Refactor decision CRD

api/v1alpha1/decision_types.go

@@ -6,90 +6,67 @@ package v1alpha1
 import (
 	corev1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
-	runtime "k8s.io/apimachinery/pkg/runtime"
 )
 
-type DecisionSpec struct {
-	// SchedulingDomain defines in which scheduling domain this decision
-	// was or is processed (e.g., nova, cinder, manila).
-	SchedulingDomain SchedulingDomain `json:"schedulingDomain"`
+// SchedulingIntents represents the Intent for a scheduling event.
+type SchedulingIntent string
 
-	// A reference to the pipeline that should be used for this decision.
+const (
+	// SchedulingIntentInitialPlacement indicates that this is the initial placement of a resource.
+	SchedulingIntentInitialPlacement SchedulingIntent = "InitialPlacement"
+	// SchedulingIntentLiveMigration  indicates that this scheduling event is triggered by a live migration operation.
+	SchedulingIntentLiveMigration SchedulingIntent = "LiveMigration"
+	// SchedulingIntentResize indicates that this scheduling event is triggered by a resize operation.
+	SchedulingIntentResize SchedulingIntent = "Resize"
+	// SchedulingIntentRebuild indicates that this scheduling event is triggered by a rebuild operation.
+	SchedulingIntentRebuild SchedulingIntent = "Rebuild"
+	// SchedulingIntentEvacuate indicates that this scheduling event is triggered by an evacuate operation.
+	SchedulingIntentEvacuate SchedulingIntent = "Evacuate"
+	// SchedulingIntentUnknown indicates that the Intent for this scheduling event is unknown.
+	SchedulingIntentUnknown SchedulingIntent = "Unknown"
+)
+
+// SchedulingHistoryEntry represents a single entry in the scheduling history of a resource.
+type SchedulingHistoryEntry struct {
+	// The hosts that were selected in this scheduling event, in order of preference.
+	OrderedHosts []string `json:"orderedHosts"`
+	// Timestamp of when the scheduling event occurred.
+	Timestamp metav1.Time `json:"timestamp"`
+	// A reference to the pipeline that was used for this decision.
 	// This reference can be used to look up the pipeline definition and its
 	// scheduler step configuration for additional context.
 	PipelineRef corev1.ObjectReference `json:"pipelineRef"`
+	// The Intent for this scheduling event.
+	Intent SchedulingIntent `json:"intent"`
+}
+
+type DecisionSpec struct {
+	// SchedulingDomain defines in which scheduling domain this decision
+	// was or is processed (e.g., nova, cinder, manila).
+	SchedulingDomain SchedulingDomain `json:"schedulingDomain"`
 
 	// An identifier for the underlying resource to be scheduled.
 	// For example, this can be the UUID of a nova instance or cinder volume.
-	// This can be used to correlate multiple decisions for the same resource.
 	ResourceID string `json:"resourceID"`
-
-	// If the type is "nova", this field contains the raw nova decision request.
-	// +kubebuilder:validation:Optional
-	NovaRaw *runtime.RawExtension `json:"novaRaw,omitempty"`
-	// If the type is "cinder", this field contains the raw cinder decision request.
-	// +kubebuilder:validation:Optional
-	CinderRaw *runtime.RawExtension `json:"cinderRaw,omitempty"`
-	// If the type is "manila", this field contains the raw manila decision request.
-	// +kubebuilder:validation:Optional
-	ManilaRaw *runtime.RawExtension `json:"manilaRaw,omitempty"`
-	// If the type is "machine", this field contains the machine reference.
-	// +kubebuilder:validation:Optional
-	MachineRef *corev1.ObjectReference `json:"machineRef,omitempty"`
-	// If the type is "pod", this field contains the pod reference.
-	// +kubebuilder:validation:Optional
-	PodRef *corev1.ObjectReference `json:"podRef,omitempty"`
-}
-
-type StepResult struct {
-	// object reference to the scheduler step.
-	StepName string `json:"stepName"`
-	// Activations of the step for each host.
-	Activations map[string]float64 `json:"activations"`
-}
-
-type DecisionResult struct {
-	// Raw input weights to the pipeline.
-	// +kubebuilder:validation:Optional
-	RawInWeights map[string]float64 `json:"rawInWeights"`
-	// Normalized input weights to the pipeline.
-	// +kubebuilder:validation:Optional
-	NormalizedInWeights map[string]float64 `json:"normalizedInWeights"`
-	// Outputs of the decision pipeline including the activations used
-	// to make the final ordering of compute hosts.
-	// +kubebuilder:validation:Optional
-	StepResults []StepResult `json:"stepResults,omitempty"`
-	// Aggregated output weights from the pipeline.
-	// +kubebuilder:validation:Optional
-	AggregatedOutWeights map[string]float64 `json:"aggregatedOutWeights"`
-	// Final ordered list of hosts from most preferred to least preferred.
-	// +kubebuilder:validation:Optional
-	OrderedHosts []string `json:"orderedHosts,omitempty"`
-	// The first element of the ordered hosts is considered the target host.
-	// +kubebuilder:validation:Optional
-	TargetHost *string `json:"targetHost,omitempty"`
 }
 
 const (
-	// The decision was successfully processed.
+	// The decision is ready and tracking the resource.
 	DecisionConditionReady = "Ready"
+	// The decision has failed to make a placement decision for the resource.
+	DecisionConditionFailed = "Failed"
 )
 
 type DecisionStatus struct {
-	// The result of this decision.
-	// +kubebuilder:validation:Optional
-	Result *DecisionResult `json:"result,omitempty"`
-
-	// If there were previous decisions for the underlying resource, they can
-	// be resolved here to provide historical context for the decision.
+	// The target host selected for the resource. Can be empty if no host could be determined.
 	// +kubebuilder:validation:Optional
-	History *[]corev1.ObjectReference `json:"history,omitempty"`
+	TargetHost string `json:"targetHost,omitempty"`
 
