Merge pull request llm-d-incubation#193 from MikeSpreitzer/tweak-policy

🌱 Tweak launcher pod population policy

Author: MikeSpreitzer

api/fma/v1alpha1/launcherpoolpolicy_types.go

@@ -0,0 +1,141 @@
+/*
+Copyright 2025 The llm-d Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+	http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package v1alpha1
+
+import (
+	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+)
+
+// LauncherPopulationPolicy defines the policy for pro-active creation of launcher Pods
+// for a given type of Node.
+// Here we introduce and use a particular definition of "type" for Nodes.
+// All the LauncherPopulationPolicy objects together define a map,
+// from (Node, LauncherConfig) to count.
+// Call this map `PopulationPolicy`.
+// When multiple CountForLauncher apply to the same (Node, LauncherConfig) pair
+// the maximum of their counts is what appears in `PopulationPolicy`.
+// When no CountForLauncher applies to a given (Node, LauncherConfig),
+// `PopulationPolicy` implicitly maps that pair to zero.
+//
+// The collective meaning of all the LauncherPopulationPolicy objects
+// and all the server-requesting Pods is that for a given (Node, LauncherConfig)
+// the number of launchers that should exist is the larger of
+// (a) what `PopulationPolicy` says for that pair, and
+// (b) the number needed to satisfy the server-requesting Pods.
+//
+// +genclient
+// +kubebuilder:object:root=true
+// +kubebuilder:subresource:status
+// +kubebuilder:resource:shortName=lpp
+type LauncherPopulationPolicy struct {
+	metav1.TypeMeta   `json:",inline"`
+	metav1.ObjectMeta `json:"metadata,omitempty"`
+
+	Spec   LauncherPopulationPolicySpec   `json:"spec,omitempty"`
+	Status LauncherPopulationPolicyStatus `json:"status,omitempty"`
+}
+
+// LauncherPopulationPolicySpec defines policy for one type of Node.
+type LauncherPopulationPolicySpec struct {
+	// Selector describes the hardware characteristics of target nodes.
+	//
+	// Introduce an EnhancedNodeSelector that supports combining label-based
+	// matching with resource field conditions.
+	// For example:
+	// enhancedNodeSelector:
+	//  # 1. Label selector (compatible with existing metav1.LabelSelector)
+	//  labelSelector:
+	//    matchLabels:
+	//      nvidia.com/gpu.family: ada-lovelace
+	//    matchExpressions:
+	//      - key: node.kubernetes.io/instance-type
+	//        operator: In
+	//        values: ["gx3-48x240x2l40s", "gx3-96x480x4l40s"]
+	//
+	//  # 2. Resource condition selector (new capability)
+	//  allocatableResources:
+	//    cpu: {min: "16", max: "64"}
+	//    memory: {min: 128Gi, max: 512Gi}
+	//    "nvidia.com/gpu": {min: 8, max: 8}
+	// +required
+	EnhancedNodeSelector EnhancedNodeSelector `json:"enhancedNodeSelector"`
+
+	// CountForLauncher declares the desired number of launchers on the
+	// relevant Node, for various LauncherConfigs.
+	// +required
+	CountForLauncher []CountForLauncher `json:"countForLauncher"`
+}
+
+// EnhancedNodeSelector defines node selector with label selector and resource requirements.
+type EnhancedNodeSelector struct {
+	// LabelSelector defines the label selector for a node.
+	// +required
+	LabelSelector metav1.LabelSelector `json:"labelSelector"`
+	// ResourceRequirements defines the resource requirements for a node.
+	// +optional
+	AllocatableResources ResourceRanges `json:"allocatableResources,omitempty"`
+}
+
+// ResourceRanges defines the required range for some resources.
+// These are the same sort of resources that appear in the `.status.allocatable`
+// of a Node object.
+type ResourceRanges map[corev1.ResourceName]ResourceRange
+
+// ResourceRange defines a range by inclusive minimum and maximum quantity values.
+type ResourceRange struct {
+	// Min specifies the minimum quantity required,
+	// or is `null` to signal that there is no lower bound.
+	// +optional
+	Min *resource.Quantity `json:"min,omitempty"`
+
+	// Max specifies the maximum quantity allowed,
+	// or is `null` to signal that there is no upper bound.
+	// +optional
+	Max *resource.Quantity `json:"max,omitempty"`
+}
+
+type CountForLauncher struct {
+	// LauncherConfigName is the name of the LauncherConfig this policy applies to.
+	// +required
+	LauncherConfigName string `json:"launcherConfigName"`
+
+	// LauncherCount is the total number of launcher pods to maintain.
+	// +required
+	LauncherCount int32 `json:"launcherCount"`
+}
+
+type LauncherPopulationPolicyStatus struct {
+	// `observedGeneration` is the `metadata.generation` last seen by the controller.
+	// +optional
+	ObservedGeneration int64 `json:"observedGeneration,omitempty"`
+
+	// `errors` reports problems seen in the desired state of this object;
+	// in particular, in the version reported by `observedGeneration`.
+	// +optional
+	Errors []string `json:"errors,omitempty"`
+	// Add status fields if needed (e.g., current idle pod counts)
+}
+
+// LauncherPopulationPolicyList contains a list of LauncherPopulationPolicy resources.
+// +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
+type LauncherPopulationPolicyList struct {
+	metav1.TypeMeta `json:",inline"`
+	metav1.ListMeta `json:"metadata,omitempty"`
+	Items           []LauncherPopulationPolicy `json:"items"`
+}