fix metrics query bug for pods with duplicate name in one node

Signed-off-by: junotx <junotx@126.com>

pkg/kapis/monitoring/v1alpha3/helper.go

@@ -55,6 +55,7 @@ type reqParams struct {
 	page                      string
 	limit                     string
 	metricFilter              string
+	namespacedResourcesFilter string
 	resourceFilter            string
 	nodeName                  string
 	workspaceName             string
@@ -107,6 +108,7 @@ func parseRequestParams(req *restful.Request) reqParams {
 	r.page = req.QueryParameter("page")
 	r.limit = req.QueryParameter("limit")
 	r.metricFilter = req.QueryParameter("metrics_filter")
+	r.namespacedResourcesFilter = req.QueryParameter("namespaced_resources_filter")
 	r.resourceFilter = req.QueryParameter("resources_filter")
 	r.nodeName = req.PathParameter("node")
 	r.workspaceName = req.PathParameter("workspace")
@@ -171,6 +173,7 @@ func (h handler) makeQueryOptions(r reqParams, lvl monitoring.Level) (q queryOpt
 		q.identifier = model.IdentifierPod
 		q.namedMetrics = model.PodMetrics
 		q.option = monitoring.PodOption{
+			NamespacedResourcesFilter: r.namespacedResourcesFilter,
 			ResourceFilter:            r.resourceFilter,
 			NodeName:                  r.nodeName,
 			NamespaceName:             r.namespaceName,

pkg/kapis/monitoring/v1alpha3/register.go

@@ -18,6 +18,8 @@
 package v1alpha3
 
 import (
+	"net/http"
+
 	"github.com/emicklei/go-restful"
 	"github.com/emicklei/go-restful-openapi"
 	"k8s.io/apimachinery/pkg/runtime/schema"
@@ -28,7 +30,6 @@ import (
 	model "kubesphere.io/kubesphere/pkg/models/monitoring"
 	"kubesphere.io/kubesphere/pkg/simple/client/monitoring"
 	"kubesphere.io/kubesphere/pkg/simple/client/openpitrix"
-	"net/http"
 )
 