-	// The number of decisions that preceded this one for the same resource.
+	// The history of scheduling events for this resource.
 	// +kubebuilder:validation:Optional
-	Precedence *int `json:"precedence,omitempty"`
+	SchedulingHistory []SchedulingHistoryEntry `json:"schedulingHistory,omitempty"`
 
-	// A human-readable explanation of the decision result.
+	// A human-readable explanation of the current scheduling decision.
 	// +kubebuilder:validation:Optional
 	Explanation string `json:"explanation,omitempty"`
 
@@ -103,12 +80,8 @@ type DecisionStatus struct {
 // +kubebuilder:resource:scope=Cluster
 // +kubebuilder:printcolumn:name="Domain",type="string",JSONPath=".spec.schedulingDomain"
 // +kubebuilder:printcolumn:name="Resource ID",type="string",JSONPath=".spec.resourceID"
-// +kubebuilder:printcolumn:name="#",type="string",JSONPath=".status.precedence"
+// +kubebuilder:printcolumn:name="Target Host",type="string",JSONPath=".status.targetHost"
 // +kubebuilder:printcolumn:name="Created",type="date",JSONPath=".metadata.creationTimestamp"
-// +kubebuilder:printcolumn:name="Pipeline",type="string",JSONPath=".spec.pipelineRef.name"
-// +kubebuilder:printcolumn:name="TargetHost",type="string",JSONPath=".status.result.targetHost"
-// +kubebuilder:printcolumn:name="Ready",type="string",JSONPath=".status.conditions[?(@.type=='Ready')].status"
-// +kubebuilder:selectablefield:JSONPath=".spec.resourceID"
 
 // Decision is the Schema for the decisions API
 type Decision struct {

cmd/main.go

@@ -41,7 +41,6 @@ import (
 	"github.com/cobaltcore-dev/cortex/internal/knowledge/extractor"
 	"github.com/cobaltcore-dev/cortex/internal/knowledge/kpis"
 	"github.com/cobaltcore-dev/cortex/internal/scheduling/cinder"
-	"github.com/cobaltcore-dev/cortex/internal/scheduling/explanation"
 	schedulinglib "github.com/cobaltcore-dev/cortex/internal/scheduling/lib"
 	"github.com/cobaltcore-dev/cortex/internal/scheduling/machines"
 	"github.com/cobaltcore-dev/cortex/internal/scheduling/manila"
@@ -439,19 +438,6 @@ func main() {
 			os.Exit(1)
 		}
 	}
-	if slices.Contains(mainConfig.EnabledControllers, "explanation-controller") {
-		// Setup a controller which will reconcile the history and explanation for
-		// decision resources.
-		explanationControllerConfig := conf.GetConfigOrDie[explanation.ControllerConfig]()
-		explanationController := &explanation.Controller{
-			Client: multiclusterClient,
-			Config: explanationControllerConfig,
-		}
-		if err := explanationController.SetupWithManager(mgr, multiclusterClient); err != nil {
-			setupLog.Error(err, "unable to create controller", "controller", "ExplanationController")
-			os.Exit(1)
-		}
-	}
 	if slices.Contains(mainConfig.EnabledControllers, "reservations-controller") {
 		monitor := reservationscontroller.NewControllerMonitor(multiclusterClient)
 		metrics.Registry.MustRegister(&monitor)

helm/bundles/cortex-cinder/values.yaml

@@ -97,7 +97,6 @@ cortex-scheduling-controllers:
         component: cinder-scheduling
     enabledControllers:
       - cinder-decisions-pipeline-controller
-      - explanation-controller
     enabledTasks:
       - cinder-decisions-cleanup-task
 

helm/bundles/cortex-ironcore/values.yaml

@@ -32,7 +32,6 @@ cortex:
     schedulingDomain: machines
     enabledControllers:
       - ironcore-decisions-pipeline-controller
-      - explanation-controller
     monitoring:
       labels:
         github_org: cobaltcore-dev

helm/bundles/cortex-manila/values.yaml

@@ -97,7 +97,6 @@ cortex-scheduling-controllers:
         component: manila-scheduling
     enabledControllers:
       - manila-decisions-pipeline-controller
-      - explanation-controller
     enabledTasks:
       - manila-decisions-cleanup-task
 

helm/bundles/cortex-nova/values.yaml

@@ -107,7 +107,6 @@ cortex-scheduling-controllers:
     enabledControllers:
       - nova-pipeline-controllers
       - nova-deschedulings-executor
-      - explanation-controller
     enabledTasks:
       - nova-decisions-cleanup-task
 

helm/bundles/cortex-pods/values.yaml

@@ -32,7 +32,6 @@ cortex:
     schedulingDomain: pods
     enabledControllers:
       - pods-decisions-pipeline-controller
-      - explanation-controller
     monitoring:
       labels:
         github_org: cobaltcore-dev