feat: VM usage and fair capacity split for overlapping flavor groups

api/v1alpha1/flavor_group_capacity_types.go

@@ -24,7 +24,11 @@ type FlavorGroupCapacitySpec struct {
 	AvailabilityZone string `json:"availabilityZone"`
 }
 
-// FlavorCapacityStatus holds per-flavor capacity numbers for one (flavor group × AZ) pair.
+// FlavorCapacityStatus holds per-flavor scheduler probe results for one (flavor group × AZ) pair.
+// These values come directly from scheduler probes and are independent of the cross-group
+// capacity split (see FreeCapacity and ExclusivelyFreeCapacity on the parent status).
+// "Placeable" means: if all remaining capacity in this AZ were used solely by this flavor,
+// this is how many would fit. It does not account for competing flavor groups.
 type FlavorCapacityStatus struct {
 	// FlavorName is the OpenStack flavor name (e.g. "hana-v2-small").
 	FlavorName string `json:"flavorName"`
@@ -37,11 +41,11 @@ type FlavorCapacityStatus struct {
 	// +kubebuilder:validation:Optional
 	PlaceableVMs int64 `json:"placeableVms,omitempty"`
 
-	// TotalCapacityHosts is the number of eligible hosts in an empty-datacenter scenario.
+	// TotalCapacityHosts is the number of eligible hosts assuming an empty datacenter.
 	// +kubebuilder:validation:Optional
 	TotalCapacityHosts int64 `json:"totalCapacityHosts,omitempty"`
 
-	// TotalCapacityVMSlots is the maximum number of VM slots in an empty-datacenter scenario.
+	// TotalCapacityVMSlots is the maximum number of VM slots assuming an empty datacenter.
 	// +kubebuilder:validation:Optional
 	TotalCapacityVMSlots int64 `json:"totalCapacityVmSlots,omitempty"`
 }
@@ -57,14 +61,33 @@ type FlavorGroupCapacityStatus struct {
 	// +kubebuilder:validation:Optional
 	CommittedCapacity int64 `json:"committedCapacity,omitempty"`
 
-	// TotalCapacity is the total capacity of all eligible hosts in an empty-datacenter scenario.
+	// TotalCapacity is the installed capacity across all eligible hosts in an empty-datacenter
+	// scenario, expressed as raw resource amounts (bytes for memory, count for cores).
 	// +kubebuilder:validation:Optional
 	TotalCapacity map[string]resource.Quantity `json:"totalCapacity,omitempty"`
 
-	// TotalInstances is the total number of VM instances running on hypervisors in this AZ,
-	// derived from Hypervisor CRD Status.Instances (not filtered by flavor group).
+	// FreeCapacity is the sum of remaining resources across all candidate hosts for this
+	// group given current allocations. Because groups can share hosts, the sum across groups
+	// may exceed actual installed capacity — this field reflects per-group availability
+	// before any cross-group fairness split.
 	// +kubebuilder:validation:Optional
-	TotalInstances int64 `json:"totalInstances,omitempty"`
+	FreeCapacity map[string]resource.Quantity `json:"freeCapacity,omitempty"`
+
+	// ExclusivelyFreeCapacity is the share of remaining resources fairly attributed to this
+	// group by the round-robin capacity split. The sum across all groups for an AZ never
+	// exceeds actual installed capacity.
+	// +kubebuilder:validation:Optional
+	ExclusivelyFreeCapacity map[string]resource.Quantity `json:"exclusivelyFreeCapacity,omitempty"`
+
+	// RunningInstances is the number of VMs running in this (flavor group × AZ) whose
+	// flavor belongs to this group.
+	// +kubebuilder:validation:Optional
+	RunningInstances int64 `json:"runningInstances,omitempty"`
+
+	// RunningResources is the total resource consumption of running VMs in this
+	// (flavor group × AZ), keyed by resource type (e.g. "memory", "cores").
+	// +kubebuilder:validation:Optional
+	RunningResources map[string]resource.Quantity `json:"runningResources,omitempty"`
 
 	// LastReconcileAt is the timestamp of the last successful reconcile.
 	// +kubebuilder:validation:Optional
@@ -80,7 +103,7 @@ type FlavorGroupCapacityStatus struct {
 // +kubebuilder:resource:scope=Cluster
 // +kubebuilder:printcolumn:name="FlavorGroup",type="string",JSONPath=".spec.flavorGroup"
 // +kubebuilder:printcolumn:name="AZ",type="string",JSONPath=".spec.availabilityZone"
-// +kubebuilder:printcolumn:name="TotalInstances",type="integer",JSONPath=".status.totalInstances"
+// +kubebuilder:printcolumn:name="Running",type="integer",JSONPath=".status.runningInstances"
 // +kubebuilder:printcolumn:name="LastReconcile",type="date",JSONPath=".status.lastReconcileAt"
 // +kubebuilder:printcolumn:name="Ready",type="string",JSONPath=".status.conditions[?(@.type=='Ready')].status"
 

cmd/manager/main.go

@@ -771,7 +771,7 @@ func main() {
 			setupLog.Error(err, "failed to register capacity monitor metrics, continuing without metrics")
 		}
 
