Add instance type offering availability to metrics

pkg/providers/instancetype/instancetype.go

@@ -118,16 +118,15 @@ func (p *Provider) List(ctx context.Context, kc *corev1beta1.KubeletConfiguratio
 		return item.([]*cloudprovider.InstanceType), nil
 	}
 	result := lo.Map(instanceTypes, func(i *ec2.InstanceTypeInfo, _ int) *cloudprovider.InstanceType {
+		instanceTypeVCPU.With(prometheus.Labels{
+			instanceTypeLabel: *i.InstanceType,
+		}).Set(float64(aws.Int64Value(i.VCpuInfo.DefaultVCpus)))
+		instanceTypeMemory.With(prometheus.Labels{
+			instanceTypeLabel: *i.InstanceType,
+		}).Set(float64(aws.Int64Value(i.MemoryInfo.SizeInMiB) * 1024 * 1024))
+
 		return NewInstanceType(ctx, i, kc, p.region, nodeClass, p.createOfferings(ctx, i, instanceTypeOfferings[aws.StringValue(i.InstanceType)], zones, subnetZones))
 	})
-	for _, instanceType := range instanceTypes {
-		InstanceTypeVCPU.With(prometheus.Labels{
-			InstanceTypeLabel: *instanceType.InstanceType,
-		}).Set(float64(aws.Int64Value(instanceType.VCpuInfo.DefaultVCpus)))
-		InstanceTypeMemory.With(prometheus.Labels{
-			InstanceTypeLabel: *instanceType.InstanceType,
-		}).Set(float64(aws.Int64Value(instanceType.MemoryInfo.SizeInMiB) * 1024 * 1024))
-	}
 	p.cache.SetDefault(key, result)
 	return result, nil
 }
@@ -167,6 +166,16 @@ func (p *Provider) createOfferings(ctx context.Context, instanceType *ec2.Instan
 				Price:        price,
 				Available:    available,
 			})
+			instanceTypeOfferingAvailable.With(prometheus.Labels{
+				instanceTypeLabel: *instanceType.InstanceType,
+				capacityTypeLabel: capacityType,
+				zoneLabel:         zone,
+			}).Set(float64(lo.Ternary(available, 1, 0)))
+			instanceTypeOfferingPriceEstimate.With(prometheus.Labels{
+				instanceTypeLabel: *instanceType.InstanceType,
+				capacityTypeLabel: capacityType,
+				zoneLabel:         zone,
+			}).Set(price)
 		}
 	}
 	return offerings

pkg/providers/instancetype/metrics.go

@@ -23,34 +23,61 @@ import (
 
 const (
 	cloudProviderSubsystem = "cloudprovider"
+	instanceTypeLabel      = "instance_type"
+	capacityTypeLabel      = "capacity_type"
+	zoneLabel              = "zone"
 )
 
 var (
-	InstanceTypeLabel = "instance_type"
-
-	InstanceTypeVCPU = prometheus.NewGaugeVec(
+	instanceTypeVCPU = prometheus.NewGaugeVec(
 		prometheus.GaugeOpts{
 			Namespace: metrics.Namespace,
 			Subsystem: cloudProviderSubsystem,
 			Name:      "instance_type_cpu_cores",
 			Help:      "VCPUs cores for a given instance type.",
 		},
 		[]string{
-			InstanceTypeLabel,
-		})
-
-	InstanceTypeMemory = prometheus.NewGaugeVec(
+			instanceTypeLabel,
+		},
+	)
+	instanceTypeMemory = prometheus.NewGaugeVec(
 		prometheus.GaugeOpts{
 			Namespace: metrics.Namespace,
 			Subsystem: cloudProviderSubsystem,
 			Name:      "instance_type_memory_bytes",
 			Help:      "Memory, in bytes, for a given instance type.",
 		},
 		[]string{
-			InstanceTypeLabel,
+			instanceTypeLabel,
+		},
+	)
+	instanceTypeOfferingAvailable = prometheus.NewGaugeVec(
+		prometheus.GaugeOpts{
+			Namespace: metrics.Namespace,
+			Subsystem: cloudProviderSubsystem,
+			Name:      "instance_type_offering_available",
+			Help:      "Instance type offering availability, based on instance type, capacity type, and zone",
+		},
+		[]string{
+			instanceTypeLabel,
+			capacityTypeLabel,
+			zoneLabel,
+		},
+	)
+	instanceTypeOfferingPriceEstimate = prometheus.NewGaugeVec(
+		prometheus.GaugeOpts{
+			Namespace: metrics.Namespace,
+			Subsystem: cloudProviderSubsystem,
+			Name:      "instance_type_offering_price_estimate",
+			Help:      "Instance type offering estimated estimated hourly price used when making informed decisions on node cost calculation, based on instance type, capacity type, and zone.",
+		},
+		[]string{
+			instanceTypeLabel,
+			capacityTypeLabel,
+			zoneLabel,
 		})
 )
 