 const (
@@ -223,7 +224,7 @@ func AddToContainer(c *restful.Container, k8sClient kubernetes.Interface, monito
 		To(h.handlePodMetricsQuery).
 		Doc("Get pod-level metric data of the whole cluster's pods.").
 		Param(ws.QueryParameter("metrics_filter", "The metric name filter consists of a regexp pattern. It specifies which metric data to return. For example, the following filter matches both pod CPU usage and memory usage: `pod_cpu_usage|pod_memory_usage`. View available metrics at [kubesphere.io](https://docs.kubesphere.io/advanced-v2.0/zh-CN/api-reference/monitoring-metrics/).").DataType("string").Required(false)).
-		Param(ws.QueryParameter("resources_filter", "The pod filter consists of a regexp pattern. It specifies which pod data to return. For example, the following filter matches any pod whose name begins with redis: `redis.*`.").DataType("string").Required(false)).
+		Param(ws.QueryParameter("namespaced_resources_filter", "Specifies a namespaced resources filter in `<namespace>/<pod_name>|<namespace>/<pod_name>` format. For example, a namespaced resources filter like `ns1/pod1|ns2/pod2` will request the data of pod1 in ns1 together with pod2 in ns2.").DataType("string").Required(false)).
 		Param(ws.QueryParameter("start", "Start time of query. Use **start** and **end** to retrieve metric data over a time span. It is a string with Unix time format, eg. 1559347200. ").DataType("string").Required(false)).
 		Param(ws.QueryParameter("end", "End time of query. Use **start** and **end** to retrieve metric data over a time span. It is a string with Unix time format, eg. 1561939200. ").DataType("string").Required(false)).
 		Param(ws.QueryParameter("step", "Time interval. Retrieve metric data at a fixed interval within the time range of start and end. It requires both **start** and **end** are provided. The format is [0-9]+[smhdwy]. Defaults to 10m (i.e. 10 min).").DataType("string").DefaultValue("10m").Required(false)).
@@ -298,6 +299,7 @@ func AddToContainer(c *restful.Container, k8sClient kubernetes.Interface, monito
 		Param(ws.PathParameter("node", "Node name.").DataType("string").Required(true)).
 		Param(ws.QueryParameter("metrics_filter", "The metric name filter consists of a regexp pattern. It specifies which metric data to return. For example, the following filter matches both pod CPU usage and memory usage: `pod_cpu_usage|pod_memory_usage`. View available metrics at [kubesphere.io](https://docs.kubesphere.io/advanced-v2.0/zh-CN/api-reference/monitoring-metrics/).").DataType("string").Required(false)).
 		Param(ws.QueryParameter("resources_filter", "The pod filter consists of a regexp pattern. It specifies which pod data to return. For example, the following filter matches any pod whose name begins with redis: `redis.*`.").DataType("string").Required(false)).
+		Param(ws.QueryParameter("namespaced_resources_filter", "Specifies a namespaced resources filter in `<namespace>/<pod_name>|<namespace>/<pod_name>` format. For example, a namespaced resources filter like `ns1/pod1|ns2/pod2` will request the data of pod1 in ns1 together with pod2 in ns2.").DataType("string").Required(false)).
 		Param(ws.QueryParameter("start", "Start time of query. Use **start** and **end** to retrieve metric data over a time span. It is a string with Unix time format, eg. 1559347200. ").DataType("string").Required(false)).
 		Param(ws.QueryParameter("end", "End time of query. Use **start** and **end** to retrieve metric data over a time span. It is a string with Unix time format, eg. 1561939200. ").DataType("string").Required(false)).
 		Param(ws.QueryParameter("step", "Time interval. Retrieve metric data at a fixed interval within the time range of start and end. It requires both **start** and **end** are provided. The format is [0-9]+[smhdwy]. Defaults to 10m (i.e. 10 min).").DataType("string").DefaultValue("10m").Required(false)).
@@ -313,7 +315,8 @@ func AddToContainer(c *restful.Container, k8sClient kubernetes.Interface, monito
 
 	ws.Route(ws.GET("/nodes/{node}/pods/{pod}").
 		To(h.handlePodMetricsQuery).
-		Doc("Get pod-level metric data of a specific pod. Navigate to the pod by the node where it is scheduled.").
+		Deprecate().
+		Doc("Get pod-level metric data of a specific pod. Navigate to the pod by the node where it is scheduled. Deprecated: please use `/namespaces/{namespace}/pods/{pod}` or `/pods` instead.").
 		Param(ws.PathParameter("node", "Node name.").DataType("string").Required(true)).
 		Param(ws.PathParameter("pod", "Pod name.").DataType("string").Required(true)).
 		Param(ws.QueryParameter("metrics_filter", "The metric name filter consists of a regexp pattern. It specifies which metric data to return. For example, the following filter matches both pod CPU usage and memory usage: `pod_cpu_usage|pod_memory_usage`. View available metrics at [kubesphere.io](https://docs.kubesphere.io/advanced-v2.0/zh-CN/api-reference/monitoring-metrics/).").DataType("string").Required(false)).

pkg/simple/client/monitoring/prometheus/prometheus.go

@@ -19,13 +19,15 @@ package prometheus
 import (
 	"context"
 	"fmt"
+	"strings"
+	"sync"
+	"time"
+
 	"github.com/prometheus/client_golang/api"
 	apiv1 "github.com/prometheus/client_golang/api/prometheus/v1"
 	"github.com/prometheus/common/model"
 	"k8s.io/klog"
 	"kubesphere.io/kubesphere/pkg/simple/client/monitoring"
-	"sync"
-	"time"
 )
 
