runtimeNamespace is the effective Dynamo runtime namespace for this\
component. Worker components may include a generation suffix; non-workers and\
Grove-backed workers use the base namespace. During rolling updates, worker\
status keeps the old active revision namespace until cutover completes.
**Validation:** Optional: \{\} \
replicas is the total number of non-terminated replicas.
**Validation:** Minimum: 0 \
updatedReplicas is the number of replicas at the current/desired revision.
**Validation:** Minimum: 0 \
readyReplicas is the number of ready replicas. Populated for\
`PodClique`, `Deployment`, and `LeaderWorkerSet`; not available for\
`PodCliqueScalingGroup`.
**Validation:** Minimum: 0 \
Optional: \{\} \
availableReplicas is the number of available replicas. Populated for\
`Deployment` and `PodCliqueScalingGroup`; not available for\
`PodClique` or `LeaderWorkerSet`.
**Validation:** Minimum: 0 \
Optional: \{\} \
scheduledReplicas is the number of replicas the backend scheduler has\
scheduled, expressed strictly in Dynamo component-replica units (not\
raw backend pod counts). It is a diagnostic aid for distinguishing\
capacity/scheduling shortfalls from runtime readiness.\
It is optional and omitted (nil) when the active backend cannot derive\
it reliably in component-replica units — for example before the backing\
resource's status has been observed, or for backends that do not report\
a scheduling count. A nil value therefore means "not reported", never\
"zero scheduled"; consumers must not treat absence as a scheduling\
failure.
**Validation:** Minimum: 0 \
Optional: \{\} \
#### ComponentType
ComponentType identifies the role of a Dynamo component within a graph.
In v1beta1 this is a strict enum. Unlike v1alpha1 (where `subComponentType`
was used as a workaround for disaggregated serving), `prefill` and `decode`
are first-class values: users can set them directly and downstream consumers
(e.g., the EPP) can filter on the pod label `nvidia.com/dynamo-component-type`.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[frontend worker prefill decode planner epp]
**Appears in:** [DynamoComponentDeploymentSharedSpec](#v1beta1-dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#v1beta1-dynamocomponentdeploymentspec)
**Allowed values**
* frontend
* worker
* prefill
* decode
* planner
* epp
#### DGDRPhase
DGDRPhase represents the lifecycle phase of a DynamoGraphDeploymentRequest.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Pending Profiling Ready Deploying Deployed Failed]
**Appears in:** [DynamoGraphDeploymentRequestStatus](#v1beta1-dynamographdeploymentrequeststatus)
**Allowed values**
* Pending
* Profiling
* Ready
* Deploying
* Deployed
* Failed
#### DGDState
DGDState is the high-level lifecycle state of a DynamoGraphDeployment.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[initializing pending successful failed]
**Appears in:** [DynamoGraphDeploymentStatus](#v1beta1-dynamographdeploymentstatus)
**Allowed values**
* initializing
* pending
* successful
* failed
#### DeploymentInfoStatus
DeploymentInfoStatus tracks the state of the deployed DynamoGraphDeployment.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestStatus](#v1beta1-dynamographdeploymentrequeststatus)
Replicas is the desired number of replicas.
**Validation:** Optional: \{\} \
AvailableReplicas is the number of replicas that are available and ready.
**Validation:** Optional: \{\} \
#### DynamoCheckpointIdentity
Deprecated: omit in DGD component checkpoint configs. Auto needs no
replacement; use checkpointRef for explicit restores.
Duplicated from v1alpha1; DynamoCheckpoint itself remains v1alpha1.
**Kind:** `type`
**Appears in:** [ComponentCheckpointConfig](#componentcheckpointconfig)
model is the model identifier (e.g. "meta-llama/Llama-3-70B").\
Deprecated: legacy identity only.
**Validation:** MinLength: 1 \
Required: \{\} \
backendFramework is the runtime framework (`vllm`, `sglang`, `trtllm`).\
Deprecated: legacy identity only.
**Validation:** Enum: \[vllm sglang trtllm] \
Required: \{\} \
dynamoVersion is the Dynamo platform version.\
Deprecated: legacy identity only.
**Validation:** Optional: \{\} \
tensorParallelSize is the tensor parallel configuration.\
Deprecated: checkpoint launch uses the pod template instead.
**Validation:** Minimum: 1 \
Optional: \{\} \
pipelineParallelSize is the pipeline parallel configuration.\
Deprecated: checkpoint launch uses the pod template instead.
**Validation:** Minimum: 1 \
Optional: \{\} \
dtype is the data type (`fp16`, `bf16`, `fp8`, etc.).\
Deprecated: legacy identity only.
**Validation:** Optional: \{\} \
maxModelLen is the maximum sequence length.\
Deprecated: legacy identity only.
**Validation:** Minimum: 1 \
Optional: \{\} \
extraParameters are additional parameters that affect the checkpoint hash.\
Deprecated: legacy identity only.
**Validation:** Optional: \{\} \
#### DynamoComponentDeployment
DynamoComponentDeployment is the Schema for the dynamocomponentdeployments API.
v1beta1 is a served version: the API server accepts reads and writes
against it, and transparently converts to/from v1alpha1 (still the
storage version until a later MR flips it). Conversion goes through the
operator's conversion webhook; see api/v1alpha1/\*\_conversion.go.
**Kind:** `resource`
`nvidia.com/v1beta1`
`DynamoComponentDeployment`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
spec defines the desired state for this Dynamo component deployment.
See [DynamoComponentDeploymentSpec](#v1beta1-dynamocomponentdeploymentspec).
#### DynamoComponentDeploymentSharedSpec
DynamoComponentDeploymentSharedSpec is the shared configuration used by both
standalone DCDs and by the components embedded in a DynamoGraphDeployment.
In v1beta1 the ten per-component pod-configuration fields that existed in
v1alpha1 (resources, envs, envFromSecret, livenessProbe, readinessProbe,
volumeMounts, annotations, labels, extraPodMetadata, extraPodSpec) are
replaced with a single `podTemplate` field holding a native
`corev1.PodTemplateSpec`. The operator injects its defaults into the
container named `"main"` and merges user overrides using strategic-merge-by-name
semantics. Users can add sidecars, init containers, and pod-level configuration
directly in `podTemplate` without any `extraPodSpec`-style escape hatch.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSpec](#v1beta1-dynamocomponentdeploymentspec), [DynamoGraphDeploymentSpec](#v1beta1-dynamographdeploymentspec)
name is the stable logical identifier for this component within its\
DynamoGraphDeployment. It must be unique within the parent's\
`spec.components` list.\
For standalone DynamoComponentDeployment objects, the defaulting webhook\
populates `name` from `metadata.name` on admission, so users\
typically do not need to set it explicitly.\
`name` is decoupled from the underlying Kubernetes resource name so that\
the operator can rename child workloads (e.g. suffixing worker DCDs with\
a hash during rolling updates) without losing the stable identity that\
downstream consumers (labels, status maps, DGDSA references, planner\
RBAC, EPP filters) depend on.
**Validation:** MaxLength: 63 \
MinLength: 1 \
Pattern: `^[A-Za-z0-9]([-A-Za-z0-9]*[A-Za-z0-9])?$` \
Required: \{\} \
type indicates the role of this component within a Dynamo graph. Drives\
port mapping, frontend detection, planner RBAC, and the pod label\
`nvidia.com/dynamo-component-type`. Because `prefill` and `decode` are\
first-class values, users can set them directly.
See [ComponentType](#componenttype).
**Validation:** Enum: \[frontend worker prefill decode planner epp] \
Optional: \{\} \
globalDynamoNamespace places the component in the global Dynamo\
namespace rather than the per-deployment namespace derived from the\
DGD name.
**Validation:** Optional: \{\} \
podTemplate is the pod template used to create the component's pods.\
The operator injects its defaults (image, command, env, ports, probes,\
resources, volume mounts) into the container named `"main"` inside\
`podTemplate.spec.containers`, merging user overrides by name. If no\
container named `"main"` is present, the operator auto-generates it\
with standard defaults. All other containers in `podTemplate.spec.containers`\
are treated as user-managed sidecars: the operator does not inject\
defaults into them, so sidecars must specify required fields (e.g. `image`)\
themselves. The validation webhook rejects pod templates where a\
non-`"main"` container is missing a required field such as `image`.
See [PodTemplateSpec](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#podtemplatespec-v1-core).
**Validation:** Optional: \{\} \
replicas is the desired number of Pods for this component. When\
`scalingAdapter` is set on this component, this field is managed by\
the DynamoGraphDeploymentScalingAdapter and should not be modified\
directly.
**Validation:** Minimum: 0 \
Optional: \{\} \
minAvailable maps to Grove PodClique minAvailable for single-node and\
Grove PodCliqueScalingGroup minAvailable for multi-node components.\
This field determines 1) the minimum number of replicas guaranteed to be\
gang-scheduled, and 2) when violating minAvailable replicas triggers gang\
termination.\
For Grove-backed DynamoGraphDeployment components, minAvailable defaults to\
1 when omitted and is immutable after creation. Positive replica counts must\
be greater than or equal to minAvailable. Replicas may be scaled to 0 as a\
special scale-to-zero state; minAvailable remains configured but is not\
enforced again until replicas is scaled back to a positive value.\
For non-Grove deployments, setting this field will result in a validation error.
**Validation:** Minimum: 1 \
Optional: \{\} \
multinode configures multinode components.
See [MultinodeSpec](#v1beta1-multinodespec).
**Validation:** Optional: \{\} \
sharedMemorySize controls the size of the tmpfs mounted at `/dev/shm`.\
`nil` selects the operator default (8Gi), a positive quantity sets a\
custom size, and `"0"` disables the shared-memory volume entirely.\
Simpler replacement for v1alpha1's `SharedMemorySpec` struct with its\
`disabled bool` + `size Quantity` pattern.
See [Quantity](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#quantity-resource-api).
**Validation:** Optional: \{\} \
modelRef references a model served by this component. When specified,\
a headless service is created for endpoint discovery.
See [ModelReference](#v1beta1-modelreference).
**Validation:** Optional: \{\} \
scalingAdapter opts this component into using the\
DynamoGraphDeploymentScalingAdapter. When set (even as an empty object,\
`scalingAdapter: \{\}`), a DGDSA is created and owns the `replicas` field\
so that external autoscalers (HPA/KEDA/Planner) can drive scaling via\
the Scale subresource. Omit the field to opt out.
See [ScalingAdapter](#v1beta1-scalingadapter).
**Validation:** Optional: \{\} \
eppConfig holds EPP-specific configuration for Endpoint Picker Plugin\
components. Only meaningful when `type` is `epp`.
See [EPPConfig](#v1beta1-eppconfig).
**Validation:** Optional: \{\} \
frontendSidecar optionally designates a container in\
`podTemplate.spec.containers` as the frontend sidecar. The value must\
match the `name` of a container in that list; the operator merges its\
frontend-sidecar defaults (auto-generated Dynamo env vars, ports,\
health probes) into that container the same way it merges into `"main"`.\
The full container definition (image, args, envFrom, env) lives in\
`podTemplate` -- this eliminates the redundant `image`, `args`,\
`envFromSecret`, and `envs` fields from v1alpha1's `FrontendSidecarSpec`.\
The validation webhook rejects values that do not match any container\
name in `podTemplate.spec.containers`.
**Validation:** Optional: \{\} \
compilationCache configures a PVC-backed compilation cache. The operator\
handles backend-specific mount paths and environment variables, so\
users do not need to hand-wire them into `podTemplate`. Extracted from\
v1alpha1's `volumeMount.useAsCompilationCache` flag.
See [CompilationCacheConfig](#compilationcacheconfig).