 func init() {
-	crmetrics.Registry.MustRegister(InstanceTypeVCPU, InstanceTypeMemory)
+	crmetrics.Registry.MustRegister(instanceTypeVCPU, instanceTypeMemory, instanceTypeOfferingAvailable, instanceTypeOfferingPriceEstimate)
 }

pkg/providers/instancetype/suite_test.go

@@ -741,34 +741,71 @@ var _ = Describe("InstanceTypes", func() {
 			Expect(it.Capacity.Pods().Value()).To(BeNumerically("==", 110))
 		}
 	})
-
-	It("should expose vcpu metrics for instance types", func() {
-		instanceInfo, err := awsEnv.InstanceTypesProvider.List(ctx, nodePool.Spec.Template.Spec.Kubelet, nodeClass)
-		Expect(err).To(BeNil())
-		Expect(len(instanceInfo)).To(BeNumerically(">", 0))
-		for _, info := range instanceInfo {
-			metric, ok := FindMetricWithLabelValues("karpenter_cloudprovider_instance_type_cpu_cores", map[string]string{
-				instancetype.InstanceTypeLabel: info.Name,
-			})
-			Expect(ok).To(BeTrue())
-			Expect(metric).To(Not(BeNil()))
-			value := metric.GetGauge().Value
-			Expect(aws.Float64Value(value)).To(BeNumerically(">", 0))
-		}
-	})
-	It("should expose memory metrics for instance types", func() {
-		instanceInfo, err := awsEnv.InstanceTypesProvider.List(ctx, nodePool.Spec.Template.Spec.Kubelet, nodeClass)
-		Expect(err).To(BeNil())
-		Expect(len(instanceInfo)).To(BeNumerically(">", 0))
-		for _, info := range instanceInfo {
-			metric, ok := FindMetricWithLabelValues("karpenter_cloudprovider_instance_type_memory_bytes", map[string]string{
-				instancetype.InstanceTypeLabel: info.Name,
-			})
-			Expect(ok).To(BeTrue())
-			Expect(metric).To(Not(BeNil()))
-			value := metric.GetGauge().Value
-			Expect(aws.Float64Value(value)).To(BeNumerically(">", 0))
-		}
+	Context("Metrics", func() {
+		It("should expose vcpu metrics for instance types", func() {
+			instanceTypes, err := awsEnv.InstanceTypesProvider.List(ctx, nodePool.Spec.Template.Spec.Kubelet, nodeClass)
+			Expect(err).To(BeNil())
+			Expect(len(instanceTypes)).To(BeNumerically(">", 0))
+			for _, it := range instanceTypes {
+				metric, ok := FindMetricWithLabelValues("karpenter_cloudprovider_instance_type_cpu_cores", map[string]string{
+					"instance_type": it.Name,
+				})
+				Expect(ok).To(BeTrue())
+				Expect(metric).To(Not(BeNil()))
+				value := metric.GetGauge().Value
+				Expect(aws.Float64Value(value)).To(BeNumerically(">", 0))
+			}
+		})
+		It("should expose memory metrics for instance types", func() {
+			instanceTypes, err := awsEnv.InstanceTypesProvider.List(ctx, nodePool.Spec.Template.Spec.Kubelet, nodeClass)
+			Expect(err).To(BeNil())
+			Expect(len(instanceTypes)).To(BeNumerically(">", 0))
+			for _, it := range instanceTypes {
+				metric, ok := FindMetricWithLabelValues("karpenter_cloudprovider_instance_type_memory_bytes", map[string]string{
+					"instance_type": it.Name,
+				})
+				Expect(ok).To(BeTrue())
+				Expect(metric).To(Not(BeNil()))
+				value := metric.GetGauge().Value
+				Expect(aws.Float64Value(value)).To(BeNumerically(">", 0))
+			}
+		})
+		It("should expose availability metrics for instance types", func() {
+			instanceTypes, err := awsEnv.InstanceTypesProvider.List(ctx, nodePool.Spec.Template.Spec.Kubelet, nodeClass)
+			Expect(err).To(BeNil())
+			Expect(len(instanceTypes)).To(BeNumerically(">", 0))
+			for _, it := range instanceTypes {
+				for _, of := range it.Offerings {
+					metric, ok := FindMetricWithLabelValues("karpenter_cloudprovider_instance_type_offering_available", map[string]string{
+						"instance_type": it.Name,
+						"capacity_type": of.CapacityType,
+						"zone":          of.Zone,
+					})
+					Expect(ok).To(BeTrue())
+					Expect(metric).To(Not(BeNil()))
+					value := metric.GetGauge().Value
+					Expect(aws.Float64Value(value)).To(BeNumerically("==", lo.Ternary(of.Available, 1, 0)))
+				}
+			}
+		})
+		It("should expose pricing metrics for instance types", func() {
+			instanceTypes, err := awsEnv.InstanceTypesProvider.List(ctx, nodePool.Spec.Template.Spec.Kubelet, nodeClass)
+			Expect(err).To(BeNil())
+			Expect(len(instanceTypes)).To(BeNumerically(">", 0))
+			for _, it := range instanceTypes {
+				for _, of := range it.Offerings {
+					metric, ok := FindMetricWithLabelValues("karpenter_cloudprovider_instance_type_offering_price_estimate", map[string]string{
+						"instance_type": it.Name,
+						"capacity_type": of.CapacityType,
+						"zone":          of.Zone,
+					})
+					Expect(ok).To(BeTrue())
+					Expect(metric).To(Not(BeNil()))
+					value := metric.GetGauge().Value
+					Expect(aws.Float64Value(value)).To(BeNumerically("==", of.Price))
+				}
+			}
+		})
 	})
 	It("should launch instances in local zones", func() {
 		ExpectApplied(ctx, env.Client, nodePool, nodeClass)