 // prometheus implements monitoring interface backed by Prometheus
@@ -49,7 +51,7 @@ func (p prometheus) GetMetric(expr string, ts time.Time) monitoring.Metric {
 	if err != nil {
 		parsedResp.Error = err.Error()
 	} else {
-		parsedResp.MetricData = parseQueryResp(value)
+		parsedResp.MetricData = parseQueryResp(value, nil)
 	}
 
 	return parsedResp
@@ -68,7 +70,7 @@ func (p prometheus) GetMetricOverTime(expr string, start, end time.Time, step ti
 	if err != nil {
 		parsedResp.Error = err.Error()
 	} else {
-		parsedResp.MetricData = parseQueryRangeResp(value)
+		parsedResp.MetricData = parseQueryRangeResp(value, nil)
 	}
 	return parsedResp
 }
@@ -90,7 +92,7 @@ func (p prometheus) GetNamedMetrics(metrics []string, ts time.Time, o monitoring
 			if err != nil {
 				parsedResp.Error = err.Error()
 			} else {
-				parsedResp.MetricData = parseQueryResp(value)
+				parsedResp.MetricData = parseQueryResp(value, genMetricFilter(o))
 			}
 
 			mtx.Lock()
@@ -129,7 +131,7 @@ func (p prometheus) GetNamedMetricsOverTime(metrics []string, start, end time.Ti
 			if err != nil {
 				parsedResp.Error = err.Error()
 			} else {
-				parsedResp.MetricData = parseQueryRangeResp(value)
+				parsedResp.MetricData = parseQueryRangeResp(value, genMetricFilter(o))
 			}
 
 			mtx.Lock()
@@ -200,12 +202,15 @@ func (p prometheus) GetMetricLabelSet(expr string, start, end time.Time) []map[s
 	return res
 }
 
-func parseQueryRangeResp(value model.Value) monitoring.MetricData {
+func parseQueryRangeResp(value model.Value, metricFilter func(metric model.Metric) bool) monitoring.MetricData {
 	res := monitoring.MetricData{MetricType: monitoring.MetricTypeMatrix}
 
 	data, _ := value.(model.Matrix)
 
 	for _, v := range data {
+		if metricFilter != nil && !metricFilter(v.Metric) {
+			continue
+		}
 		mv := monitoring.MetricValue{
 			Metadata: make(map[string]string),
 		}
@@ -224,12 +229,15 @@ func parseQueryRangeResp(value model.Value) monitoring.MetricData {
 	return res
 }
 
-func parseQueryResp(value model.Value) monitoring.MetricData {
+func parseQueryResp(value model.Value, metricFilter func(metric model.Metric) bool) monitoring.MetricData {
 	res := monitoring.MetricData{MetricType: monitoring.MetricTypeVector}
 
 	data, _ := value.(model.Vector)
 
 	for _, v := range data {
+		if metricFilter != nil && !metricFilter(v.Metric) {
+			continue
+		}
 		mv := monitoring.MetricValue{
 			Metadata: make(map[string]string),
 		}
@@ -245,3 +253,26 @@ func parseQueryResp(value model.Value) monitoring.MetricData {
 
 	return res
 }
+
+func genMetricFilter(o monitoring.QueryOption) func(metric model.Metric) bool {
+	if o != nil {
+		if po, ok := o.(monitoring.PodOption); ok {
+			if po.NamespacedResourcesFilter != "" {
+				namespacedPodsMap := make(map[string]struct{})
+				for _, s := range strings.Split(po.NamespacedResourcesFilter, "|") {
+					namespacedPodsMap[s] = struct{}{}
+				}
+				return func(metric model.Metric) bool {
+					if len(metric) == 0 {
+						return false
+					}
+					_, ok := namespacedPodsMap[string(metric["namespace"])+"/"+string(metric["pod"])]
+					return ok
+				}
+			}
+		}
+	}
+	return func(metric model.Metric) bool {
+		return true
+	}
+}

pkg/simple/client/monitoring/prometheus/promql.go

@@ -15,8 +15,9 @@ package prometheus
 
 import (
 	"fmt"
-	"kubesphere.io/kubesphere/pkg/simple/client/monitoring"
 	"strings"
+
+	"kubesphere.io/kubesphere/pkg/simple/client/monitoring"
 )
 
 const (
@@ -349,12 +350,6 @@ func makePodMetricExpr(tmpl string, o monitoring.QueryOptions) string {
 		}
 	}
 