**Validation:** Optional: \{\} \
topologyConstraint applies to this component.\
`topologyConstraint.packDomain` is required. When both this and\
`spec.topologyConstraint.packDomain` are set, this field's `packDomain`\
must be narrower than or equal to the spec-level value.
See [TopologyConstraint](#v1beta1-topologyconstraint).
**Validation:** Optional: \{\} \
experimental groups opt-in preview features whose API shape and\
behavior may change in breaking ways between v1beta1 releases,\
including disappearing without a name-preserving graduation path.\
In v1beta1 this block holds `gpuMemoryService` and `failover` (which\
remain tightly coupled -- failover requires GMS -- and are expected to\
evolve together as the DRA-based GPU sharing story matures), and\
`checkpoint` (whose API shape is still settling). Fields here are\
explicitly NOT covered by the normal v1beta1 deprecation policy; do not\
depend on them for production workloads.
See [ExperimentalSpec](#experimentalspec).
**Validation:** Optional: \{\} \
#### DynamoComponentDeploymentSpec
DynamoComponentDeploymentSpec defines the desired state of a DynamoComponentDeployment.
**Kind:** `type`
**Appears in:** [DynamoComponentDeployment](#v1beta1-dynamocomponentdeployment)
backendFramework specifies the backend framework.
**Validation:** Enum: \[sglang vllm trtllm] \
name is the stable logical identifier for this component within its\
DynamoGraphDeployment. It must be unique within the parent's\
`spec.components` list.\
For standalone DynamoComponentDeployment objects, the defaulting webhook\
populates `name` from `metadata.name` on admission, so users\
typically do not need to set it explicitly.\
`name` is decoupled from the underlying Kubernetes resource name so that\
the operator can rename child workloads (e.g. suffixing worker DCDs with\
a hash during rolling updates) without losing the stable identity that\
downstream consumers (labels, status maps, DGDSA references, planner\
RBAC, EPP filters) depend on.
**Validation:** MaxLength: 63 \
MinLength: 1 \
Pattern: `^[A-Za-z0-9]([-A-Za-z0-9]*[A-Za-z0-9])?$` \
Required: \{\} \
type indicates the role of this component within a Dynamo graph. Drives\
port mapping, frontend detection, planner RBAC, and the pod label\
`nvidia.com/dynamo-component-type`. Because `prefill` and `decode` are\
first-class values, users can set them directly.
See [ComponentType](#componenttype).
**Validation:** Enum: \[frontend worker prefill decode planner epp] \
Optional: \{\} \
globalDynamoNamespace places the component in the global Dynamo\
namespace rather than the per-deployment namespace derived from the\
DGD name.
**Validation:** Optional: \{\} \
podTemplate is the pod template used to create the component's pods.\
The operator injects its defaults (image, command, env, ports, probes,\
resources, volume mounts) into the container named `"main"` inside\
`podTemplate.spec.containers`, merging user overrides by name. If no\
container named `"main"` is present, the operator auto-generates it\
with standard defaults. All other containers in `podTemplate.spec.containers`\
are treated as user-managed sidecars: the operator does not inject\
defaults into them, so sidecars must specify required fields (e.g. `image`)\
themselves. The validation webhook rejects pod templates where a\
non-`"main"` container is missing a required field such as `image`.
See [PodTemplateSpec](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#podtemplatespec-v1-core).
**Validation:** Optional: \{\} \
replicas is the desired number of Pods for this component. When\
`scalingAdapter` is set on this component, this field is managed by\
the DynamoGraphDeploymentScalingAdapter and should not be modified\
directly.
**Validation:** Minimum: 0 \
Optional: \{\} \
minAvailable maps to Grove PodClique minAvailable for single-node and\
Grove PodCliqueScalingGroup minAvailable for multi-node components.\
This field determines 1) the minimum number of replicas guaranteed to be\
gang-scheduled, and 2) when violating minAvailable replicas triggers gang\
termination.\
For Grove-backed DynamoGraphDeployment components, minAvailable defaults to\
1 when omitted and is immutable after creation. Positive replica counts must\
be greater than or equal to minAvailable. Replicas may be scaled to 0 as a\
special scale-to-zero state; minAvailable remains configured but is not\
enforced again until replicas is scaled back to a positive value.\
For non-Grove deployments, setting this field will result in a validation error.
**Validation:** Minimum: 1 \
Optional: \{\} \
multinode configures multinode components.
See [MultinodeSpec](#v1beta1-multinodespec).
**Validation:** Optional: \{\} \
sharedMemorySize controls the size of the tmpfs mounted at `/dev/shm`.\
`nil` selects the operator default (8Gi), a positive quantity sets a\
custom size, and `"0"` disables the shared-memory volume entirely.\
Simpler replacement for v1alpha1's `SharedMemorySpec` struct with its\
`disabled bool` + `size Quantity` pattern.
See [Quantity](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#quantity-resource-api).
**Validation:** Optional: \{\} \
modelRef references a model served by this component. When specified,\
a headless service is created for endpoint discovery.
See [ModelReference](#v1beta1-modelreference).
**Validation:** Optional: \{\} \
scalingAdapter opts this component into using the\
DynamoGraphDeploymentScalingAdapter. When set (even as an empty object,\
`scalingAdapter: \{\}`), a DGDSA is created and owns the `replicas` field\
so that external autoscalers (HPA/KEDA/Planner) can drive scaling via\
the Scale subresource. Omit the field to opt out.
See [ScalingAdapter](#v1beta1-scalingadapter).
**Validation:** Optional: \{\} \
eppConfig holds EPP-specific configuration for Endpoint Picker Plugin\
components. Only meaningful when `type` is `epp`.
See [EPPConfig](#v1beta1-eppconfig).
**Validation:** Optional: \{\} \
frontendSidecar optionally designates a container in\
`podTemplate.spec.containers` as the frontend sidecar. The value must\
match the `name` of a container in that list; the operator merges its\
frontend-sidecar defaults (auto-generated Dynamo env vars, ports,\
health probes) into that container the same way it merges into `"main"`.\
The full container definition (image, args, envFrom, env) lives in\
`podTemplate` -- this eliminates the redundant `image`, `args`,\
`envFromSecret`, and `envs` fields from v1alpha1's `FrontendSidecarSpec`.\
The validation webhook rejects values that do not match any container\
name in `podTemplate.spec.containers`.
**Validation:** Optional: \{\} \
compilationCache configures a PVC-backed compilation cache. The operator\
handles backend-specific mount paths and environment variables, so\
users do not need to hand-wire them into `podTemplate`. Extracted from\
v1alpha1's `volumeMount.useAsCompilationCache` flag.
See [CompilationCacheConfig](#compilationcacheconfig).
**Validation:** Optional: \{\} \
topologyConstraint applies to this component.\
`topologyConstraint.packDomain` is required. When both this and\
`spec.topologyConstraint.packDomain` are set, this field's `packDomain`\
must be narrower than or equal to the spec-level value.
See [TopologyConstraint](#v1beta1-topologyconstraint).
**Validation:** Optional: \{\} \
experimental groups opt-in preview features whose API shape and\
behavior may change in breaking ways between v1beta1 releases,\
including disappearing without a name-preserving graduation path.\
In v1beta1 this block holds `gpuMemoryService` and `failover` (which\
remain tightly coupled -- failover requires GMS -- and are expected to\
evolve together as the DRA-based GPU sharing story matures), and\
`checkpoint` (whose API shape is still settling). Fields here are\
explicitly NOT covered by the normal v1beta1 deprecation policy; do not\
depend on them for production workloads.
See [ExperimentalSpec](#experimentalspec).
**Validation:** Optional: \{\} \
#### DynamoGraphDeployment
DynamoGraphDeployment is the Schema for the dynamographdeployments API.
v1beta1 is a served version: the API server accepts reads and writes
against it, and transparently converts to/from v1alpha1 (still the
storage version until a later MR flips it). Conversion goes through the
operator's conversion webhook; see api/v1alpha1/\*\_conversion.go.
**Kind:** `resource`
`nvidia.com/v1beta1`
`DynamoGraphDeployment`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
spec defines the desired state for this graph deployment.
See [DynamoGraphDeploymentSpec](#v1beta1-dynamographdeploymentspec).
status reflects the current observed state of this graph deployment.
See [DynamoGraphDeploymentStatus](#v1beta1-dynamographdeploymentstatus).
#### DynamoGraphDeploymentComponentRef
DynamoGraphDeploymentComponentRef identifies a specific component within a
DynamoGraphDeployment. Renamed from v1alpha1's `DynamoGraphDeploymentServiceRef`
to align with the v1beta1 `services -> components` and
`serviceName -> componentName` renames.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentScalingAdapterSpec](#v1beta1-dynamographdeploymentscalingadapterspec)
name is the `metadata.name` of the target DynamoGraphDeployment.
**Validation:** MinLength: 1 \
Required: \{\} \
componentName is the `componentName` of the entry within the target\
DGD's `spec.components` list to scale.
**Validation:** MinLength: 1 \
Required: \{\} \
#### DynamoGraphDeploymentExperimentalSpec
DynamoGraphDeploymentExperimentalSpec groups graph-level opt-in preview
features whose API shape and behavior may change in breaking ways between
v1beta1 releases. Component-level experimental features live under
`spec.components[*].experimental`.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentSpec](#v1beta1-dynamographdeploymentspec)
kvTransferPolicy configures topology-aware routing for KV-cache\
transfers between prefill and decode workers.
See [KvTransferPolicy](#v1beta1-kvtransferpolicy).
**Validation:** Optional: \{\} \
#### v1beta1 DynamoGraphDeploymentRequest
DynamoGraphDeploymentRequest is the Schema for the dynamographdeploymentrequests API.
It provides a simplified, SLA-driven interface for deploying inference models on Dynamo.
Users specify a model and optional performance targets; the controller handles profiling,
configuration selection, and deployment.
Lifecycle:
1. Pending: Spec validated, preparing for profiling
2. Profiling: Profiling job is running to discover optimal configurations
3. Ready: Profiling complete, generated DGD spec available in status
4. Deploying: DGD is being created and rolled out (when autoApply=true)
5. Deployed: DGD is running and healthy
6. Failed: An unrecoverable error occurred
**Kind:** `resource`
`nvidia.com/v1beta1`
`DynamoGraphDeploymentRequest`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
Spec defines the desired state for this deployment request.
See [DynamoGraphDeploymentRequestSpec](#v1beta1-dynamographdeploymentrequestspec).
Status reflects the current observed state of this deployment request.
See [DynamoGraphDeploymentRequestStatus](#v1beta1-dynamographdeploymentrequeststatus).
#### v1beta1 DynamoGraphDeploymentRequestSpec
DynamoGraphDeploymentRequestSpec defines the desired state of a DynamoGraphDeploymentRequest.
Only the Model field is required; all other fields are optional and have sensible defaults.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequest](#v1beta1-dynamographdeploymentrequest)
Model specifies the model to deploy (e.g., "Qwen/Qwen3-0.6B", "meta-llama/Llama-3-70b").\
Can be a HuggingFace ID or a private model name.
**Validation:** MinLength: 1 \
Required: \{\} \
Backend specifies the inference backend to use for profiling and deployment.
See [BackendType](#backendtype).
**Validation:** Enum: \[auto sglang trtllm vllm] \
Optional: \{\} \
Image is the container image reference for the profiling job (planner image).\
Example: "nvcr.io/nvidia/ai-dynamo/dynamo-planner:1.2.1".\
For Dynamo \< 1.1.0, use dynamo-frontend.
**Validation:** Optional: \{\} \
ModelCache provides optional PVC configuration for pre-downloaded model weights.\
When provided, weights are loaded from the PVC instead of downloading from HuggingFace.
See [ModelCacheSpec](#modelcachespec).
