Feature/remove k8s cache (#32539)

* replaced internal expiring cache with non expiring dictionary in memory * fixed a bug that prevented to export pod/container metrics when node/state_node metricsets were disabled (cherry picked from commit 5503761) # Conflicts: # metricbeat/module/kubernetes/container/container.go # metricbeat/module/kubernetes/container/container_test.go # metricbeat/module/kubernetes/container/data.go # metricbeat/module/kubernetes/pod/data.go # metricbeat/module/kubernetes/pod/pod.go # metricbeat/module/kubernetes/pod/pod_test.go # metricbeat/module/kubernetes/state_cronjob/state_cronjob.go # metricbeat/module/kubernetes/state_persistentvolume/state_persistentvolume.go # metricbeat/module/kubernetes/state_persistentvolumeclaim/state_persistentvolumeclaim.go # metricbeat/module/kubernetes/util/kubernetes.go
elastic · Aug 15, 2022 · 66333a7 · 66333a7
1 parent 3cbdcbe
commit 66333a7
Show file tree

Hide file tree

Showing 27 changed files with 1,399 additions and 273 deletions.
diff --git a/CHANGELOG.next.asciidoc b/CHANGELOG.next.asciidoc
@@ -54,6 +54,13 @@ https://github.com/elastic/beats/compare/v8.1.3\...main[Check the HEAD diff]
 
 - Improve handling of disabled commands in Zookeeper Metricbeat module. {pull}31013[#31013]
 - Fix kubernetes module's internal cache expiration issue. This avoid metrics like `kubernetes.container.cpu.usage.limit.pct` from not being populated. {pull}31785[31785]
+- update kubernetes apiserver metricset to not collect deprecated metrics and fix dashboard {pull}31973[31973]
+- Check for nil metadata in GCP. {pull}32281[32281]
+- update kubernetes controllermanager metricset to not collect deprecated metrics and fix dashboard {pull}32037[32037]
+- Fix to ARN parsing for Cloudwatch resource names with leading slashes {pull}32358[32358]
+- Fix an infinite loop in AWS billing metricset. {pull}32626[32626]
+- Add missing metrics in AWS Transit Gateway module {pull}32617[32617]
+- Replace internal expiring cache used by the Kubernetes module with in-memory dictionary {pull}32539[32539]
 
