capacity metrics

This adds gauges that describe how many objects the
external-provisioner manages, how many of those already exist, and how
many are to be deleted. Example:

  # HELP csistoragecapacities_desired_current [ALPHA] Number of CSIStorageCapacity objects that exist and are supposed to be managed automatically.
  # TYPE csistoragecapacities_desired_current gauge
  csistoragecapacities_desired_current 2
  # HELP csistoragecapacities_desired_goal [ALPHA] Number of CSIStorageCapacity objects that are supposed to be managed automatically.
  # TYPE csistoragecapacities_desired_goal gauge
  csistoragecapacities_desired_goal 2
  # HELP csistoragecapacities_obsolete [ALPHA] Number of CSIStorageCapacity objects that exist and will be deleted automatically.
  # TYPE csistoragecapacities_obsolete gauge
  csistoragecapacities_obsolete 0

cmd/csi-provisioner/csi-provisioner.go

@@ -468,6 +468,7 @@ func main() {
 			*capacityPollInterval,
 			*capacityImmediateBinding,
 		)
+		legacyregistry.CustomMustRegister(capacityController)
 	}
 
 	// Start HTTP server, regardless whether we are the leader or not.

go.sum

@@ -125,6 +125,7 @@ github.com/cockroachdb/datadriven v0.0.0-20190809214429-80d97fb3cbaa/go.mod h1:z
 github.com/codahale/hdrhistogram v0.0.0-20161010025455-3a0bb77429bd/go.mod h1:sE/e/2PUdi/liOCUjSTXgM1o87ZssimdTWN964YiIeI=
 github.com/codegangsta/negroni v1.0.0/go.mod h1:v0y3T5G7Y1UlFfyxFn/QLRU4a2EuNau2iZY63YTKWo0=
 github.com/container-storage-interface/spec v1.2.0/go.mod h1:6URME8mwIBbpVyZV93Ce5St17xBiQJQY67NDsuohiy4=
+github.com/container-storage-interface/spec v1.3.0 h1:wMH4UIoWnK/TXYw8mbcIHgZmB6kHOeIsYsiaTJwa6bc=
 github.com/container-storage-interface/spec v1.3.0/go.mod h1:6URME8mwIBbpVyZV93Ce5St17xBiQJQY67NDsuohiy4=
 github.com/containerd/cgroups v0.0.0-20200531161412-0dbf7f05ba59/go.mod h1:pA0z1pT8KYB3TCXK/ocprsh7MAkoW8bZVzPdih9snmM=
 github.com/containerd/console v0.0.0-20180822173158-c12b1e7919c1/go.mod h1:Tj/on1eG8kiEhd0+fhSDzsPAFESxzBBvdyEgyryXffw=
@@ -220,7 +221,6 @@ github.com/go-logfmt/logfmt v0.4.0/go.mod h1:3RMwSq7FuexP4Kalkev3ejPJsZTpXXBr9+V
 github.com/go-logfmt/logfmt v0.5.0/go.mod h1:wCYkCAKZfumFQihp8CzCvQ3paCTfi41vtzG1KdI/P7A=
 github.com/go-logr/logr v0.1.0/go.mod h1:ixOQHD9gLJUVQQ2ZOR7zLEifBX6tGkNJF4QyIY7sIas=
 github.com/go-logr/logr v0.2.0/go.mod h1:z6/tIYblkpsD+a4lm/fGIIU9mZ+XfAiaFtq7xTgseGU=
-github.com/go-logr/logr v0.3.0 h1:q4c+kbcR0d5rSurhBR8dIgieOaYpXtsdTYfx22Cu6rs=
 github.com/go-logr/logr v0.3.0/go.mod h1:z6/tIYblkpsD+a4lm/fGIIU9mZ+XfAiaFtq7xTgseGU=
 github.com/go-logr/logr v0.4.0 h1:K7/B1jt6fIBQVd4Owv2MqGQClcgf0R266+7C/QjRcLc=
 github.com/go-logr/logr v0.4.0/go.mod h1:z6/tIYblkpsD+a4lm/fGIIU9mZ+XfAiaFtq7xTgseGU=
@@ -1163,7 +1163,6 @@ k8s.io/klog v1.0.0/go.mod h1:4Bi6QPql/J/LkTDqv7R/cd3hPo4k2DG6Ptcz060Ez5I=
 k8s.io/klog/v2 v2.0.0/go.mod h1:PBfzABfn139FHAV07az/IF9Wp1bkk3vpT2XSJ76fSDE=
 k8s.io/klog/v2 v2.2.0/go.mod h1:Od+F08eJP+W3HUb4pSrPpgp9DGU4GzlpG/TmITuYh/Y=
 k8s.io/klog/v2 v2.3.0/go.mod h1:Od+F08eJP+W3HUb4pSrPpgp9DGU4GzlpG/TmITuYh/Y=
-k8s.io/klog/v2 v2.4.0 h1:7+X0fUguPyrKEC4WjH8iGDg3laWgMo5tMnRTIGTTxGQ=
 k8s.io/klog/v2 v2.4.0/go.mod h1:Od+F08eJP+W3HUb4pSrPpgp9DGU4GzlpG/TmITuYh/Y=
 k8s.io/klog/v2 v2.5.0 h1:8mOnjf1RmUPW6KRqQCfYSZq/K20Unmp3IhuZUhxl8KI=
 k8s.io/klog/v2 v2.5.0/go.mod h1:hy9LJ/NvuK+iVyP4Ehqva4HxZG/oXyIS3n3Jmire4Ec=

pkg/capacity/capacity.go

@@ -41,6 +41,7 @@ import (
 	"k8s.io/client-go/kubernetes"
 	"k8s.io/client-go/tools/cache"
 	"k8s.io/client-go/util/workqueue"
+	"k8s.io/component-base/metrics"
 	"k8s.io/klog/v2"
 )
 