**Validation:** Optional: \{\} \
Hardware describes the hardware resources available for profiling and deployment.\
Typically auto-filled by the operator from cluster discovery.
See [HardwareSpec](#hardwarespec).
**Validation:** Optional: \{\} \
Workload defines the expected workload characteristics for SLA-based profiling.
See [WorkloadSpec](#workloadspec).
**Validation:** Optional: \{\} \
SLA defines service-level agreement targets that drive profiling optimization.
See [SLASpec](#slaspec).
**Validation:** Optional: \{\} \
Overrides allows customizing the profiling job and the generated DynamoGraphDeployment.
See [OverridesSpec](#overridesspec).
**Validation:** Optional: \{\} \
Features controls optional Dynamo platform features in the generated deployment.
See [FeaturesSpec](#featuresspec).
**Validation:** Optional: \{\} \
SearchStrategy controls the profiling search depth.\
"rapid" performs a fast sweep; "thorough" explores more configurations.
See [SearchStrategy](#searchstrategy).
**Validation:** Enum: \[rapid thorough] \
Optional: \{\} \
AutoApply indicates whether to automatically create a DynamoGraphDeployment\
after profiling completes. If false, the generated spec is stored in status\
for manual review and application.
**Validation:** Optional: \{\} \
#### v1beta1 DynamoGraphDeploymentRequestStatus
DynamoGraphDeploymentRequestStatus represents the observed state of a DynamoGraphDeploymentRequest.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequest](#v1beta1-dynamographdeploymentrequest)
Phase is the high-level lifecycle phase of the deployment request.
See [DGDRPhase](#dgdrphase).
**Validation:** Enum: \[Pending Profiling Ready Deploying Deployed Failed] \
Optional: \{\} \
ProfilingPhase indicates the current sub-phase of the profiling pipeline.\
Only meaningful when Phase is "Profiling". Cleared when profiling completes or fails.
See [ProfilingPhase](#profilingphase).
**Validation:** Enum: \[Initializing SweepingPrefill SweepingDecode SelectingConfig BuildingCurves GeneratingDGD Done] \
Optional: \{\} \
DGDName is the name of the generated or created DynamoGraphDeployment.
**Validation:** Optional: \{\} \
ProfilingJobName is the name of the Kubernetes Job running the profiler.
**Validation:** Optional: \{\} \
Conditions contains the latest observed conditions of the deployment request.\
Standard condition types include: Succeeded, Validation, Profiling, SpecGenerated, DeploymentReady.
See [Condition](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#condition-v1-meta).
**Validation:** Optional: \{\} \
ProfilingResults contains the selected deployment configuration produced by profiling.\
Deprecated compatibility fields may remain on objects created by older releases.
See [ProfilingResultsStatus](#profilingresultsstatus).
**Validation:** Optional: \{\} \
DeploymentInfo tracks the state of the deployed DynamoGraphDeployment.\
Populated when a DGD has been created (either via autoApply or manually).
See [DeploymentInfoStatus](#deploymentinfostatus).
**Validation:** Optional: \{\} \
ObservedGeneration is the most recent generation observed by the controller.
**Validation:** Optional: \{\} \
#### DynamoGraphDeploymentScalingAdapter
DynamoGraphDeploymentScalingAdapter provides a scaling interface for individual
components within a DynamoGraphDeployment. It implements the Kubernetes scale
subresource, enabling integration with HPA, KEDA, and custom autoscalers.
The adapter acts as an intermediary between autoscalers and the DGD,
ensuring that only the adapter controller modifies the DGD's component replicas.
This prevents conflicts when multiple autoscaling mechanisms are in play.
v1alpha1 remains the storage version; conversion between served versions is
handled by the operator's conversion webhook
(see api/v1alpha1/dynamographdeploymentscalingadapter\_conversion.go).
**Kind:** `resource`
`nvidia.com/v1beta1`
`DynamoGraphDeploymentScalingAdapter`
Refer to Kubernetes API documentation for fields of `metadata`.
See [ObjectMeta](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#objectmeta-v1-meta).
See [DynamoGraphDeploymentScalingAdapterSpec](#v1beta1-dynamographdeploymentscalingadapterspec).
See [DynamoGraphDeploymentScalingAdapterStatus](#v1beta1-dynamographdeploymentscalingadapterstatus).
#### DynamoGraphDeploymentScalingAdapterSpec
DynamoGraphDeploymentScalingAdapterSpec defines the desired state of a
DynamoGraphDeploymentScalingAdapter.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentScalingAdapter](#v1beta1-dynamographdeploymentscalingadapter)
replicas is the desired number of replicas for the target component.\
This field is modified by external autoscalers (HPA/KEDA/Planner) or\
manually by users.
**Validation:** Minimum: 0 \
Required: \{\} \
dgdRef references the DynamoGraphDeployment and the specific component to scale.
See [DynamoGraphDeploymentComponentRef](#dynamographdeploymentcomponentref).
**Validation:** Required: \{\} \
#### DynamoGraphDeploymentScalingAdapterStatus
DynamoGraphDeploymentScalingAdapterStatus defines the observed state of a
DynamoGraphDeploymentScalingAdapter.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentScalingAdapter](#v1beta1-dynamographdeploymentscalingadapter)
replicas is the current number of replicas for the target component.\
This is synced from the DGD's component replicas and is required for\
the scale subresource.
selector is a label selector string for the pods managed by this\
adapter. Required for HPA compatibility via the scale subresource.
**Validation:** Optional: \{\} \
lastScaleTime is the last time the adapter scaled the target component.
See [Time](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#time-v1-meta).
**Validation:** Optional: \{\} \
#### DynamoGraphDeploymentSpec
DynamoGraphDeploymentSpec defines the desired state of a DynamoGraphDeployment.
**Kind:** `type`
**Appears in:** [DynamoGraphDeployment](#v1beta1-dynamographdeployment)
annotations to propagate to all child resources (PCS, DCD, Deployments,\
and pod templates). Component-level (`podTemplate`) values take precedence\
on conflict.
**Validation:** Optional: \{\} \
labels to propagate to all child resources. Same precedence rules as `annotations`.
**Validation:** Optional: \{\} \
priorityClassName is the name of the PriorityClass to use for Grove PodCliqueSets.\
Requires the Grove pathway.
**Validation:** Optional: \{\} \
components are the components deployed as part of this graph. Each entry\
carries its own stable logical `name`, and names must be unique within\
the list. Component types are generally repeatable, except `type: epp`\
which may appear at most once.
See [DynamoComponentDeploymentSharedSpec](#v1beta1-dynamocomponentdeploymentsharedspec).
**Validation:** MaxItems: 25 \
Optional: \{\} \
env is prepended to every component's environment. Component-specific\
env entries with the same name take precedence and may reference values\
from this list.
See [EnvVar](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#envvar-v1-core).
**Validation:** Optional: \{\} \
backendFramework specifies the backend framework (e.g. "sglang", "vllm", "trtllm").
**Validation:** Enum: \[sglang vllm trtllm] \
restart specifies the restart policy for the graph deployment.
See [Restart](#v1beta1-restart).
**Validation:** Optional: \{\} \
topologyConstraint is the deployment-level topology constraint. When\
set, `spec.topologyConstraint.clusterTopologyName` names the ClusterTopology\
CR to use. `spec.topologyConstraint.packDomain` is optional at this\
level and can be omitted when only components carry constraints.\
Components without their own `topologyConstraint` inherit from this value.
See [SpecTopologyConstraint](#v1beta1-spectopologyconstraint).
**Validation:** Optional: \{\} \
experimental groups graph-level preview features whose API shape and\
behavior may change in breaking ways between v1beta1 releases.
See [DynamoGraphDeploymentExperimentalSpec](#v1beta1-dynamographdeploymentexperimentalspec).
**Validation:** Optional: \{\} \
#### DynamoGraphDeploymentStatus
DynamoGraphDeploymentStatus defines the observed state of a DynamoGraphDeployment.
Unchanged between v1alpha1 and v1beta1.
**Kind:** `type`
**Appears in:** [DynamoGraphDeployment](#v1beta1-dynamographdeployment)
observedGeneration is the most recent generation observed by the controller.
**Validation:** Optional: \{\} \
state is a high-level textual status of the graph deployment lifecycle.
See [DGDState](#v1beta1-dgdstate).
**Validation:** Enum: \[initializing pending successful failed] \
conditions contains the latest observed conditions of the graph deployment.\
Merged by type on patch updates.
See [Condition](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#condition-v1-meta).
**Validation:** Optional: \{\} \
components contains per-component replica status information, keyed by component name.
See [ComponentReplicaStatus](#componentreplicastatus).
**Validation:** Optional: \{\} \
restart contains the status of a graph-level restart.
See [RestartStatus](#v1beta1-restartstatus).
**Validation:** Optional: \{\} \
checkpoints contains per-component checkpoint status, keyed by component name.
See [ComponentCheckpointStatus](#componentcheckpointstatus).
**Validation:** Optional: \{\} \
rollingUpdate tracks the progress of operator-managed rolling updates.\
Currently only supported for single-node, non-Grove deployments (DCD/Deployment).
See [RollingUpdateStatus](#v1beta1-rollingupdatestatus).
**Validation:** Optional: \{\} \
placement groups DGD-level scheduler placement signals (score, reporting\
state, and any future placement fields).
See [PlacementStatus](#v1beta1-placementstatus).
**Validation:** Optional: \{\} \
#### EPPConfig
EPPConfig contains configuration for EPP (Endpoint Picker Plugin) components.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#v1beta1-dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#v1beta1-dynamocomponentdeploymentspec)
configMapRef references a user-provided ConfigMap containing EPP\
configuration. Mutually exclusive with `config`.
See [ConfigMapKeySelector](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#configmapkeyselector-v1-core).
**Validation:** Optional: \{\} \
config allows specifying EPP `EndpointPickerConfig` directly as a\
structured object. The operator marshals this to YAML and creates a\
ConfigMap automatically. Mutually exclusive with `configMapRef`. One of\
`configMapRef` or `config` must be specified.
**Validation:** Type: object \
Optional: \{\} \
#### ExperimentalSpec
ExperimentalSpec groups opt-in preview features whose API shape and behavior
may change in breaking ways between v1beta1 releases (including disappearing
without a name-preserving graduation path). Fields placed under
`experimental` are explicitly NOT covered by the normal v1beta1 deprecation
policy and should not be relied on for production workloads. Features
graduate out of this block (and become first-class fields on the shared
spec) once their API is considered stable.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#v1beta1-dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#v1beta1-dynamocomponentdeploymentspec)
gpuMemoryService configures the GPU Memory Service (GMS). When set, GPU\
access for GMS clients is managed via DRA.
See [GPUMemoryServiceSpec](#v1beta1-gpumemoryservicespec).
**Validation:** Optional: \{\} \
failover configures active-passive GPU failover for this component.\
Requires `gpuMemoryService` to also be set, and `failover.mode` must\
match `gpuMemoryService.mode` (enforced by the validation webhook).
See [FailoverSpec](#v1beta1-failoverspec).
**Validation:** Optional: \{\} \
checkpoint configures container-image snapshotting and restore for\
this component. Set `checkpoint.enabled: true` to opt in. Without\
checkpointRef, the DGD controller creates a DGD-scoped DynamoCheckpoint\
CR and later restores pods in the same DGD generation from that\
checkpoint. With checkpointRef, the DGD restores from that existing\
checkpoint instead. The user-facing shape of this field is still settling,\
which is why it lives under `experimental` in v1beta1 instead of at the\
top level.
See [ComponentCheckpointConfig](#componentcheckpointconfig).
**Validation:** Optional: \{\} \
#### FailoverSpec
FailoverSpec configures active-passive failover for a worker component.