 *Packetbeat*
 

diff --git a/metricbeat/module/kubernetes/_meta/test/stats_summary_multiple_containers.json b/metricbeat/module/kubernetes/_meta/test/stats_summary_multiple_containers.json
@@ -0,0 +1,163 @@
+{
+    "node": {
+     "nodeName": "gke-beats-default-pool-a5b33e2e-hdww",
+     "systemContainers": [
+      {
+       "name": "kubelet",
+       "startTime": "2017-04-18T12:53:49Z",
+       "cpu": {
+        "time": "2017-04-20T08:06:46Z",
+        "usageNanoCores": 11263994,
+        "usageCoreNanoSeconds": 2357800908948
+       },
+       "memory": {
+        "time": "2017-04-20T08:06:46Z",
+        "usageBytes": 36683776,
+        "workingSetBytes": 36495360,
+        "rssBytes": 35512320,
+        "pageFaults": 100835242,
+        "majorPageFaults": 0
+       },
+       "userDefinedMetrics": null
+      }
+     ],
+     "startTime": "2017-04-18T12:53:49Z",
+     "cpu": {
+      "time": "2017-04-20T08:06:41Z",
+      "usageNanoCores": 18691146,
+      "usageCoreNanoSeconds": 4189523881380
+     },
+     "memory": {
+      "time": "2017-04-20T08:06:41Z",
+      "availableBytes": 1768316928,
+      "usageBytes": 2764943360,
+      "workingSetBytes": 2111090688,
+      "rssBytes": 2150400,
+      "pageFaults": 131567,
+      "majorPageFaults": 103
+     },
+     "network": {
+      "time": "2017-04-20T08:06:41Z",
+      "rxBytes": 1115133198,
+      "rxErrors": 0,
+      "txBytes": 812729002,
+      "txErrors": 0
+     },
+     "fs": {
+      "availableBytes": 98727014400,
+      "capacityBytes": 101258067968,
+      "usedBytes": 2514276352,
+      "inodesFree": 18446744073709551615,
+      "inodes": 6258720,
+      "inodesUsed": 138624
+     },
+     "runtime": {
+      "imageFs": {
+       "availableBytes": 98727014400,
+       "capacityBytes": 101258067968,
+       "usedBytes": 860204379,
+       "inodesFree": 18446744073709551615,
+       "inodes": 6258720,
+       "inodesUsed": 138624
+      }
+     }
+    },
+    "pods": [
+     {
+      "podRef": {
+       "name": "nginx-deployment-2303442956-pcqfc",
+       "namespace": "default",
+       "uid": "beabc196-2456-11e7-a3ad-42010a840235"
+      },
+      "startTime": "2017-04-18T16:47:44Z",
+      "containers": [
+       {
+        "name": "nginx",
+        "startTime": "2017-04-18T16:47:44Z",
+        "cpu": {
+         "time": "2017-04-20T08:06:34Z",
+         "usageNanoCores": 11263994,
+         "usageCoreNanoSeconds": 43959424
+        },
+        "memory": {
+         "time": "2017-04-20T08:06:34Z",
+         "usageBytes": 1462272,
+         "workingSetBytes": 1454080,
+         "rssBytes": 1409024,
+         "pageFaults": 841,
+         "majorPageFaults": 0
+        },
+        "rootfs": {
+         "availableBytes": 98727014400,
+         "capacityBytes": 101258067968,
+         "usedBytes": 61440,
+         "inodesFree": 6120096,
+         "inodes": 6258720,
+         "inodesUsed": 21
+        },
+        "logs": {
+         "availableBytes": 98727014400,
+         "capacityBytes": 101258067968,
+         "usedBytes": 28672,
+         "inodesFree": 6120096,
+         "inodes": 6258720,
+         "inodesUsed": 138624
+        },
+        "userDefinedMetrics": null
+       },
+       {
+        "name": "sidecar",
+        "startTime": "2017-04-18T16:47:44Z",
+        "cpu": {
+         "time": "2017-04-20T08:06:34Z",
+         "usageNanoCores": 11263994,
+         "usageCoreNanoSeconds": 43959424
+        },
+        "memory": {
+         "time": "2017-04-20T08:06:34Z",
+         "usageBytes": 1462272,
+         "workingSetBytes": 1454080,
+         "rssBytes": 1409024,
+         "pageFaults": 841,
+         "majorPageFaults": 0
+        },
+        "rootfs": {
+         "availableBytes": 98727014400,
+         "capacityBytes": 101258067968,
+         "usedBytes": 61440,
+         "inodesFree": 6120096,
+         "inodes": 6258720,
+         "inodesUsed": 21
+        },
+        "logs": {
+         "availableBytes": 98727014400,
+         "capacityBytes": 101258067968,
+         "usedBytes": 28672,
+         "inodesFree": 6120096,
+         "inodes": 6258720,
+         "inodesUsed": 138624
+        },
+        "userDefinedMetrics": null
+       }
+      ],
+      "network": {
+       "time": "2017-04-20T08:06:41Z",
+       "rxBytes": 107056,
+       "rxErrors": 0,
+       "txBytes": 72447,
+       "txErrors": 0
+      },
+      "volume": [
+       {
+        "availableBytes": 1939689472,
+        "capacityBytes": 1939701760,
+        "usedBytes": 12288,
+        "inodesFree": 473551,
+        "inodes": 473560,
+        "inodesUsed": 9,
+        "name": "default-token-sg8x5"
+       }
+      ]
+     }
+    ]
+   }
diff --git a/metricbeat/module/kubernetes/container/container.go b/metricbeat/module/kubernetes/container/container.go
@@ -77,7 +77,7 @@ func New(base mb.BaseMetricSet) (mb.MetricSet, error) {
 	return &MetricSet{
 		BaseMetricSet: base,
 		http:          http,
-		enricher:      util.NewContainerMetadataEnricher(base, mod.GetPerfMetricsCache(), true),
+		enricher:      util.NewContainerMetadataEnricher(base, mod.GetMetricsRepo(), true),
 		mod:           mod,
 	}, nil
 }
@@ -95,7 +95,11 @@ func (m *MetricSet) Fetch(reporter mb.ReporterV2) {
 		return
 	}
 
+<<<<<<< HEAD
 	events, err := eventMapping(body, m.mod.GetPerfMetricsCache())
+=======
+	events, err := eventMapping(body, m.mod.GetMetricsRepo(), m.Logger())
+>>>>>>> 5503761995 (Feature/remove k8s cache (#32539))
 	if err != nil {
 		m.Logger().Error(err)
 		reporter.Error(err)

diff --git a/metricbeat/module/kubernetes/container/container_test.go b/metricbeat/module/kubernetes/container/container_test.go
@@ -24,7 +24,6 @@ import (
 	"io/ioutil"
 	"os"
 	"testing"
-	"time"
 