-		capacityController := capacity.NewController(multiclusterClient, capacityConfig)
+		capacityController := capacity.NewController(multiclusterClient, capacityConfig, commitmentsVMSource)
 		if err := mgr.Add(manager.RunnableFunc(func(ctx context.Context) error {
 			return capacityController.Start(ctx)
 		})); err != nil {

helm/library/cortex/files/crds/cortex.cloud_flavorgroupcapacities.yaml

@@ -21,8 +21,8 @@ spec:
     - jsonPath: .spec.availabilityZone
       name: AZ
       type: string
-    - jsonPath: .status.totalInstances
-      name: TotalInstances
+    - jsonPath: .status.runningInstances
+      name: Running
       type: integer
     - jsonPath: .status.lastReconcileAt
       name: LastReconcile
@@ -135,12 +135,28 @@ spec:
                   - type
                   type: object
                 type: array
+              exclusivelyFreeCapacity:
+                additionalProperties:
+                  anyOf:
+                  - type: integer
+                  - type: string
+                  pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
+                  x-kubernetes-int-or-string: true
+                description: |-
+                  ExclusivelyFreeCapacity is the share of remaining resources fairly attributed to this
+                  group by the round-robin capacity split. The sum across all groups for an AZ never
+                  exceeds actual installed capacity.
+                type: object
               flavors:
                 description: Flavors holds per-flavor capacity data for all flavors
                   in the group.
                 items:
-                  description: FlavorCapacityStatus holds per-flavor capacity numbers
-                    for one (flavor group × AZ) pair.
+                  description: |-
+                    FlavorCapacityStatus holds per-flavor scheduler probe results for one (flavor group × AZ) pair.
+                    These values come directly from scheduler probes and are independent of the cross-group
+                    capacity split (see FreeCapacity and ExclusivelyFreeCapacity on the parent status).
+                    "Placeable" means: if all remaining capacity in this AZ were used solely by this flavor,
+                    this is how many would fit. It does not account for competing flavor groups.
                   properties:
                     flavorName:
                       description: FlavorName is the OpenStack flavor name (e.g. "hana-v2-small").
@@ -157,39 +173,64 @@ spec:
                       type: integer
                     totalCapacityHosts:
                       description: TotalCapacityHosts is the number of eligible hosts
-                        in an empty-datacenter scenario.
+                        assuming an empty datacenter.
                       format: int64
                       type: integer
                     totalCapacityVmSlots:
                       description: TotalCapacityVMSlots is the maximum number of VM
-                        slots in an empty-datacenter scenario.
+                        slots assuming an empty datacenter.
                       format: int64
                       type: integer
                   required:
                   - flavorName
                   type: object
                 type: array
+              freeCapacity:
+                additionalProperties:
+                  anyOf:
+                  - type: integer
+                  - type: string
+                  pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
+                  x-kubernetes-int-or-string: true
+                description: |-
+                  FreeCapacity is the sum of remaining resources across all candidate hosts for this
+                  group given current allocations. Because groups can share hosts, the sum across groups
+                  may exceed actual installed capacity — this field reflects per-group availability
+                  before any cross-group fairness split.
+                type: object
               lastReconcileAt:
                 description: LastReconcileAt is the timestamp of the last successful
                   reconcile.
                 format: date-time
                 type: string
-              totalCapacity:
+              runningInstances:
+                description: |-
+                  RunningInstances is the number of VMs running in this (flavor group × AZ) whose
+                  flavor belongs to this group.
+                format: int64
+                type: integer
+              runningResources:
                 additionalProperties:
                   anyOf:
                   - type: integer
                   - type: string
                   pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
                   x-kubernetes-int-or-string: true
-                description: TotalCapacity is the total capacity of all eligible hosts
-                  in an empty-datacenter scenario.
+                description: |-
+                  RunningResources is the total resource consumption of running VMs in this
+                  (flavor group × AZ), keyed by resource type (e.g. "memory", "cores").
                 type: object
-              totalInstances:
+              totalCapacity:
+                additionalProperties:
+                  anyOf:
+                  - type: integer
+                  - type: string
+                  pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
+                  x-kubernetes-int-or-string: true
                 description: |-
-                  TotalInstances is the total number of VM instances running on hypervisors in this AZ,
-                  derived from Hypervisor CRD Status.Instances (not filtered by flavor group).
-                format: int64
-                type: integer
+                  TotalCapacity is the installed capacity across all eligible hosts in an empty-datacenter
+                  scenario, expressed as raw resource amounts (bytes for memory, count for cores).
+                type: object
             type: object
         required:
         - spec