The main container is cloned into two engine containers (active + standby)
sharing GPUs via DRA, and the standby acquires the flock when the active
engine fails. Failover requires that gpuMemoryService is also set, and that
failover.mode matches gpuMemoryService.mode. Also requires the
`nvidia.com/dynamo-kube-discovery-mode: container` annotation on the DGD.
See ExperimentalSpec for the stability caveat.
**Kind:** `type`
**Appears in:** [ExperimentalSpec](#experimentalspec)
mode selects the failover deployment topology. Must match\
`spec.experimental.gpuMemoryService.mode` (or\
`spec.components[*].experimental.gpuMemoryService.mode` inside a\
DynamoGraphDeployment).
See [GPUMemoryServiceMode](#v1beta1-gpumemoryservicemode).
**Validation:** Enum: \[IntraPod InterPod] \
Optional: \{\} \
numShadows is the number of shadow (standby) engine containers per\
rank. Reserved for future use; the operator currently creates exactly\
one shadow.
**Validation:** Maximum: 1 \
Minimum: 1 \
Optional: \{\} \
#### FeaturesSpec
FeaturesSpec controls optional Dynamo platform features in the generated deployment.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestSpec](#v1beta1-dynamographdeploymentrequestspec)
Planner contains the raw Planner configuration passed to the Planner service.\
Its schema is defined by dynamo.planner.config.planner\_config.PlannerConfig.\
See [Planner Guide](/dynamo/dev/knowledge-base/modular-components/planner/planner-guide#plannerconfig-reference).\
DGDR passes this object through without field-level validation; the Planner\
service validates it at startup.\
The presence of this field (non-null) enables the planner in the generated DGD.
See [RawExtension](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#rawextension-runtime-pkg).
**Validation:** Type: object \
Optional: \{\} \
Mocker configures the simulated (mocker) backend for testing without GPUs.
See [MockerSpec](#mockerspec).
**Validation:** Optional: \{\} \
#### GMSClientPodSpec
GMSClientPodSpec declares an additional GMS client pod for inter-pod GMS.
**Kind:** `type`
**Appears in:** [GPUMemoryServiceSpec](#v1beta1-gpumemoryservicespec)
name identifies this client pod.
**Validation:** MaxLength: 63 \
MinLength: 1 \
Pattern: `^[a-z0-9]([-a-z0-9]*[a-z0-9])?$` \
podTemplate configures the pod to run as a GMS client.
See [PodTemplateSpec](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#podtemplatespec-v1-core).
**Validation:** Schemaless: \{\} \
Type: object \
#### GPUMemoryServiceMode
GPUMemoryServiceMode selects the GMS deployment topology.
**Kind:** `enum`
**Underlying type:** `string`
**Appears in:** [FailoverSpec](#v1beta1-failoverspec), [GPUMemoryServiceSpec](#v1beta1-gpumemoryservicespec)
**Allowed values**
* IntraPod GMSModeIntraPod runs GMS as a sidecar within the same pod.\
* InterPod GMSModeInterPod runs GMS as rank-local pods that share GPUs through DRA.\
Extra client pod rendering is reserved for a follow-up change.\
#### GPUMemoryServiceSpec
GPUMemoryServiceSpec configures the GPU Memory Service (GMS) for a
worker component. The operator injects GMS wiring and replaces the main
container's GPU resources with a DRA `ResourceClaim` for shared GPU access.
See ExperimentalSpec for the stability caveat.
**Kind:** `type`
**Appears in:** [ExperimentalSpec](#experimentalspec)
mode selects the GMS deployment topology.
See [GPUMemoryServiceMode](#v1beta1-gpumemoryservicemode).
**Validation:** Enum: \[IntraPod InterPod] \
Optional: \{\} \
deviceClassName is the DRA `DeviceClass` to request GPUs from.
**Validation:** Optional: \{\} \
extraClientContainers lists additional user-declared containers that should\
be wired as GMS clients in service pods. Checkpoint Job clients are declared\
under checkpoint.job.gmsClientContainers. In each rendered pod, only\
matching container names are wired; absent names are ignored.
**Validation:** items:MaxLength: 63 \
items:MinLength: 1 \
items:Pattern: `^[a-z0-9]([-a-z0-9]*[a-z0-9])?$` \
Optional: \{\} \
extraClientPods declares additional GMS client pods for inter-pod GMS. This field is\
reserved for future use and is rejected until inter-pod client orchestration is wired.
See [GMSClientPodSpec](#v1beta1-gmsclientpodspec).
**Validation:** Optional: \{\} \
#### GPUSKUType
GPUSKUType is the AIC hardware system identifier for a supported GPU.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[gb200\_sxm gb10 b200\_sxm h200\_sxm h100\_sxm h100\_pcie a100\_sxm a100\_pcie a30 l40s l40 l4 v100\_sxm v100\_pcie t4 mi200 mi300]
**Appears in:** [HardwareSpec](#hardwarespec)
**Allowed values**
* gb200\_sxm --- Blackwell ---\
* gb10
* b200\_sxm
* h200\_sxm --- Hopper ---\
* h100\_sxm
* h100\_pcie
* a100\_sxm --- Ampere ---\
* a100\_pcie
* a30
* l40s --- Ada ---\
* l40
* l4
* v100\_sxm --- Older NVIDIA ---\
* v100\_pcie
* t4
* mi200 --- AMD ---\
* mi300
#### HardwareSpec
HardwareSpec describes the GPU hardware for profiling and deployment.
All fields are auto-detected from cluster GPU nodes when omitted
(requires cluster-wide mode with GPU discovery enabled).
gpuSku is a selector (restricts which nodes are considered);
the other fields are pure overrides passed to the profiler.
If all four fields are set, discovery is skipped.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestSpec](#v1beta1-dynamographdeploymentrequestspec)
GPUSKU selects the GPU type to target.\
When omitted, auto-detected by selecting the GPU with the highest\
node count, then highest VRAM. In mixed-GPU clusters, set this to\
choose which GPU type to use. Discovery and totalGpus are then\
restricted to nodes matching this SKU.
See [GPUSKUType](#gpuskutype).
**Validation:** Enum: \[gb200\_sxm gb10 b200\_sxm h200\_sxm h100\_sxm h100\_pcie a100\_sxm a100\_pcie a30 l40s l40 l4 v100\_sxm v100\_pcie t4 mi200 mi300] \
Optional: \{\} \
VRAMMB is the VRAM per GPU in MiB.\
When omitted, auto-detected from cluster GPU nodes.
**Validation:** Optional: \{\} \
TotalGPUs is the GPU budget for profiling and deployment.\
The profiler uses this to determine parallelism and replica count.\
When omitted, computed by counting GPUs on discovered nodes\
(filtered by gpuSku when set), temporarily capped at 32 to\
limit profiler search space. This cap may be removed in a future\
release. Set this field explicitly to override.
**Validation:** Optional: \{\} \
NumGPUsPerNode is the number of GPUs per node.\
When omitted, auto-detected from cluster GPU nodes.
**Validation:** Optional: \{\} \
Interconnect describes the primary GPU-to-GPU interconnect *within a node*.\
Semantics / usage:\
- This is capability metadata used for profiling, planning, and deployment decisions.\
- It does NOT configure or enable any GPU interconnect; it only describes what is available/assumed.\
- When omitted, the operator may attempt best-effort discovery (currently distinguishes "nvlink"\
vs "pcie" based on DCGM NVLink link count). If discovery is unavailable, it may remain empty.\
Impact of wrong / missing values:\
- If set more optimistically than reality (e.g., "nvlink" when only PCIe is present), performance\
models may overestimate intra-node bandwidth and choose overly aggressive parallelism or layouts,\
resulting in degraded performance compared to expectations.\
- If set more pessimistically than reality (e.g., "pcie" when NVLink is present), the system may\
choose conservative plans and leave performance on the table.\
- If unset and undiscovered, consumers should treat the interconnect as unknown and fall back to\
conservative assumptions.\
Example values: "pcie", "nvlink". Other values may be accepted but may not be auto-detected.
**Validation:** Optional: \{\} \
RDMA indicates whether the cluster has RDMA-capable networking available for Dynamo data movement.\
Semantics / usage:\
- This is capability metadata used for profiling, planning, and deployment decisions.\
- It does NOT install, enable, or configure RDMA (e.g., drivers, SR-IOV, NVIDIA network operator,\
GPUDirect settings). It only expresses availability/intent.\
- When omitted, the operator may attempt best-effort discovery (e.g., via node labels indicating\
RDMA/SR-IOV capability and/or presence of NVIDIA network-operator RDMA components). If discovery\
is unavailable, it may remain unset.\
Impact of wrong / missing values:\
- False positive (set true when RDMA is not actually usable end-to-end) may cause plans or\
deployments to assume RDMA is available; depending on the runtime transport selection and\
fallback behavior, this can lead to connection/setup failures or performance regressions.\
- False negative (set false when RDMA is available) will typically avoid RDMA-optimized paths and\
fall back to non-RDMA transports, usually remaining functional but potentially slower.\
- If unset and undiscovered, consumers should treat RDMA availability as unknown and use\
conservative defaults / fallback transports.
**Validation:** Optional: \{\} \
#### KvTransferEnforcement
KvTransferEnforcement controls how the selected prefill worker's topology is
applied to decode routing.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[required preferred]
**Appears in:** [KvTransferPolicy](#v1beta1-kvtransferpolicy)
**Allowed values**
* required KvTransferEnforcementRequired enforces same-domain decode worker\
selection.\
* preferred KvTransferEnforcementPreferred biases decode worker selection toward the\
same domain.\
#### KvTransferPolicy
KvTransferPolicy configures topology-aware routing for KV-cache transfers
between prefill and decode workers. This is a graph-wide concern placed
under `spec.experimental` while the API is incubating.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentExperimentalSpec](#v1beta1-dynamographdeploymentexperimentalspec)
clusterTopologyName references a Grove ClusterTopology CR. The operator\
reads the CR's topology levels and projects them through Dynamo-owned pod\
labels for worker topology metadata.
**Validation:** MinLength: 1 \
Optional: \{\} \
labelKey is a Kubernetes node label key (e.g.\
"topology.kubernetes.io/zone") whose value identifies the topology\
domain for each worker. The operator copies the node label onto worker\
pods so the runtime can publish it as worker metadata. The label\
should correspond to the topology level named in `domain`.
**Validation:** MaxLength: 317 \
MinLength: 1 \
Pattern: `^(([a-z0-9]([-a-z0-9]\{0,61\}[a-z0-9])?)(\.[a-z0-9]([-a-z0-9]\{0,61\}[a-z0-9])?)*/)?([A-Za-z0-9]([-A-Za-z0-9_.]\{0,61\}[A-Za-z0-9])?)$` \
Optional: \{\} \
domain is the logical name for the topology level to enforce\
(e.g. "zone", "rack"). The router uses this to match workers that\
share the same value for the label identified by `labelKey`.
See [TopologyDomain](#v1beta1-topologydomain).
**Validation:** Pattern: `^[a-z0-9]([a-z0-9-]*[a-z0-9])?$` \
enforcement controls how the selected prefill worker's topology is\
applied to decode routing. "required" only allows decode workers in the\
same topology domain as the selected prefill worker. "preferred" keeps\
all decode workers eligible, but biases selection toward workers in the\
same topology domain. Defaults to "required".
See [KvTransferEnforcement](#v1beta1-kvtransferenforcement).
**Validation:** Enum: \[required preferred] \
Optional: \{\} \
preferredWeight is required and used only when enforcement is\
"preferred". Higher values create a stronger same-domain routing\
preference, but do not guarantee same-domain selection. The value is not\
a probability; worker selection still depends on load and other routing\
inputs. A value of 0 disables the topology preference; 1 is the strongest\
supported preference.
