Merge branch 'develop' into bolt/multi-savings

pkg/costmodel/cluster.go

@@ -296,7 +296,7 @@ func ClusterNodes(cp cloud.Provider, client prometheus.Client, duration, offset
 	queryNodeRAMCost := fmt.Sprintf(`sum_over_time((avg(kube_node_status_capacity_memory_bytes) by (cluster_id, node) * on(cluster_id, node) group_right avg(node_ram_hourly_cost) by (cluster_id, node, instance_type, provider_id))[%s:%dm]%s) / 1024 / 1024 / 1024 * %f`, durationStr, minsPerResolution, offsetStr, hourlyToCumulative)
 	queryNodeRAMBytes := fmt.Sprintf(`avg_over_time(avg(kube_node_status_capacity_memory_bytes) by (cluster_id, node)[%s:%dm]%s)`, durationStr, minsPerResolution, offsetStr)
 	queryNodeGPUCost := fmt.Sprintf(`sum_over_time((avg(node_gpu_hourly_cost) by (cluster_id, node, provider_id))[%s:%dm]%s)`, durationStr, minsPerResolution, offsetStr)
-	queryNodeLabels := fmt.Sprintf(`count_over_time(kube_node_labels[%s:%dm]%s)`, durationStr, minsPerResolution, offsetStr)
+	queryNodeLabels := fmt.Sprintf(`avg_over_time(kubecost_node_is_spot[%s:%dm]%s)`, durationStr, minsPerResolution, offsetStr)
 
 	resChNodeCPUCost := ctx.Query(queryNodeCPUCost)
 	resChNodeCPUCores := ctx.Query(queryNodeCPUCores)
@@ -462,13 +462,14 @@ func ClusterNodes(cp cloud.Provider, client prometheus.Client, duration, offset
 		}
 
 		// GCP preemptible label
-		pre, _ := result.GetString("label_cloud_google_com_gke_preemptible")
+		pre := result.Values[0].Value
+
 		cluster, err := result.GetString("cluster_id")
 		if err != nil {
 			cluster = env.GetClusterID()
 		}
 		key := fmt.Sprintf("%s/%s", cluster, nodeName)
-		if node, ok := nodeMap[key]; pre == "true" && ok {
+		if node, ok := nodeMap[key]; pre > 0.0 && ok {
 			node.Preemptible = true
 		}
 

pkg/costmodel/metrics.go

@@ -426,6 +426,11 @@ func StartCostModelMetricRecording(a *Accesses) bool {
 				a.RAMPriceRecorder.WithLabelValues(nodeName, nodeName, nodeType, nodeRegion, node.ProviderID).Set(ramCost)
 				a.GPUPriceRecorder.WithLabelValues(nodeName, nodeName, nodeType, nodeRegion, node.ProviderID).Set(gpuCost)
 				a.NodeTotalPriceRecorder.WithLabelValues(nodeName, nodeName, nodeType, nodeRegion, node.ProviderID).Set(totalCost)
+				if node.IsSpot() {
+					a.NodeSpotRecorder.WithLabelValues(nodeName, nodeName, nodeType, nodeRegion, node.ProviderID).Set(1.0)
+				} else {
+					a.NodeSpotRecorder.WithLabelValues(nodeName, nodeName, nodeType, nodeRegion, node.ProviderID).Set(0.0)
+				}
 				labelKey := getKeyFromLabelStrings(nodeName, nodeName, nodeType, nodeRegion, node.ProviderID)
 				nodeSeen[labelKey] = true
 			}
@@ -509,6 +514,12 @@ func StartCostModelMetricRecording(a *Accesses) bool {
 					} else {
 						klog.Infof("FAILURE TO REMOVE %s from totalprice", labelString)
 					}
+					ok = a.NodeSpotRecorder.DeleteLabelValues(labels...)
+					if ok {
+						klog.Infof("removed %s from spot records", labelString)
+					} else {
+						klog.Infof("FAILURE TO REMOVE %s from spot records", labelString)
+					}
 					ok = a.CPUPriceRecorder.DeleteLabelValues(labels...)
 					if ok {
 						klog.Infof("removed %s from cpuprice", labelString)

pkg/costmodel/router.go

@@ -68,6 +68,7 @@ type Accesses struct {
 	PersistentVolumePriceRecorder *prometheus.GaugeVec
 	GPUPriceRecorder              *prometheus.GaugeVec
 	NodeTotalPriceRecorder        *prometheus.GaugeVec
+	NodeSpotRecorder              *prometheus.GaugeVec
 	RAMAllocationRecorder         *prometheus.GaugeVec
 	CPUAllocationRecorder         *prometheus.GaugeVec
 	GPUAllocationRecorder         *prometheus.GaugeVec
@@ -652,6 +653,7 @@ func (p *Accesses) ClusterInfo(w http.ResponseWriter, r *http.Request, ps httpro
 	writeReportingFlags(data)
 
 	// Include Thanos Offset Duration if Applicable
+	data["thanosEnabled"] = fmt.Sprintf("%t", thanos.IsEnabled())
 	if thanos.IsEnabled() {
 		data["thanosOffset"] = thanos.Offset()
 	}
@@ -878,6 +880,11 @@ func Initialize(additionalConfigWatchers ...ConfigWatchers) {
 		Help: "node_total_hourly_cost Total node cost per hour",
 	}, []string{"instance", "node", "instance_type", "region", "provider_id"})
 
+	spotGv := prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Name: "kubecost_node_is_spot",
+		Help: "kubecost_node_is_spot Cloud provider info about node preemptibility",
+	}, []string{"instance", "node", "instance_type", "region", "provider_id"})
+
 	pvGv := prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Name: "pv_hourly_cost",
 		Help: "pv_hourly_cost Cost per GB per hour on a persistent disk",
@@ -924,6 +931,7 @@ func Initialize(additionalConfigWatchers ...ConfigWatchers) {
 	prometheus.MustRegister(gpuGv)
 	prometheus.MustRegister(totalGv)
 	prometheus.MustRegister(pvGv)
+	prometheus.MustRegister(spotGv)
 	prometheus.MustRegister(RAMAllocation)
 	prometheus.MustRegister(CPUAllocation)
 	prometheus.MustRegister(PVAllocation)
@@ -952,6 +960,7 @@ func Initialize(additionalConfigWatchers ...ConfigWatchers) {
 		RAMPriceRecorder:              ramGv,
 		GPUPriceRecorder:              gpuGv,
 		NodeTotalPriceRecorder:        totalGv,
+		NodeSpotRecorder:              spotGv,
 		RAMAllocationRecorder:         RAMAllocation,
 		CPUAllocationRecorder:         CPUAllocation,
 		GPUAllocationRecorder:         GPUAllocation,