pkg/providers/pricing/pricing.go

@@ -33,7 +33,6 @@ import (
 	"github.com/aws/aws-sdk-go/service/ec2/ec2iface"
 	"github.com/aws/aws-sdk-go/service/pricing"
 	"github.com/aws/aws-sdk-go/service/pricing/pricingiface"
-	"github.com/prometheus/client_golang/prometheus"
 	"github.com/samber/lo"
 	"go.uber.org/multierr"
 	"knative.dev/pkg/logging"
@@ -204,14 +203,6 @@ func (p *Provider) UpdateOnDemandPricing(ctx context.Context) error {
 	}
 
 	p.onDemandPrices = lo.Assign(onDemandPrices, onDemandMetalPrices)
-	for instanceType, price := range p.onDemandPrices {
-		InstancePriceEstimate.With(prometheus.Labels{
-			InstanceTypeLabel: instanceType,
-			CapacityTypeLabel: ec2.UsageClassTypeOnDemand,
-			RegionLabel:       p.region,
-			TopologyLabel:     "",
-		}).Set(price)
-	}
 	if p.cm.HasChanged("on-demand-prices", p.onDemandPrices) {
 		logging.FromContext(ctx).With("instance-type-count", len(p.onDemandPrices)).Debugf("updated on-demand pricing")
 	}
@@ -341,12 +332,6 @@ func (p *Provider) UpdateSpotPricing(ctx context.Context) error {
 				prices[instanceType] = map[string]float64{}
 			}
 			prices[instanceType][az] = spotPrice
-			InstancePriceEstimate.With(prometheus.Labels{
-				InstanceTypeLabel: instanceType,
-				CapacityTypeLabel: ec2.UsageClassTypeSpot,
-				RegionLabel:       p.region,
-				TopologyLabel:     az,
-			}).Set(spotPrice)
 		}
 		return true
 	})

pkg/providers/pricing/suite_test.go

@@ -123,12 +123,10 @@ var _ = Describe("Pricing", func() {
 		price, ok := awsEnv.PricingProvider.OnDemandPrice("c98.large")
 		Expect(ok).To(BeTrue())
 		Expect(price).To(BeNumerically("==", 1.20))
-		Expect(getPricingEstimateMetricValue("c98.large", ec2.UsageClassTypeOnDemand, "")).To(BeNumerically("==", 1.20))
 
 		price, ok = awsEnv.PricingProvider.OnDemandPrice("c99.large")
 		Expect(ok).To(BeTrue())
 		Expect(price).To(BeNumerically("==", 1.23))
-		Expect(getPricingEstimateMetricValue("c99.large", ec2.UsageClassTypeOnDemand, "")).To(BeNumerically("==", 1.23))
 	})
 	It("should update spot pricing with response from the pricing API", func() {
 		now := time.Now()
@@ -171,12 +169,10 @@ var _ = Describe("Pricing", func() {
 		price, ok := awsEnv.PricingProvider.SpotPrice("c98.large", "test-zone-1b")
 		Expect(ok).To(BeTrue())
 		Expect(price).To(BeNumerically("==", 1.10))
-		Expect(getPricingEstimateMetricValue("c98.large", ec2.UsageClassTypeSpot, "test-zone-1b")).To(BeNumerically("==", 1.10))
 