-	// For monitoring pods in the whole cluster
-	// Get /pods
-	if o.NamespaceName == "" && o.NodeName == "" {
-		podSelector = fmt.Sprintf(`pod=~"%s"`, o.ResourceFilter)
-	}
-
 	// For monitoring pods in the specific namespace
 	// GET /namespaces/{namespace}/workloads/{kind}/{workload}/pods or
 	// GET /namespaces/{namespace}/pods/{pod} or
@@ -365,17 +360,59 @@ func makePodMetricExpr(tmpl string, o monitoring.QueryOptions) string {
 		} else {
 			podSelector = fmt.Sprintf(`pod=~"%s", namespace="%s"`, o.ResourceFilter, o.NamespaceName)
 		}
+	} else {
+		var namespaces, pods []string
+		if o.NamespacedResourcesFilter != "" {
+			for _, np := range strings.Split(o.NamespacedResourcesFilter, "|") {
+				if nparr := strings.SplitN(np, "/", 2); len(nparr) > 1 {
+					namespaces = append(namespaces, nparr[0])
+					pods = append(pods, nparr[1])
+				} else {
+					pods = append(pods, np)
+				}
+			}
 		}
-
 		// For monitoring pods on the specific node
 		// GET /nodes/{node}/pods/{pod}
+		// GET /nodes/{node}/pods
 		if o.NodeName != "" {
 			if o.PodName != "" {
-			podSelector = fmt.Sprintf(`pod="%s", node="%s"`, o.PodName, o.NodeName)
+				if nparr := strings.SplitN(o.PodName, "/", 2); len(nparr) > 1 {
+					podSelector = fmt.Sprintf(`namespace="%s",pod="%s", node="%s"`, nparr[0], nparr[1], o.NodeName)
 				} else {
-			podSelector = fmt.Sprintf(`pod=~"%s", node="%s"`, o.ResourceFilter, o.NodeName)
+					podSelector = fmt.Sprintf(`pod="%s", node="%s"`, o.PodName, o.NodeName)
+				}
+			} else {
+				var ps []string
+				ps = append(ps, fmt.Sprintf(`node="%s"`, o.NodeName))
+				if o.ResourceFilter != "" {
+					ps = append(ps, fmt.Sprintf(`pod=~"%s"`, o.ResourceFilter))
+				}
+
+				if len(namespaces) > 0 {
+					ps = append(ps, fmt.Sprintf(`namespace=~"%s"`, strings.Join(namespaces, "|")))
+				}
+				if len(pods) > 0 {
+					ps = append(ps, fmt.Sprintf(`pod=~"%s"`, strings.Join(pods, "|")))
+				}
+				podSelector = strings.Join(ps, ",")
+			}
+		} else {
+			// For monitoring pods in the whole cluster
+			// Get /pods
+			var ps []string
+			if len(namespaces) > 0 {
+				ps = append(ps, fmt.Sprintf(`namespace=~"%s"`, strings.Join(namespaces, "|")))
+			}
+			if len(pods) > 0 {
+				ps = append(ps, fmt.Sprintf(`pod=~"%s"`, strings.Join(pods, "|")))
+			}
+			if len(ps) > 0 {
+				podSelector = strings.Join(ps, ",")
 			}
 		}
+	}
+
 	return strings.NewReplacer("$1", workloadSelector, "$2", podSelector).Replace(tmpl)
 }
 

pkg/simple/client/monitoring/prometheus/promql_test.go

@@ -139,7 +139,7 @@ func TestMakeExpr(t *testing.T) {
 			name: "pod_net_bytes_transmitted",
 			opts: monitoring.QueryOptions{
 				Level:                     monitoring.LevelPod,
-				ResourceFilter: "elasticsearch-0",
+				NamespacedResourcesFilter: "logging/elasticsearch-0|ks/redis",
 			},
 		},
 		{