**Validation:** Maximum: 1 \
Minimum: 0 \
Optional: \{\} \
#### MockerSpec
MockerSpec configures the simulated (mocker) backend.
**Kind:** `type`
**Appears in:** [FeaturesSpec](#featuresspec)
Enabled indicates whether to deploy mocker workers instead of real inference workers.\
Useful for large-scale testing without GPUs.
**Validation:** Optional: \{\} \
#### ModelCacheSpec
ModelCacheSpec references a PVC containing pre-downloaded model weights.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestSpec](#v1beta1-dynamographdeploymentrequestspec)
PVCName is the name of the PersistentVolumeClaim containing model weights.\
The PVC must exist in the same namespace as the DGDR.
**Validation:** Optional: \{\} \
PVCModelPath is the path to the model checkpoint directory within the PVC\
(e.g. "deepseek-r1" or "models/Llama-3.1-405B-FP8").
**Validation:** Optional: \{\} \
PVCMountPath is the mount path for the PVC inside the container.
**Validation:** Optional: \{\} \
#### ModelReference
ModelReference identifies a model served by a component.
When specified, a headless service is created for endpoint discovery.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#v1beta1-dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#v1beta1-dynamocomponentdeploymentspec)
name is the base model identifier (e.g. "llama-3-70b-instruct-v1").
**Validation:** MinLength: 1 \
Required: \{\} \
revision is the model revision/version.
**Validation:** Optional: \{\} \
#### MultinodeSpec
MultinodeSpec configures a multinode component.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#v1beta1-dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#v1beta1-dynamocomponentdeploymentspec)
nodeCount is the number of nodes to deploy for the multinode component.\
Total GPUs used is `nodeCount * container GPU request`.
**Validation:** Minimum: 2 \
Optional: \{\} \
#### OptimizationType
OptimizationType defines the optimization target for SLA-based profiling.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[latency throughput]
**Appears in:** [SLASpec](#slaspec)
**Allowed values**
* latency
* throughput
#### OverridesSpec
OverridesSpec allows customizing the profiling job and the generated DynamoGraphDeployment.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestSpec](#v1beta1-dynamographdeploymentrequestspec)
ProfilingJob allows overriding the profiling Job specification.\
Fields set here are merged into the controller-generated Job spec.
See [JobSpec](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#jobspec-v1-batch).
**Validation:** Optional: \{\} \
DGD provides a partial, versioned DynamoGraphDeployment override for the\
profiler-generated deployment. Set apiVersion to nvidia.com/v1alpha1 or\
nvidia.com/v1beta1 and kind to DynamoGraphDeployment.\
The profiler merges the override using the schema for its declared version.\
If the generated DGD uses another supported version, the complete DGD is\
converted before the merge and converted back afterward. The final DGD\
selected or created by a DGDR is nvidia.com/v1beta1.\
The override can update DGD fields, but topology entries are limited to\
services or components already present in the generated DGD. Metadata labels\
and annotations are merged, metadata.name selects the final DGD name, and\
other identity or runtime metadata is ignored.\
V1alpha1 worker argument lists retain legacy append behavior. V1beta1 follows\
structural schema merge behavior, including map-list merging and atomic-list\
replacement.\
The raw embedded resource preserves either supported schema. The API server\
validates that it has apiVersion and kind; override processing validates the\
DGD kind, supported version, and field schema.
See [RawExtension](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#rawextension-runtime-pkg).
**Validation:** EmbeddedResource: \{\} \
Optional: \{\} \
#### ParetoConfig
ParetoConfig is retained for compatibility with status objects produced by
older profiler releases.
Deprecated: The profiler no longer generates Pareto configurations.
**Kind:** `type`
**Appears in:** [ProfilingResultsStatus](#profilingresultsstatus)
Config is the full deployment configuration for this Pareto point.
See [RawExtension](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#rawextension-runtime-pkg).
**Validation:** Type: object \
#### PlacementScoreState
PlacementScoreState describes whether placement score is available and how
complete the reported score is for a graph deployment.
Every backend must set this field after the first reconciliation:
* Reported: a score is available for every scored placement unit.
* Partial: a score is available for some but not all placement units.
* Unsupported: the backend does not surface a placement score at all.
* Unknown: the backend supports scores but the current value is
indeterminate (e.g. read failure, not yet populated by the
scheduler). When set, PlacementStatus.Score must be cleared.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Reported Partial Unsupported Unknown]
**Appears in:** [PlacementStatus](#v1beta1-placementstatus)
**Allowed values**
* Reported
* Partial
* Unsupported
* Unknown
#### PlacementStatus
PlacementStatus groups DGD-level scheduler placement fields under a single
status object so future placement signals (e.g. scheduler contract version,
last-report timestamp, per-unit reports) can be added without a schema break.
The score source is an open question in DEP #10064 (Grove mirror, typed Grove
scheduler API, or unstructured provider). Until a source is selected and
implemented, the DGD controller does not write this field; the schema and
conversion are landed here so downstream consumers can rely on the shape.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentStatus](#v1beta1-dynamographdeploymentstatus)
score is the DGD-level scheduler placement score aggregated from\
relevant scheduler placement units. Normalized to \[0.0, 1.0] where higher\
is better and 1.0 represents the best possible placement. Aggregation\
uses the minimum across placement units so the value is a worst-placement\
signal for the graph. Scores are only comparable across DGDs that share\
the same scheduler scoring contract and version.
**Validation:** Maximum: 1 \
Minimum: 0 \
Optional: \{\} \
state indicates placement score reporting state. See PlacementScoreState\
for the semantics of each value.
See [PlacementScoreState](#v1beta1-placementscorestate).
**Validation:** Enum: \[Reported Partial Unsupported Unknown] \
Optional: \{\} \
#### ProfilingPhase
ProfilingPhase represents a sub-phase within the profiling pipeline.
When the DGDR Phase is "Profiling", this value indicates which step
of the profiling pipeline is currently executing.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Initializing SweepingPrefill SweepingDecode SelectingConfig BuildingCurves GeneratingDGD Done]
**Appears in:** [DynamoGraphDeploymentRequestStatus](#v1beta1-dynamographdeploymentrequeststatus)
**Allowed values**
* Initializing Profiler is loading the DGD template, detecting GPU hardware,\
and resolving the model architecture from HuggingFace.\
* SweepingPrefill Sweeping parallelization strategies (TP/TEP/DEP) across GPU counts\
for prefill, measuring TTFT at each configuration.\
* SweepingDecode Sweeping parallelization strategies and concurrency levels\
for decode, measuring ITL at each configuration.\
* SelectingConfig Filtering results against SLA targets and selecting the most\
cost-efficient configuration that meets TTFT/ITL requirements.\
* BuildingCurves Building detailed interpolation curves (ISL→TTFT for prefill,\
KV-usage×context-length→ITL for decode) using the selected configs.\
* GeneratingDGD Packaging profiling data into a ConfigMap and generating\
the final DGD YAML with planner integration.\
* Done Profiling pipeline finished successfully.\
#### ProfilingResultsStatus
ProfilingResultsStatus contains the output of the profiling process.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestStatus](#v1beta1-dynamographdeploymentrequeststatus)
Pareto is retained for compatibility with existing status objects.\
Deprecated: The controller no longer populates this field.
See [ParetoConfig](#paretoconfig).
**Validation:** Optional: \{\} \
SelectedConfig is the recommended configuration chosen by the profiler\
based on the SLA targets. This is the configuration used for deployment\
when autoApply is true.
See [RawExtension](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#rawextension-runtime-pkg).
**Validation:** Type: object \
Optional: \{\} \
#### Restart
Restart specifies the restart policy for a graph deployment.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentSpec](#v1beta1-dynamographdeploymentspec)
id is an arbitrary string that triggers a restart when changed. Any\
modification to this value initiates a restart of the graph deployment\
according to the configured strategy.
**Validation:** MinLength: 1 \
Required: \{\} \
strategy specifies the restart strategy for the graph deployment.
See [RestartStrategy](#v1beta1-restartstrategy).
**Validation:** Optional: \{\} \
#### RestartPhase
RestartPhase enumerates phases of a graph-level restart.
**Kind:** `enum`
**Underlying type:** `string`
**Appears in:** [RestartStatus](#v1beta1-restartstatus)
**Allowed values**
* Pending
* Restarting
* Completed
* Failed
* Superseded
#### RestartStatus
RestartStatus contains the status of a graph-level restart.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentStatus](#v1beta1-dynamographdeploymentstatus)
observedID is the restart ID currently being processed. Matches `Restart.id` in the spec.
phase is the phase of the restart.
See [RestartPhase](#v1beta1-restartphase).
inProgress contains the names of the components currently being restarted.
**Validation:** Optional: \{\} \
#### RestartStrategy
RestartStrategy defines how components are restarted.
**Kind:** `type`
**Appears in:** [Restart](#v1beta1-restart)
type specifies the restart strategy type.
See [RestartStrategyType](#v1beta1-restartstrategytype).
**Validation:** Enum: \[Sequential Parallel] \
Optional: \{\} \
order is the complete ordered set of component names for sequential\
restarts. Omit or leave empty to use the controller's default order.\
This field must not be set for parallel restarts.
**Validation:** Optional: \{\} \
#### RestartStrategyType
RestartStrategyType enumerates restart strategies.
**Kind:** `enum`
**Underlying type:** `string`
**Appears in:** [RestartStrategy](#v1beta1-restartstrategy)
**Allowed values**
* Sequential
* Parallel
#### RollingUpdatePhase
RollingUpdatePhase represents the current phase of a rolling update.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[Pending InProgress Completed Failed ]
**Appears in:** [RollingUpdateStatus](#v1beta1-rollingupdatestatus)
**Allowed values**
* Pending
* InProgress
* Completed
* Failed
* \`\`
#### RollingUpdateStatus
RollingUpdateStatus tracks the progress of an operator-managed rolling update.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentStatus](#v1beta1-dynamographdeploymentstatus)
phase indicates the current phase of the rolling update.
See [RollingUpdatePhase](#v1beta1-rollingupdatephase).
**Validation:** Enum: \[Pending InProgress Completed Failed ] \
Optional: \{\} \
startTime is when the rolling update began.
See [Time](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#time-v1-meta).
**Validation:** Optional: \{\} \
endTime is when the rolling update completed (successfully or failed).
See [Time](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#time-v1-meta).
**Validation:** Optional: \{\} \
updatedComponents is the list of components that have completed the\
rolling update.
**Validation:** Optional: \{\} \
#### SLASpec
SLASpec defines the service-level agreement targets for profiling optimization.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestSpec](#v1beta1-dynamographdeploymentrequestspec)
TTFT is the Time To First Token target in milliseconds.
**Validation:** Optional: \{\} \
ITL is the Inter-Token Latency target in milliseconds.
**Validation:** Optional: \{\} \
E2ELatency is the target end-to-end request latency in milliseconds.\
Alternative to specifying TTFT + ITL.
**Validation:** Optional: \{\} \
OptimizationType is the optimization target for SLA profiling.\
Valid values: latency, throughput.
See [OptimizationType](#optimizationtype).
**Validation:** Enum: \[latency throughput] \
Optional: \{\} \
#### ScalingAdapter
ScalingAdapter opts a component into using the DynamoGraphDeploymentScalingAdapter
(DGDSA). When `scalingAdapter` is set on a component (even as an empty
object, `scalingAdapter: {}`), the DGDSA is created and owns the `replicas`
field so that external autoscalers (HPA/KEDA/Planner) can drive scaling via
the Scale subresource. Omitting the field opts the component out.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#v1beta1-dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#v1beta1-dynamocomponentdeploymentspec)
#### SearchStrategy
SearchStrategy controls the profiling search depth.
