controller

Documentation

Index

• Constants
• func AcceleratorTypesToNodeAffinity(a *v1alpha1.AcceleratorTypes) (*corev1.NodeAffinity, error)
• func InterpolateBaseConfigMap(ctx context.Context, cm *corev1.ConfigMap, msvc *msv1alpha1.ModelService) (*corev1.ConfigMap, error)
• func InterpolateModelService(ctx context.Context, msvc *msv1alpha1.ModelService) (*msv1alpha1.ModelService, error)
• func MergeBackendRefSlices(dest, src []gatewayv1.BackendRef) ([]gatewayv1.BackendRef, error)
• func MergeContainerSlices(dest, src []corev1.Container) ([]corev1.Container, error)
• func MergeGatewayRefSlices(dest, src []gatewayv1.ParentReference) ([]gatewayv1.ParentReference, error)
• type BaseConfig
  • func BaseConfigFromCM(cm *corev1.ConfigMap) (*BaseConfig, error)
  • func (interpolatedBaseConfig *BaseConfig) MergeChildResources(ctx context.Context, modelService *msv1alpha1.ModelService, ...) *BaseConfig
• type ModelServiceReconciler
  • func (r *ModelServiceReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctrl.Result, error)
  • func (r *ModelServiceReconciler) SetupWithManager(mgr ctrl.Manager) error
• type RBACOptions
• type TemplateFuncs
• type TemplateVars
• type URIType
  • func UriType(uri string) URIType

Constants

const DECODE_ROLE = "decode"

const ENV_HF_HOME = "HF_HOME"

const ENV_HF_TOKEN = "HF_TOKEN"

const HF_PREFIX string = "hf://"

const MODEL_ARTIFACT_URI_HF = "hf"

const MODEL_ARTIFACT_URI_HF_PREFIX = MODEL_ARTIFACT_URI_HF + "://"

const MODEL_ARTIFACT_URI_OCI = "oci"

const MODEL_ARTIFACT_URI_OCI_PREFIX = MODEL_ARTIFACT_URI_OCI + "://"

const MODEL_ARTIFACT_URI_PVC = "pvc"

const MODEL_ARTIFACT_URI_PVC_PREFIX = MODEL_ARTIFACT_URI_PVC + "://"

const PREFILL_ROLE = "prefill"

const PVC_PREFIX string = "pvc://"

Functions

func AcceleratorTypesToNodeAffinity

func AcceleratorTypesToNodeAffinity(a *v1alpha1.AcceleratorTypes) (*corev1.NodeAffinity, error)

ToNodeAffinity generates a NodeAffinity rule that requires nodes to match the specified accelerator label key and one of the allowed values.

Returns an error if LabelKey is empty or LabelValues is empty.

func InterpolateBaseConfigMap

func InterpolateBaseConfigMap(ctx context.Context, cm *corev1.ConfigMap, msvc *msv1alpha1.ModelService) (*corev1.ConfigMap, error)

InterpolateBaseConfigMap data strings using msvc template variable values

func InterpolateModelService

func InterpolateModelService(ctx context.Context, msvc *msv1alpha1.ModelService) (*msv1alpha1.ModelService, error)

InterpolateModelService interpolates strings using msvc template variable values

func MergeBackendRefSlices

func MergeBackendRefSlices(dest, src []gatewayv1.BackendRef) ([]gatewayv1.BackendRef, error)

MergeBackendRefSlices merges src slice containing gatewayv1.ParentRefs into dest in place

func MergeContainerSlices

func MergeContainerSlices(dest, src []corev1.Container) ([]corev1.Container, error)

MergeContainerSlices merges src slice into dest in place

func MergeGatewayRefSlices

func MergeGatewayRefSlices(dest, src []gatewayv1.ParentReference) ([]gatewayv1.ParentReference, error)

MergeGatewayRefSlices merges src slice containing gatewayv1.ParentRefs into dest in place