pkg/simple/client/monitoring/prometheus/testdata/promqls.go

@@ -31,7 +31,7 @@ var PromQLs = map[string]string{
 	"pod_cpu_usage":                         `round(sum by (namespace, pod) (irate(container_cpu_usage_seconds_total{job="kubelet", pod!="", image!=""}[5m])) * on (namespace, pod) group_left(owner_kind, owner_name) kube_pod_owner{owner_kind="ReplicaSet", owner_name=~"^elasticsearch-[^-]{1,10}$"} * on (namespace, pod) group_left(node) kube_pod_info{pod=~"elasticsearch-0", namespace="default"}, 0.001)`,
 	"pod_memory_usage":                      `sum by (namespace, pod) (container_memory_usage_bytes{job="kubelet", pod!="", image!=""}) * on (namespace, pod) group_left(owner_kind, owner_name) kube_pod_owner{} * on (namespace, pod) group_left(node) kube_pod_info{pod="elasticsearch-12345", namespace="default"}`,
 	"pod_memory_usage_wo_cache":             `sum by (namespace, pod) (container_memory_working_set_bytes{job="kubelet", pod!="", image!=""}) * on (namespace, pod) group_left(owner_kind, owner_name) kube_pod_owner{} * on (namespace, pod) group_left(node) kube_pod_info{pod="elasticsearch-12345", node="i-2dazc1d6"}`,
-	"pod_net_bytes_transmitted":             `sum by (namespace, pod) (irate(container_network_transmit_bytes_total{pod!="", interface!~"^(cali.+|tunl.+|dummy.+|kube.+|flannel.+|cni.+|docker.+|veth.+|lo.*)", job="kubelet"}[5m])) * on (namespace, pod) group_left(owner_kind, owner_name) kube_pod_owner{} * on (namespace, pod) group_left(node) kube_pod_info{pod=~"elasticsearch-0"}`,
+	"pod_net_bytes_transmitted":             `sum by (namespace, pod) (irate(container_network_transmit_bytes_total{pod!="", interface!~"^(cali.+|tunl.+|dummy.+|kube.+|flannel.+|cni.+|docker.+|veth.+|lo.*)", job="kubelet"}[5m])) * on (namespace, pod) group_left(owner_kind, owner_name) kube_pod_owner{} * on (namespace, pod) group_left(node) kube_pod_info{namespace=~"logging|ks",pod=~"elasticsearch-0|redis"}`,
 	"container_cpu_usage":                   `round(sum by (namespace, pod, container) (irate(container_cpu_usage_seconds_total{job="kubelet", container!="POD", container!="", image!="", pod="elasticsearch-12345", namespace="default", container="syscall"}[5m])), 0.001)`,
 	"container_memory_usage":                `sum by (namespace, pod, container) (container_memory_usage_bytes{job="kubelet", container!="POD", container!="", image!="", pod="elasticsearch-12345", namespace="default", container=~"syscall"})`,
 	"pvc_inodes_available":                  `max by (namespace, persistentvolumeclaim) (kubelet_volume_stats_inodes_free) * on (namespace, persistentvolumeclaim) group_left (storageclass) kube_persistentvolumeclaim_info{namespace="default", persistentvolumeclaim="db-123"}`,

pkg/simple/client/monitoring/query_options.go

@@ -37,6 +37,7 @@ type QueryOption interface {
 type QueryOptions struct {
 	Level Level
 
+	NamespacedResourcesFilter string
 	ResourceFilter            string
 	NodeName                  string
 	WorkspaceName             string
@@ -108,6 +109,7 @@ func (wo WorkloadOption) Apply(o *QueryOptions) {
 }
 
 type PodOption struct {
+	NamespacedResourcesFilter string
 	ResourceFilter            string
 	NodeName                  string
 	NamespaceName             string
@@ -118,6 +120,7 @@ type PodOption struct {
 
 func (po PodOption) Apply(o *QueryOptions) {
 	o.Level = LevelPod
+	o.NamespacedResourcesFilter = po.NamespacedResourcesFilter
 	o.ResourceFilter = po.ResourceFilter
 	o.NodeName = po.NodeName
 	o.NamespaceName = po.NamespaceName