 	"github.com/stretchr/testify/assert"
 
@@ -41,12 +40,29 @@ func TestEventMapping(t *testing.T) {
 	body, err := ioutil.ReadAll(f)
 	assert.NoError(t, err, "cannot read test file "+testFile)
 
-	cache := util.NewPerfMetricsCache(120 * time.Second)
-	cache.NodeCoresAllocatable.Set("gke-beats-default-pool-a5b33e2e-hdww", 2)
-	cache.NodeMemAllocatable.Set("gke-beats-default-pool-a5b33e2e-hdww", 146227200)
-	cache.ContainerMemLimit.Set(util.ContainerUID("default", "nginx-deployment-2303442956-pcqfc", "nginx"), 14622720)
+	metricsRepo := util.NewMetricsRepo()
 
+<<<<<<< HEAD
 	events, err := eventMapping(body, cache)
+=======
+	nodeName := "gke-beats-default-pool-a5b33e2e-hdww"
+
+	nodeMetrics := util.NewNodeMetrics()
+	nodeMetrics.CoresAllocatable = util.NewFloat64Metric(2)
+	nodeMetrics.MemoryAllocatable = util.NewFloat64Metric(146227200)
+	addNodeMetric(metricsRepo, nodeName, nodeMetrics)
+
+	namespace := "default"
+	podName := "nginx-deployment-2303442956-pcqfc"
+	podId := util.NewPodId(namespace, podName)
+	containerName := "nginx"
+
+	containerMetrics := util.NewContainerMetrics()
+	containerMetrics.MemoryLimit = util.NewFloat64Metric(14622720)
+	addContainerMetric(metricsRepo, nodeName, podId, containerName, containerMetrics)
+
+	events, err := eventMapping(body, metricsRepo, logger)
+>>>>>>> 5503761995 (Feature/remove k8s cache (#32539))
 	assert.NoError(t, err, "error mapping "+testFile)
 
 	assert.Len(t, events, 1, "got wrong number of events")
@@ -94,3 +110,15 @@ func testValue(t *testing.T, event common.MapStr, field string, value interface{
 	assert.NoError(t, err, "Could not read field "+field)
 	assert.EqualValues(t, data, value, "Wrong value for field "+field)
 }
+
+func addContainerMetric(metricsRepo *util.MetricsRepo, nodeName string, podId util.PodId, containerName string, metrics *util.ContainerMetrics) {
+	nodeStore, _ := metricsRepo.AddNodeStore(nodeName)
+	podStore, _ := nodeStore.AddPodStore(podId)
+	containerStore, _ := podStore.AddContainerStore(containerName)
+	containerStore.SetContainerMetrics(metrics)
+}
+
+func addNodeMetric(metricsRepo *util.MetricsRepo, nodeName string, nodeMetrics *util.NodeMetrics) {
+	nodeStore, _ := metricsRepo.AddNodeStore(nodeName)
+	nodeStore.SetNodeMetrics(nodeMetrics)
+}
diff --git a/metricbeat/module/kubernetes/container/data.go b/metricbeat/module/kubernetes/container/data.go
@@ -27,8 +27,13 @@ import (
 	"github.com/elastic/beats/v7/metricbeat/module/kubernetes/util"
 )
 
+<<<<<<< HEAD
 func eventMapping(content []byte, perfMetrics *util.PerfMetricsCache) ([]common.MapStr, error) {
 	events := []common.MapStr{}
+=======
+func eventMapping(content []byte, metricsRepo *util.MetricsRepo, logger *logp.Logger) ([]mapstr.M, error) {
+	events := []mapstr.M{}
+>>>>>>> 5503761995 (Feature/remove k8s cache (#32539))
 	var summary kubernetes.Summary
 
 	err := json.Unmarshal(content, &summary)
@@ -37,9 +42,23 @@ func eventMapping(content []byte, perfMetrics *util.PerfMetricsCache) ([]common.
 	}
 