 		price, ok = awsEnv.PricingProvider.SpotPrice("c99.large", "test-zone-1a")
 		Expect(ok).To(BeTrue())
 		Expect(price).To(BeNumerically("==", 1.23))
-		Expect(getPricingEstimateMetricValue("c99.large", ec2.UsageClassTypeSpot, "test-zone-1a")).To(BeNumerically("==", 1.23))
 	})
 	It("should update zonal pricing with data from the spot pricing API", func() {
 		now := time.Now()
@@ -207,7 +203,6 @@ var _ = Describe("Pricing", func() {
 		price, ok := awsEnv.PricingProvider.SpotPrice("c98.large", "test-zone-1a")
 		Expect(ok).To(BeTrue())
 		Expect(price).To(BeNumerically("==", 1.20))
-		Expect(getPricingEstimateMetricValue("c98.large", ec2.UsageClassTypeSpot, "test-zone-1a")).To(BeNumerically("==", 1.20))
 
 		_, ok = awsEnv.PricingProvider.SpotPrice("c98.large", "test-zone-1b")
 		Expect(ok).ToNot(BeTrue())
@@ -300,20 +295,5 @@ var _ = Describe("Pricing", func() {
 		price, ok = awsEnv.PricingProvider.SpotPrice("c98.large", "test-zone-1b")
 		Expect(ok).To(BeTrue())
 		Expect(price).To(BeNumerically("==", 1.10))
-		Expect(getPricingEstimateMetricValue("c98.large", ec2.UsageClassTypeSpot, "test-zone-1b")).To(BeNumerically("==", 1.10))
 	})
 })
-
-func getPricingEstimateMetricValue(instanceType string, capacityType string, zone string) float64 {
-	var value *float64
-	metric, ok := FindMetricWithLabelValues("karpenter_cloudprovider_instance_type_price_estimate", map[string]string{
-		pricing.InstanceTypeLabel: instanceType,
-		pricing.CapacityTypeLabel: capacityType,
-		pricing.RegionLabel:       fake.DefaultRegion,
-		pricing.TopologyLabel:     zone,
-	})
-	Expect(ok).To(BeTrue())
-	value = metric.GetGauge().Value
-	Expect(value).To(Not(BeNil()))
-	return *value
-}

website/content/en/preview/reference/metrics.md

@@ -151,8 +151,11 @@ Current count of nodes in cluster state
 
 ## Cloudprovider Metrics
 
-### `karpenter_cloudprovider_instance_type_price_estimate`
-Estimated hourly price used when making informed decisions on node cost calculation. This is updated once on startup and then every 12 hours.
+### `karpenter_cloudprovider_instance_type_offering_price_estimate`
+Instance type offering estimated estimated hourly price used when making informed decisions on node cost calculation, based on instance type, capacity type, and zone.
+
+### `karpenter_cloudprovider_instance_type_offering_available`
+Instance type offering availability, based on instance type, capacity type, and zone
 
 ### `karpenter_cloudprovider_instance_type_memory_bytes`
 Memory, in bytes, for a given instance type.

website/content/en/preview/upgrading/upgrade-guide.md

@@ -37,6 +37,14 @@ kubectl apply -f https://raw.githubusercontent.com/aws/karpenter{{< githubRelRef
 WHEN CREATING A NEW SECTION OF THE UPGRADE GUIDANCE FOR NEWER VERSIONS, ENSURE THAT YOU COPY THE BETA API ALERT SECTION FROM THE LAST RELEASE TO PROPERLY WARN USERS OF THE RISK OF UPGRADING WITHOUT GOING TO 0.32.x FIRST
 -->
 
+### Upgrading to `0.36.0`+
+
+{{% alert title="Warning" color="warning" %}}
+`0.33.0`+ _only_ supports Karpenter v1beta1 APIs and will not work with existing Provisioner, AWSNodeTemplate or Machine alpha APIs. Do not upgrade to `0.35.0`+ without first [upgrading to `0.32.x`]({{<ref "#upgrading-to-0320" >}}). This version supports both the alpha and beta APIs, allowing you to migrate all of your existing APIs to beta APIs without experiencing downtime.
+{{% /alert %}}
+
+* Karpenter changed the name of the `karpenter_cloudprovider_instance_type_price_estimate` metric to `karpenter_cloudprovider_instance_type_offering_price_estimate` to align with the new `karpenter_cloudprovider_instance_type_offering_available` metric. The `region` label was also dropped from the metric, since this can be inferred from the environment that Karpenter is running in.
+
 ### Upgrading to `0.35.0`+
 
 {{% alert title="Warning" color="warning" %}}