@@ -73,6 +74,8 @@ const (
 // and storage class name as some other object. That should never happen,
 // but the controller is prepared to clean that up, just in case.
 type Controller struct {
+	metrics.BaseStableCollector
+
 	csiController    CSICapacityClient
 	driverName       string
 	client           kubernetes.Interface
@@ -107,6 +110,28 @@ var (
 		Factor:   1.1,
 		Steps:    10,
 	}
+
+	objectsGoalDesc = metrics.NewDesc(
+		"csistoragecapacities_desired_goal",
+		"Number of CSIStorageCapacity objects that are supposed to be managed automatically.",
+		nil, nil,
+		metrics.ALPHA,
+		"",
+	)
+	objectsCurrentDesc = metrics.NewDesc(
+		"csistoragecapacities_desired_current",
+		"Number of CSIStorageCapacity objects that exist and are supposed to be managed automatically.",
+		nil, nil,
+		metrics.ALPHA,
+		"",
+	)
+	objectsObsoleteDesc = metrics.NewDesc(
+		"csistoragecapacities_obsolete",
+		"Number of CSIStorageCapacity objects that exist and will be deleted automatically.",
+		nil, nil,
+		metrics.ALPHA,
+		"",
+	)
 )
 
 // CSICapacityClient is the relevant subset of csi.ControllerClient.
@@ -115,6 +140,8 @@ type CSICapacityClient interface {
 }
 
 // NewController creates a new controller for CSIStorageCapacity objects.
+// It implements metrics.StableCollector and thus can be registered in
+// a registry.
 func NewCentralCapacityController(
 	csiController CSICapacityClient,
 	driverName string,
@@ -358,8 +385,11 @@ func (c *Controller) addWorkItem(segment *topology.Segment, sc *storagev1.Storag
 		storageClassName: sc.Name,
 	}
 	// Ensure that we have an entry for it...
-	capacity := c.capacities[item]
-	c.capacities[item] = capacity
+	_, found := c.capacities[item]
+	if !found {
+		c.capacities[item] = nil
+	}
+
 	// ... and then tell our workers to update
 	// or create that capacity object.
 	klog.V(5).Infof("Capacity Controller: enqueuing %+v", item)
@@ -613,6 +643,83 @@ func (c *Controller) onCDelete(ctx context.Context, capacity *storagev1alpha1.CS
 	}
 }
 
+// DescribeWithStability implements the metrics.StableCollector interface.
+func (c *Controller) DescribeWithStability(ch chan<- *metrics.Desc) {
+	ch <- objectsGoalDesc
+	ch <- objectsCurrentDesc
+	ch <- objectsObsoleteDesc
+}
+
+// CollectWithStability implements the metrics.StableCollector interface.
+func (c *Controller) CollectWithStability(ch chan<- metrics.Metric) {
+	c.capacitiesLock.Lock()
+	defer c.capacitiesLock.Unlock()
+
+	ch <- metrics.NewLazyConstMetric(objectsGoalDesc,
+		metrics.GaugeValue,
+		float64(c.getObjectsGoal()),
+	)
+	ch <- metrics.NewLazyConstMetric(objectsCurrentDesc,
+		metrics.GaugeValue,
+		float64(c.getObjectsCurrent()),
+	)
+	ch <- metrics.NewLazyConstMetric(objectsObsoleteDesc,
+		metrics.GaugeValue,
+		float64(c.getObjectsObsolete()),
+	)
+}
+
+// getObjectsGoal is called during metrics gathering and calculates the number
+// of CSIStorageCapacity objects which are are meant to
+// to exist.
+func (c *Controller) getObjectsGoal() int64 {
+	return int64(len(c.capacities))
+}
+
+// getObjectsCurrent is called during metrics gathering and calculates the number
+// of CSIStorageCapacity objects which are currently exist and are meant to
+// continue to exist.
+func (c *Controller) getObjectsCurrent() int64 {
+	current := int64(0)
+	for _, capacity := range c.capacities {
+		if capacity != nil {
+			current++
+		}
+	}
+	return current
+}
+
+// getObsoleteObjects is called during metrics gathering and calculates the number
+// of CSIStorageCapacity objects which currently exist (according to our informer)
+// and which are no longer needed.
+func (c *Controller) getObjectsObsolete() int64 {
+	obsolete := int64(0)
+	capacities, _ := c.cInformer.Lister().List(labels.Everything())
+	if capacities == nil {
+		// Shouldn't happen, local operation.
+		return 0
+	}
+	for _, capacity := range capacities {
+		if !c.isControlledByUs(capacity.OwnerReferences) {
+			continue
+		}
+		if c.isObsolete(capacity) {
+			obsolete++
+		}
+	}
+	return obsolete
+}
+
+func (c *Controller) isObsolete(capacity *storagev1alpha1.CSIStorageCapacity) bool {
+	for item, _ := range c.capacities {
+		if item.storageClassName == capacity.StorageClassName &&
+			reflect.DeepEqual(item.segment.GetLabelSelector(), capacity.NodeTopology) {
+			return false
+		}
+	}
+	return true
+}
+
 // isControlledByUs implements the same logic as https://pkg.go.dev/k8s.io/apimachinery/pkg/apis/meta/v1?tab=doc#IsControlledBy,
 // just with the expected owner identified directly with the UID.
 func (c *Controller) isControlledByUs(owners []metav1.OwnerReference) bool {