**Kind:** `enum`
**Underlying type:** `string`
**Validation:** Enum: \[rapid thorough]
**Appears in:** [DynamoGraphDeploymentRequestSpec](#v1beta1-dynamographdeploymentrequestspec)
**Allowed values**
* rapid
* thorough
#### SpecTopologyConstraint
SpecTopologyConstraint defines deployment-level topology placement requirements.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentSpec](#v1beta1-dynamographdeploymentspec)
clusterTopologyName is the name of the ClusterTopology resource that\
defines the topology hierarchy for this deployment.
**Validation:** MinLength: 1 \
packDomain is the default topology domain to pack pods within.\
Optional; omit when only components carry constraints.
See [TopologyDomain](#v1beta1-topologydomain).
**Validation:** Pattern: `^[a-z0-9]([a-z0-9-]*[a-z0-9])?$` \
Optional: \{\} \
#### TopologyConstraint
TopologyConstraint defines component-level topology placement requirements.
The topology profile is inherited from the deployment-level
`SpecTopologyConstraint`.
**Kind:** `type`
**Appears in:** [DynamoComponentDeploymentSharedSpec](#v1beta1-dynamocomponentdeploymentsharedspec), [DynamoComponentDeploymentSpec](#v1beta1-dynamocomponentdeploymentspec)
packDomain is the topology domain to pack pods within. Must match a\
domain defined in the referenced ClusterTopology CR.
See [TopologyDomain](#v1beta1-topologydomain).
**Validation:** Pattern: `^[a-z0-9]([a-z0-9-]*[a-z0-9])?$` \
#### TopologyDomain
TopologyDomain is a free-form topology level identifier.
Common examples: "region", "zone", "datacenter", "block", "rack", "host", "numa".
When used with a ClusterTopology CR, domain names are defined in the CR's
hierarchy; when used with `spec.experimental.kvTransferPolicy.labelKey`
alone, the value is a user-chosen logical name for the topology level.
**Kind:** `type`
**Underlying type:** `string`
**Validation:** Pattern: `^[a-z0-9]([a-z0-9-]*[a-z0-9])?$`
**Appears in:** [KvTransferPolicy](#v1beta1-kvtransferpolicy), [SpecTopologyConstraint](#v1beta1-spectopologyconstraint), [TopologyConstraint](#v1beta1-topologyconstraint)
#### WorkloadSpec
WorkloadSpec defines the workload characteristics for SLA-based profiling.
**Kind:** `type`
**Appears in:** [DynamoGraphDeploymentRequestSpec](#v1beta1-dynamographdeploymentrequestspec)
ISL is the Input Sequence Length (number of tokens).
**Validation:** Optional: \{\} \
OSL is the Output Sequence Length (number of tokens).
**Validation:** Optional: \{\} \
Concurrency is the target concurrency level.\
Required (or RequestRate) when the planner is disabled.
**Validation:** Optional: \{\} \
RequestRate is the target request rate (req/s).\
Required (or Concurrency) when the planner is disabled.
**Validation:** Optional: \{\} \
## operator.config.dynamo.nvidia.com/v1alpha1
**Resource Types**
#### [OperatorConfiguration](#operatorconfiguration)
#### CertProvisionMode
CertProvisionMode controls how webhook TLS certificates are managed.
**Kind:** `enum`
**Underlying type:** `string`
**Appears in:** [WebhookServer](#webhookserver)
**Allowed values**
* auto CertProvisionModeAuto uses the built-in cert-controller to generate and rotate certificates.\
* manual CertProvisionModeManual expects certificates to be provided externally (e.g., cert-manager, admin).\
#### CheckpointConfiguration
CheckpointConfiguration holds checkpoint/restore settings.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
Enabled indicates if checkpoint functionality is enabled
Seccomp controls the localhost seccomp profile applied to checkpoint and\
restore pods. A nil value means "use the default profile"; set\
Seccomp.Disabled=true to disable seccomp injection entirely.
See [CheckpointSeccompConfiguration](#checkpointseccompconfiguration).
Storage optionally configures the namespace-local checkpoint PVC that\
workload pods mount. When omitted, the operator preserves the legacy\
behavior of discovering storage from a snapshot-agent DaemonSet in the\
workload namespace.
See [CheckpointStorageConfiguration](#checkpointstorageconfiguration).
CleanupImage is the image used by best-effort artifact cleanup Jobs for\
automatically-created checkpoints. It must provide a POSIX shell and `rm`.
#### CheckpointOCIConfig
Deprecated: CheckpointOCIConfig is retained for compatibility and ignored by
the current snapshot flow.
**Kind:** `type`
**Appears in:** [CheckpointStorageConfiguration](#checkpointstorageconfiguration)
URI is the legacy OCI URI (oci://registry/repository).
CredentialsSecretRef is the legacy docker config secret name.
#### CheckpointPVCConfig
CheckpointPVCConfig configures the namespace-local PVC mounted into
checkpoint and restore workload pods.
**Kind:** `type`
**Appears in:** [CheckpointStorageConfiguration](#checkpointstorageconfiguration)
PVCName is the PVC name in each workload namespace.
BasePath is the mount path inside checkpoint and restore workload pods.
Create tells the operator to create the PVC in workload namespaces when\
it is missing. When false, the PVC must already exist.
Size is the storage request used when Create is true.
StorageClassName is the optional StorageClass name used when Create is true.
AccessMode is the PVC access mode used when Create is true.
#### CheckpointS3Config
Deprecated: CheckpointS3Config is retained for compatibility and ignored by
the current snapshot flow.
**Kind:** `type`
**Appears in:** [CheckpointStorageConfiguration](#checkpointstorageconfiguration)
URI is the legacy S3 URI (s3://\[endpoint/]bucket/prefix).
CredentialsSecretRef is the legacy credentials secret name.
#### CheckpointSeccompConfiguration
CheckpointSeccompConfiguration controls the localhost seccomp profile applied
to checkpoint and restore pods. The profile blocks io\_uring syscalls (which
CRIU cannot dump). Default behavior (zero-value substruct, or absent
substruct) applies DefaultSeccompProfile. Set Disabled=true on OpenShift
(custom localhost profiles require privileged SCC) or when using a CRIU
build with io\_uring support. Set Profile to override the default path.
**Kind:** `type`
**Appears in:** [CheckpointConfiguration](#checkpointconfiguration)
Disabled, when true, suppresses seccomp profile injection entirely.\
Use this for clusters where custom localhost profiles are not allowed\
(e.g. OpenShift's restricted-v2 SCC) or for CRIU builds that handle\
io\_uring natively.
Profile is the localhost seccomp profile path. Empty falls back to\
DefaultSeccompProfile. Ignored when Disabled is true.
#### CheckpointStorageConfiguration
CheckpointStorageConfiguration configures checkpoint storage for operator
pod mutations. Only PVC storage is implemented today.
**Kind:** `type`
**Appears in:** [CheckpointConfiguration](#checkpointconfiguration)
Type is the storage backend type. Only pvc is implemented today.
PVC configuration for pvc-based settings.
See [CheckpointPVCConfig](#checkpointpvcconfig).
Deprecated: S3 is retained for compatibility and ignored.
See [CheckpointS3Config](#checkpoints3config).
Deprecated: OCI is retained for compatibility and ignored.
See [CheckpointOCIConfig](#checkpointociconfig).
#### DRAConfiguration
DRAConfiguration holds Dynamic Resource Allocation (resource.k8s.io/v1) settings.
NOTE: auto-detection here only verifies that the resource.k8s.io/v1 API is
registered on the apiserver (Kubernetes 1.34+). It does NOT verify that a
GPU-specific DRA resource driver (e.g. nvidia/k8s-dra-driver-gpu) is
installed, that its DeviceClass exists, or that node-level GPU drivers are
compatible. An admin can use `enabled: false` to force-off DRA integration
on clusters where the API is present but the GPU driver stack is not wired
up — this makes the operator fail GMS / inter-pod failover admissions early
with a clear error instead of letting pods Pend with a confusing
"resourceclaim not found" at schedule time.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
Enabled overrides auto-detection of the resource.k8s.io/v1 API.\
nil = auto-detect. Setting true requires detection to also succeed (the\
operator will exit at startup otherwise).
#### DiscoveryBackend
DiscoveryBackend is the type for the discovery backend.
**Kind:** `enum`
**Underlying type:** `string`
**Appears in:** [DiscoveryConfiguration](#discoveryconfiguration)
**Allowed values**
* kubernetes DiscoveryBackendKubernetes is the Kubernetes discovery backend\
* etcd DiscoveryBackendEtcd is the etcd discovery backend\
#### DiscoveryConfiguration
DiscoveryConfiguration holds discovery backend settings.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
Backend is the discovery backend: "kubernetes" or "etcd"
See [DiscoveryBackend](#discoverybackend).
#### GPUConfiguration
GPUConfiguration holds GPU discovery settings.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
DiscoveryEnabled indicates whether GPU discovery is enabled
#### GroveConfiguration
GroveConfiguration holds Grove orchestrator settings.
**Kind:** `type`
**Appears in:** [OrchestratorConfiguration](#orchestratorconfiguration)
Enabled overrides auto-detection. nil = auto-detect.
TerminationDelay configures the termination delay for Grove PodCliqueSets
See [Duration](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#duration-v1-meta).
#### InfrastructureConfiguration
InfrastructureConfiguration holds service mesh and backend addresses.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
NATSAddress is the address of the NATS server
ETCDAddress is the address of the etcd server
ModelExpressURL is the URL of the Model Express server to inject into all pods
PrometheusEndpoint is the URL of the Prometheus endpoint to use for metrics
#### IngressConfiguration
IngressConfiguration holds ingress settings.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
VirtualServiceGateway is the name of the Istio virtual service gateway
ControllerClassName is the ingress controller class name
ControllerTLSSecretName is the TLS secret for the ingress controller
HostSuffix is the suffix for ingress hostnames
#### IstioMeshConfiguration
IstioMeshConfiguration holds Istio-specific mesh settings.
**Kind:** `type`
**Appears in:** [ServiceMeshConfiguration](#servicemeshconfiguration)
TLSMode is the Istio TLS mode for DestinationRules.\
Supported values: "DISABLE", "SIMPLE", "ISTIO\_MUTUAL", "MUTUAL".\
Defaults to "SIMPLE".
InsecureSkipVerify skips TLS certificate verification in DestinationRules.\
Defaults to true (matching upstream GAIE behavior with self-signed certs).
ClientCertificate is the path (in the istio-proxy sidecar's filesystem)\
to the file holding the client-side TLS certificate used for mTLS.\
REQUIRED when TLSMode is "MUTUAL"; ignored for other modes.
PrivateKey is the path (in the istio-proxy sidecar's filesystem) to the\
file holding the client-side TLS private key used for mTLS.\
REQUIRED when TLSMode is "MUTUAL"; ignored for other modes.
CaCertificates is the optional path (in the istio-proxy sidecar's\
filesystem) to the file holding CA certificates used to verify the\
server certificate. Used only when TLSMode is "MUTUAL"; for other modes\
the field is ignored.
#### KaiSchedulerConfiguration
KaiSchedulerConfiguration holds Kai-scheduler settings.
**Kind:** `type`
**Appears in:** [OrchestratorConfiguration](#orchestratorconfiguration)
Enabled overrides auto-detection. nil = auto-detect.
#### LWSConfiguration
LWSConfiguration holds LWS orchestrator settings.
**Kind:** `type`
**Appears in:** [OrchestratorConfiguration](#orchestratorconfiguration)
Enabled overrides auto-detection. nil = auto-detect.