Types

type BaseConfig

type BaseConfig struct {
	ConfigMaps        []corev1.ConfigMap          `json:"configMaps,omitempty"`
	PrefillDeployment *appsv1.Deployment          `json:"prefillDeployment,omitempty"`
	DecodeDeployment  *appsv1.Deployment          `json:"decodeDeployment,omitempty"`
	PrefillService    *corev1.Service             `json:"prefillService,omitempty"`
	DecodeService     *corev1.Service             `json:"decodeService,omitempty"`
	HTTPRoute         *gatewayv1.HTTPRoute        `json:"httpRoute,omitempty"`
	InferencePool     *giev1alpha2.InferencePool  `json:"inferencePool,omitempty"`
	InferenceModel    *giev1alpha2.InferenceModel `json:"inferenceModel,omitempty"`
	EPPDeployment     *appsv1.Deployment          `json:"eppDeployment,omitempty"`
	EPPService        *corev1.Service             `json:"eppService,omitempty"`
	EPPServiceAccount *corev1.ServiceAccount      `json:"eppServiceAccount,omitempty"`
	PDServiceAccount  *corev1.ServiceAccount      `json:"pdServiceAccount,omitempty"`
	EPPRoleBinding    *rbacv1.RoleBinding         `json:"eppRoleBinding,omitempty"`
}

BaseConfig holds information read from the base configmap

func BaseConfigFromCM

func BaseConfigFromCM(cm *corev1.ConfigMap) (*BaseConfig, error)

BaseConfigFromCM returns a BaseConfig object if the input configmap is a valid serialization

func (*BaseConfig) MergeChildResources

func (interpolatedBaseConfig *BaseConfig) MergeChildResources(ctx context.Context, modelService *msv1alpha1.ModelService, scheme *runtime.Scheme, rbacOptions *RBACOptions) *BaseConfig

MergeChildResources merges the MSVC resources into BaseConfig resources merging means MSVC controller is overwriting some fields, such as Name and Namespace for that resource

type ModelServiceReconciler

type ModelServiceReconciler struct {
	RBACOptions RBACOptions
	client.Client
	Scheme *runtime.Scheme
}

ModelServiceReconciler reconciles a ModelService object

func (*ModelServiceReconciler) Reconcile

func (r *ModelServiceReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctrl.Result, error)

For more details, check Reconcile and its Result here: - https://pkg.go.dev/sigs.k8s.io/controller-runtime@v0.20.4/pkg/reconcile

func (*ModelServiceReconciler) SetupWithManager

func (r *ModelServiceReconciler) SetupWithManager(mgr ctrl.Manager) error

SetupWithManager sets up the controller with the Manager.

type RBACOptions

type RBACOptions struct {
	// EPPPullSecrets contains names of epp pull secrets
	// the secrets objects are assumed to be in the controller namespace
	// these are pull secrets used by epp deployment created by the controller
	EPPPullSecrets []string
	// PDPullSecrets contains names of pd pull secrets;
	// the secrets objects are assumed to be in the controller namespace
	// these are pull secrets used by pd deployment created by the controller
	PDPullSecrets []string
	// EPPClusterRole name of the epp cluster role
	// this is a cluster role used in the rolebinding for the epp deployment
	EPPClusterRole string
}

RBACOptions provides the options need to create service accounts and role binding during reconcile

type TemplateFuncs

type TemplateFuncs struct {
	// contains filtered or unexported fields
}

type TemplateVars

type TemplateVars struct {
	ModelServiceName      string `json:"modelServiceName,omitempty"`
	ModelServiceNamespace string `json:"modelServiceNamespace,omitempty"`
	ModelName             string `json:"modelName,omitempty"`
	HFModelName           string `json:"hfModelName,omitempty"`
	SanitizedModelName    string `json:"sanitizedModelName,omitempty"`
	ModelPath             string `json:"modelPath,omitempty"`
	MountedModelPath      string `json:"mountedModelPath,omitempty"`
	AuthSecretName        string `json:"authSecretName,omitempty"`
	EPPServiceName        string `json:"eppServiceName,omitempty"`
	EPPDeploymentName     string `json:"eppDeploymentName,omitempty"`
	PrefillDeploymentName string `json:"prefillDeploymentName,omitempty"`
	DecodeDeploymentName  string `json:"decodeDeploymentName,omitempty"`
	PrefillServiceName    string `json:"prefillServiceName,omitempty"`
	DecodeServiceName     string `json:"decodeServiceName,omitempty"`
	InferencePoolName     string `json:"inferencePoolName,omitempty"`
	InferenceModelName    string `json:"inferenceModelName,omitempty"`
}

Context is intended to be use for interpolating template variables in BaseConfig

type URIType

type URIType string

const (
	PVC        URIType = "pvc"
	HF         URIType = "hf"
	OCI        URIType = "oci"
	UnknownURI URIType = "unknown"
)

func UriType

func UriType(uri string) URIType

UriType returns the type of URI