#### LeaderElectionConfiguration
LeaderElectionConfiguration holds leader election settings.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
Enabled enables leader election for controller manager
ID is the leader election resource identity
Namespace is the namespace for the leader election resource
#### LoggingConfiguration
LoggingConfiguration holds logging settings.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
Level is the log level (e.g., "info", "debug")
Format is the log format (e.g., "json", "text")
#### MPIConfiguration
MPIConfiguration holds MPI SSH secret settings.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
SSHSecretName is the name of the secret containing the SSH key for MPI
SSHSecretNamespace is the namespace where the MPI SSH secret is located
#### MetricsServer
MetricsServer extends Server with secure serving option.
**Kind:** `type`
**Appears in:** [ServerConfiguration](#serverconfiguration)
BindAddress is the address the server binds to
Port is the port the server listens on
Secure enables secure serving for the metrics endpoint.\
nil = default to true (secure by default).
#### NamespaceConfiguration
NamespaceConfiguration determines operator namespace mode.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
Restricted enables namespace-restricted mode for development and testing.\
Namespace-restricted mode is not supported for production.
Scope configures the namespace ownership claim in namespace-restricted mode.
See [NamespaceScopeConfiguration](#namespacescopeconfiguration).
#### NamespaceScopeConfiguration
NamespaceScopeConfiguration configures the development/test namespace ownership claim.
**Kind:** `type`
**Appears in:** [NamespaceConfiguration](#namespaceconfiguration)
LeaseDuration is the duration of namespace scope marker lease before expiration
See [Duration](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#duration-v1-meta).
LeaseRenewInterval is the interval for renewing namespace scope marker lease
See [Duration](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#duration-v1-meta).
#### OperatorConfiguration
OperatorConfiguration is the Schema for the operator configuration.
**Kind:** `resource`
`operator.config.dynamo.nvidia.com/v1alpha1`
`OperatorConfiguration`
Server configuration (metrics, health probes, webhooks)
See [ServerConfiguration](#serverconfiguration).
Leader election configuration
See [LeaderElectionConfiguration](#leaderelectionconfiguration).
Namespace configuration (restricted vs cluster-wide)
See [NamespaceConfiguration](#namespaceconfiguration).
Orchestrator configuration with optional overrides
See [OrchestratorConfiguration](#orchestratorconfiguration).
DRA (Dynamic Resource Allocation) settings with optional override
See [DRAConfiguration](#draconfiguration).
Service mesh and infrastructure addresses
See [InfrastructureConfiguration](#infrastructureconfiguration).
Ingress configuration
See [IngressConfiguration](#ingressconfiguration).
ServiceMesh configures automatic generation of service-mesh resources\
(e.g., Istio DestinationRules) for EPP components.
See [ServiceMeshConfiguration](#servicemeshconfiguration).
RBAC configuration for cross-namespace resource management (cluster-wide mode)
See [RBACConfiguration](#rbacconfiguration).
MPI SSH secret configuration
See [MPIConfiguration](#mpiconfiguration).
Checkpoint/restore configuration
See [CheckpointConfiguration](#checkpointconfiguration).
Discovery backend configuration
See [DiscoveryConfiguration](#discoveryconfiguration).
GPU discovery configuration
See [GPUConfiguration](#gpuconfiguration).
Logging configuration
See [LoggingConfiguration](#loggingconfiguration).
HTTP/2 and TLS settings
See [SecurityConfiguration](#securityconfiguration).
#### OrchestratorConfiguration
OrchestratorConfiguration holds orchestrator override settings.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
Grove orchestrator configuration
See [GroveConfiguration](#groveconfiguration).
LWS orchestrator configuration
See [LWSConfiguration](#lwsconfiguration).
KaiScheduler configuration
See [KaiSchedulerConfiguration](#kaischedulerconfiguration).
VolcanoScheduler configuration
See [VolcanoSchedulerConfiguration](#volcanoschedulerconfiguration).
#### RBACConfiguration
RBACConfiguration holds RBAC settings for cluster-wide mode.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
PlannerClusterRoleName is the ClusterRole for planner
DGDRProfilingClusterRoleName is the ClusterRole for DGDR profiling jobs
EPPClusterRoleName is the ClusterRole for EPP
#### SecurityConfiguration
SecurityConfiguration holds HTTP/2 and TLS settings.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
EnableHTTP2 enables HTTP/2 for metrics and webhook servers
#### Server
Server holds a bind address and port.
**Kind:** `type`
**Appears in:** [MetricsServer](#metricsserver), [ServerConfiguration](#serverconfiguration), [WebhookServer](#webhookserver)
BindAddress is the address the server binds to
Port is the port the server listens on
#### ServerConfiguration
ServerConfiguration holds server bind addresses and ports.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
Metrics server configuration
See [MetricsServer](#metricsserver).
Health probe server configuration
See [Server](#server).
Webhook server configuration
See [WebhookServer](#webhookserver).
#### ServiceMeshConfiguration
ServiceMeshConfiguration holds service mesh integration settings.
The operator uses this to generate mesh-specific resources (e.g., Istio
DestinationRules) for EPP components so that sidecar proxies connect
correctly without double-TLS issues.
**Kind:** `type`
**Appears in:** [OperatorConfiguration](#operatorconfiguration)
Enabled overrides service mesh auto-detection. nil = auto-detect.
Provider selects the service mesh implementation. Supported: "istio", "".\
Empty string disables service mesh resource generation.
Istio holds Istio-specific settings. Only used when Provider is "istio".
See [IstioMeshConfiguration](#istiomeshconfiguration).
#### VolcanoSchedulerConfiguration
VolcanoSchedulerConfiguration holds Volcano scheduler settings.
**Kind:** `type`
**Appears in:** [OrchestratorConfiguration](#orchestratorconfiguration)
EXPERIMENTAL: Enabled controls Volcano scheduler integration for Grove PodCliqueSets.
#### WebhookServer
WebhookServer extends Server with host and certificate directory.
**Kind:** `type`
**Appears in:** [ServerConfiguration](#serverconfiguration)
BindAddress is the address the server binds to
Port is the port the server listens on
Host is the address the webhook server binds to
CertDir is the directory containing TLS certificates
CertProvisionMode controls certificate management: "auto" (built-in cert-controller) or "manual" (external)
See [CertProvisionMode](#certprovisionmode).
SecretName is the name of the Kubernetes Secret holding webhook TLS certificates
ServiceName is the name of the Kubernetes Service fronting the webhook server.\
Used to generate certificate SANs. Set by the Helm chart.
## Operator Default Values Injection
The Dynamo operator automatically applies default values to various fields when they are not explicitly specified in your deployments. These defaults include:
* **Health Probes**: Startup, liveness, and readiness probes are configured differently for frontend, worker, and planner components. For example, worker components receive a startup probe with a 2-hour timeout (720 failures × 10 seconds) to accommodate long model loading times.
* **Security Context**: All components receive `fsGroup: 1000` by default to ensure proper file permissions for mounted volumes. This can be overridden via the `extraPodSpec.securityContext` field.
* **Shared Memory**: All components receive an 8Gi shared memory volume mounted at `/dev/shm` by default (can be disabled or resized via the `sharedMemory` field).
* **Environment Variables**: Components automatically receive environment variables like `DYN_NAMESPACE`, `DYN_PARENT_DGD_K8S_NAME`, `DYNAMO_PORT`, and backend-specific variables.
* **Pod Configuration**: Default `terminationGracePeriodSeconds` of 60 seconds and `restartPolicy: Always`.
* **Autoscaling**: When enabled without explicit metrics, defaults to CPU-based autoscaling with 80% target utilization.
* **Backend-Specific Behavior**: For multinode deployments, probes are automatically modified or removed for worker nodes depending on the backend framework (VLLM, SGLang, or TensorRT-LLM).
### Pod Specification Defaults
All components receive the following pod-level defaults unless overridden:
* **`terminationGracePeriodSeconds`**: `60` seconds
* **`restartPolicy`**: `Always`
### Security Context
The operator automatically applies default security context settings to all components to ensure proper file permissions, particularly for mounted volumes:
* **`fsGroup`**: `1000` - Sets the group ownership of mounted volumes and any files created in those volumes
This default ensures that non-root containers can write to mounted volumes (like model caches or persistent storage) without permission issues. The `fsGroup` setting is particularly important for:
* Model downloads and caching
* Compilation cache directories
* Persistent volume claims (PVCs)
* SSH key generation in multinode deployments
#### Overriding Security Context
To override the default security context, specify your own `securityContext` in the `extraPodSpec` of your component:
```yaml
services:
YourWorker:
extraPodSpec:
securityContext:
fsGroup: 2000 # Custom group ID
runAsUser: 1000
runAsGroup: 1000
runAsNonRoot: true
```
**Important**: When you provide *any* `securityContext` object in `extraPodSpec`, the operator will not inject any defaults. This gives you complete control over the security context, including the ability to run as root (by omitting `runAsNonRoot` or setting it to `false`).
#### OpenShift and Security Context Constraints
In OpenShift environments with Security Context Constraints (SCCs), you may need to omit explicit UID/GID values to allow OpenShift's admission controllers to assign them dynamically:
```yaml
services:
YourWorker:
extraPodSpec:
securityContext:
# # Omit fsGroup to let OpenShift assign it based on SCC
# # OpenShift will inject the appropriate UID range
```
Alternatively, if you want to keep the default `fsGroup: 1000` behavior and are certain your cluster allows it, you don't need to specify anything - the operator defaults will work.
### Shared Memory Configuration
Shared memory is enabled by default for all components:
* **Enabled**: `true` (unless explicitly disabled via `sharedMemory.disabled`)
* **Size**: `8Gi`
* **Mount Path**: `/dev/shm`
* **Volume Type**: `emptyDir` with `memory` medium
To disable shared memory or customize the size, use the `sharedMemory` field in your component specification.
### Health Probes by Component Type
The operator applies different default health probes based on the component type.
#### Frontend Components
Frontend components receive the following probe configurations:
**Liveness Probe:**
* **Type**: HTTP GET
* **Path**: `/health`
* **Port**: `http` (8000)
* **Initial Delay**: 60 seconds
* **Period**: 60 seconds
* **Timeout**: 30 seconds
* **Failure Threshold**: 10
**Readiness Probe:**
* **Type**: Exec command
* **Command**: `curl -s http://localhost:${DYNAMO_PORT}/health | jq -e ".status == \"healthy\""`
* **Initial Delay**: 60 seconds
* **Period**: 60 seconds
* **Timeout**: 30 seconds
* **Failure Threshold**: 10
#### Worker Components
Worker components receive the following probe configurations:
**Liveness Probe:**
* **Type**: HTTP GET
* **Path**: `/live`
* **Port**: `system` (9090)
* **Period**: 5 seconds
* **Timeout**: 30 seconds
* **Failure Threshold**: 1
**Readiness Probe:**
* **Type**: HTTP GET
* **Path**: `/health`
* **Port**: `system` (9090)
* **Period**: 10 seconds
* **Timeout**: 30 seconds
* **Failure Threshold**: 60
**Startup Probe:**
* **Type**: HTTP GET
* **Path**: `/live`
* **Port**: `system` (9090)
* **Period**: 10 seconds
* **Timeout**: 5 seconds
* **Failure Threshold**: 720 (allows up to 2 hours for startup: 10s × 720 = 7200s)
:::\{note}
For larger models (typically >70B parameters) or slower storage systems, you may need to increase the `failureThreshold` to allow more time for model loading. Calculate the required threshold based on your expected startup time: `failureThreshold = (expected_startup_seconds / period)`. Override the startup probe in your component specification if the default 2-hour window is insufficient.
:::
#### Multinode Deployment Probe Modifications
For multinode deployments, the operator modifies probes based on the backend framework and node role:
##### VLLM Backend
The operator automatically applies distributed execution configuration based on parallelism settings:
**Tensor/Pipeline Parallel Mode (Recommended)** (when `world_size > GPUs_per_node`):
* Uses PyTorch multiprocessing (mp) backend for distributed execution (`--distributed-executor-backend mp`)
* Supports multi-node deployments with PyTorch's native distributed initialization
* **All nodes**: Run vLLM with proper `--nnodes`, `--node-rank`, `--master-addr` flags injected
* **Probes**: Worker probes adjusted; leader probes remain active
**Ray Backend**:
* Used for use cases such as Elastic EP
* Install with `pip install "ray>=2.55.0"` and configure `--distributed-executor-backend ray`
**Data Parallel Mode** (when `world_size × data_parallel_size > GPUs_per_node`):
* **Worker nodes**: All probes (liveness, readiness, startup) are removed
* **Leader nodes**: All probes remain active
##### SGLang Backend
* **Worker nodes**: All probes (liveness, readiness, startup) are removed
##### TensorRT-LLM Backend
* **Leader nodes**: All probes remain unchanged
* **Worker nodes**:
* Liveness and startup probes are removed
* Readiness probe is replaced with a TCP socket check on SSH port (2222):
* **Initial Delay**: 20 seconds
* **Period**: 20 seconds
* **Timeout**: 5 seconds
* **Failure Threshold**: 10
### Environment Variables
The operator automatically injects environment variables into component containers based on component type, backend framework, and operator configuration. User-provided `envs` values always take precedence over operator defaults.
#### All Components
These environment variables are injected into every component container regardless of type.
| Variable | Purpose | Default | Type | Source |
| ------------------------------ | ----------------------------------------------------------------- | ------------------------------------------------------------------- | -------- | --------------------------------------------------- |
| `DYN_NAMESPACE` | Dynamo service namespace used for service discovery and routing | Derived from DGD spec | `string` | Downward API annotation on checkpoint-restored pods |
| `DYN_COMPONENT` | Identifies the component type for runtime behavior | One of: `frontend`, `worker`, `prefill`, `decode`, `planner`, `epp` | `string` | Set from component spec |
| `DYN_PARENT_DGD_K8S_NAME` | Kubernetes name of the parent DynamoGraphDeployment resource | — | `string` | Set from DGD metadata |
| `DYN_PARENT_DGD_K8S_NAMESPACE` | Kubernetes namespace of the parent DynamoGraphDeployment resource | — | `string` | Set from DGD metadata |
| `POD_NAME` | Current pod name | — | `string` | Downward API (`metadata.name`) |
| `POD_NAMESPACE` | Current pod namespace | — | `string` | Downward API (`metadata.namespace`) |
| `POD_UID` | Current pod UID | — | `string` | Downward API (`metadata.uid`) |
| `DYN_DISCOVERY_BACKEND` | Service discovery backend for inter-component communication | `kubernetes` | `string` | Options: `kubernetes`, `etcd` |
#### Infrastructure (Conditional)
These are injected into all components when the corresponding infrastructure service is configured in the operator's `OperatorConfiguration`.
| Variable | Purpose | Default | Type | Condition |
| --------------------- | ---------------------------------------------- | ------- | -------- | ---------------------------------------------------------- |
| `NATS_SERVER` | NATS messaging server address | — | `string` | Set when `infrastructure.natsAddress` is configured |
| `ETCD_ENDPOINTS` | etcd endpoint addresses for distributed state | — | `string` | Set when `infrastructure.etcdAddress` is configured |
| `MODEL_EXPRESS_URL` | Model Express service URL for model management | — | `string` | Set when `infrastructure.modelExpressURL` is configured |
| `PROMETHEUS_ENDPOINT` | Prometheus endpoint for metrics collection | — | `string` | Set when `infrastructure.prometheusEndpoint` is configured |
#### Frontend Components
| Variable | Purpose | Default | Type |
| ---------------------- | -------------------------------------------------- | ----------------------- | -------- |
| `DYNAMO_PORT` | HTTP port the frontend listens on | `8000` | `int` |
| `DYN_HTTP_PORT` | HTTP port for the frontend service (alias) | `8000` | `int` |
| `DYN_NAMESPACE_PREFIX` | Namespace prefix used for frontend request routing | Same as `DYN_NAMESPACE` | `string` |
#### Worker Components
| Variable | Purpose | Default | Type | |
| --------------------------------------- | ------------------------------------------------------------------------ | -------------- | --------------------- | ------------------------------------------ |
| `DYN_SYSTEM_ENABLED` | Enables the system HTTP server for health checks and metrics | `true` | `string` (boolean) | |
| `DYN_SYSTEM_USE_ENDPOINT_HEALTH_STATUS` | Endpoints whose health status is used for readiness | `["generate"]` | `string` (JSON array) | |
| `DYN_SYSTEM_PORT` | Port for the system HTTP server (health, metrics) | `9090` | `int` | |
| `DYN_HEALTH_CHECK_ENABLED` | Disables the legacy health check mechanism in favor of the system server | `false` | `string` (boolean) | |
| `NIXL_TELEMETRY_ENABLE` | Enables or disables NIXL telemetry collection | `n` | `string` | Options: `y`, `n` |
| `NIXL_TELEMETRY_EXPORTER` | Telemetry exporter format for NIXL metrics | `prometheus` | `string` | |
| `NIXL_TELEMETRY_PROMETHEUS_PORT` | Port for NIXL Prometheus metrics endpoint | `19090` | `int` | |
| `DYN_NAMESPACE_WORKER_SUFFIX` | Hash suffix appended to worker namespace for rolling updates | — | `string` | Only set during rolling update transitions |
#### Planner Components
| Variable | Purpose | Default | Type |
| ------------------------- | -------------------------------------------------- | ------- | ----- |
| `PLANNER_PROMETHEUS_PORT` | Port for the planner's Prometheus metrics endpoint | `9085` | `int` |
#### EPP (Endpoint Picker Plugin) Components
| Variable | Purpose | Default | Type |
| --------------- | ----------------------------------------------------- | ------- | ------------------ |
| `USE_STREAMING` | Enables streaming mode for inference request proxying | `true` | `string` (boolean) |
| `RUST_LOG` | Rust log level and filter configuration | `info` | `string` |
#### VLLM Backend
| Variable | Purpose | Default | Type | Condition |
| ----------------------------- | --------------------------------------------------------- | ------- | -------- | --------------------------------------------------------- |
| `VLLM_CACHE_ROOT` | Directory for vLLM compilation cache artifacts | — | `string` | Set when a volume mount has `useAsCompilationCache: true` |
| `VLLM_NIXL_SIDE_CHANNEL_HOST` | Host IP for the NIXL side channel in multiprocessing mode | Pod IP | `string` | Multinode mp backend only (Downward API: `status.podIP`) |
#### TensorRT-LLM Backend
| Variable | Purpose | Default | Type | Condition |
| ----------------------------------- | ------------------------------------------------------------------------- | ------- | -------- | -------------------------- |
| `OMPI_MCA_orte_keep_fqdn_hostnames` | Instructs OpenMPI to preserve FQDN hostnames for inter-node communication | `1` | `string` | Multinode deployments only |
### Service Accounts
The following component types automatically receive dedicated service accounts:
* **Planner**: `planner-serviceaccount`
* **EPP**: `epp-serviceaccount`
### Image Pull Secrets
The operator automatically discovers and injects image pull secrets for container images. When a component specifies a container image, the operator:
1. Scans all Kubernetes secrets of type `kubernetes.io/dockerconfigjson` in the component's namespace
2. Extracts the docker registry server URLs from each secret's authentication configuration
3. Matches the container image's registry host against the discovered registry URLs
4. Automatically injects matching secrets as `imagePullSecrets` in the pod specification
This eliminates the need to manually specify image pull secrets for each component. The operator maintains an internal index of docker secrets and their associated registries, refreshing this index periodically.
**To disable automatic image pull secret discovery** for a specific component, add the following annotation:
```yaml
annotations:
nvidia.com/disable-image-pull-secret-discovery: "true"
```
### Autoscaling Defaults
When autoscaling is enabled but no metrics are specified, the operator applies:
* **Default Metric**: CPU utilization
* **Target Average Utilization**: `80%`
### Port Configurations
Default container ports are configured based on component type:
#### Frontend Components
* **Port**: 8000
* **Protocol**: TCP
* **Name**: `http`
#### Worker Components
* **Port**: 9090 (system)
* **Protocol**: TCP
* **Name**: `system`
* **Port**: 19090 (NIXL)
* **Protocol**: TCP
* **Name**: `nixl`
#### Planner Components
* **Port**: 9085
* **Protocol**: TCP
* **Name**: `metrics`
#### EPP Components
* **Port**: 9002 (gRPC)
* **Protocol**: TCP
* **Name**: `grpc`
* **Port**: 9003 (gRPC health)
* **Protocol**: TCP
* **Name**: `grpc-health`
* **Port**: 9090 (metrics)
* **Protocol**: TCP
* **Name**: `metrics`
### Backend-Specific Configurations
#### VLLM
* **Ray Head Port**: 6379 (for Ray-based multinode deployments)
* **MP Master Port**: 29500 (for PyTorch distributed multinode TP/PP deployments with mp backend)
* **Data Parallel RPC Port**: 13445 (for data parallel multinode deployments)
#### SGLang
* **Distribution Init Port**: 29500 (for multinode deployments)
#### TensorRT-LLM
* **SSH Port**: 2222 (for multinode MPI communication)
* **OpenMPI Environment**: `OMPI_MCA_orte_keep_fqdn_hostnames=1`
### Implementation Reference
For users who want to understand the implementation details or contribute to the operator, the default values described in this document are set in the following source files:
* **Health Probes, Security Context & Pod Specifications**: [`internal/dynamo/graph.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/dynamo/graph.go) - Contains the main logic for applying default probes, security context, environment variables, shared memory, and pod configurations
* **Component-Specific Defaults**:
* [`internal/dynamo/component_common.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/dynamo/component_common.go) - Base container and pod spec shared by all component types
* [`internal/dynamo/component_frontend.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/dynamo/component_frontend.go)
* [`internal/dynamo/component_worker.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/dynamo/component_worker.go)
* [`internal/dynamo/component_planner.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/dynamo/component_planner.go)
* [`internal/dynamo/component_epp.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/dynamo/component_epp.go)
* **Image Pull Secrets**: [`internal/secrets/docker.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/secrets/docker.go) - Implements the docker secret indexer and automatic discovery
* **Backend-Specific Behavior**:
* [`internal/dynamo/backend_vllm.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/dynamo/backend_vllm.go)
* [`internal/dynamo/backend_sglang.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/dynamo/backend_sglang.go)
* [`internal/dynamo/backend_trtllm.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/dynamo/backend_trtllm.go)
* **Checkpoint / Restore**:
* [`internal/checkpoint/podspec.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/checkpoint/podspec.go) - Checkpoint env var injection and volume setup
* [`internal/checkpoint/resolve.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/checkpoint/resolve.go) - Checkpoint resolution logic
* [`internal/checkpoint/resource.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/checkpoint/resource.go) - Checkpoint resource management
* **Constants & Annotations**: [`internal/consts/consts.go`](https://github.com/ai-dynamo/dynamo/blob/main/deploy/operator/internal/consts/consts.go) - Defines annotation keys and other constants
### Notes
* All these defaults can be overridden by explicitly specifying values in your DynamoComponentDeployment or DynamoGraphDeployment resources
* User-specified probes (via `livenessProbe`, `readinessProbe`, or `startupProbe` fields) take precedence over operator defaults
* For security context, if you provide *any* `securityContext` in `extraPodSpec`, no defaults will be injected, giving you full control
* For multinode deployments, some defaults are modified or removed as described above to accommodate distributed execution patterns
* The `extraPodSpec.mainContainer` field can be used to override probe configurations set by the operator
Raw generated Markdown source: `docs/fern/kubernetes/api-reference.md`