 	node := summary.Node
-	nodeCores := perfMetrics.NodeCoresAllocatable.Get(node.NodeName)
-	nodeMem := perfMetrics.NodeMemAllocatable.Get(node.NodeName)
+
+	nodeCores := 0.0
+	nodeMem := 0.0
+
+	nodeStore := metricsRepo.GetNodeStore(node.NodeName)
+	nodeMetrics := nodeStore.GetNodeMetrics()
+	if nodeMetrics.CoresAllocatable != nil {
+		nodeCores = nodeMetrics.CoresAllocatable.Value
+	}
+	if nodeMetrics.MemoryAllocatable != nil {
+		nodeMem = nodeMetrics.MemoryAllocatable.Value
+	}
+
 	for _, pod := range summary.Pods {
+		podId := util.NewPodId(pod.PodRef.Namespace, pod.PodRef.Name)
+		podStore := nodeStore.GetPodStore(podId)
+
 		for _, container := range pod.Containers {
 			containerEvent := common.MapStr{
 				mb.ModuleDataKey: common.MapStr{
@@ -125,17 +144,41 @@ func eventMapping(content []byte, perfMetrics *util.PerfMetricsCache) ([]common.
 				containerEvent.Put("memory.usage.node.pct", float64(container.Memory.UsageBytes)/nodeMem)
 			}
 
-			cuid := util.ContainerUID(pod.PodRef.Namespace, pod.PodRef.Name, container.Name)
-			coresLimit := perfMetrics.ContainerCoresLimit.GetWithDefault(cuid, nodeCores)
-			memLimit := perfMetrics.ContainerMemLimit.GetWithDefault(cuid, nodeMem)
+			containerStore := podStore.GetContainerStore(container.Name)
+			containerMetrics := containerStore.GetContainerMetrics()
 
+			containerCoresLimit := nodeCores
+			if containerMetrics.CoresLimit != nil {
+				containerCoresLimit = containerMetrics.CoresLimit.Value
+			}
+
+			containerMemLimit := nodeMem
+			if containerMetrics.MemoryLimit != nil {
+				containerMemLimit = containerMetrics.MemoryLimit.Value
+			}
+
+			// NOTE:
+			// we don't currently check if `containerMemLimit` > `nodeMem` as we do in `kubernetes/pod/data.go`.
+			// There we do check, since if a container doesn't have a limit set, it will inherit the node limits and the sum of all
+			// the container limits can be greater than the node limits. We assume here the user can set correct limits on containers.
+
+<<<<<<< HEAD
 			if coresLimit > 0 {
 				containerEvent.Put("cpu.usage.limit.pct", float64(container.CPU.UsageNanoCores)/1e9/coresLimit)
 			}
 
 			if memLimit > 0 {
 				containerEvent.Put("memory.usage.limit.pct", float64(container.Memory.UsageBytes)/memLimit)
 				containerEvent.Put("memory.workingset.limit.pct", float64(container.Memory.WorkingSetBytes)/memLimit)
+=======
+			if containerCoresLimit > 0 {
+				kubernetes2.ShouldPut(containerEvent, "cpu.usage.limit.pct", float64(container.CPU.UsageNanoCores)/1e9/containerCoresLimit, logger)
+			}
+
+			if containerMemLimit > 0 {
+				kubernetes2.ShouldPut(containerEvent, "memory.usage.limit.pct", float64(container.Memory.UsageBytes)/containerMemLimit, logger)
+				kubernetes2.ShouldPut(containerEvent, "memory.workingset.limit.pct", float64(container.Memory.WorkingSetBytes)/containerMemLimit, logger)
+>>>>>>> 5503761995 (Feature/remove k8s cache (#32539))
 			}
 
 			events = append(events, containerEvent)

diff --git a/metricbeat/module/kubernetes/kubernetes.go b/metricbeat/module/kubernetes/kubernetes.go
@@ -42,7 +42,7 @@ type Module interface {
 	mb.Module
 	GetStateMetricsFamilies(prometheus p.Prometheus) ([]*dto.MetricFamily, error)
 	GetKubeletStats(http *helper.HTTP) ([]byte, error)
-	GetPerfMetricsCache() *util.PerfMetricsCache
+	GetMetricsRepo() *util.MetricsRepo
 }
 
 type familiesCache struct {
@@ -86,7 +86,7 @@ type module struct {
 
 	kubeStateMetricsCache *kubeStateMetricsCache
 	kubeletStatsCache     *kubeletStatsCache
-	perfMetrics           *util.PerfMetricsCache
+	metricsRepo           *util.MetricsRepo
 	cacheHash             uint64
 }
 
@@ -97,25 +97,18 @@ func ModuleBuilder() func(base mb.BaseModule) (mb.Module, error) {
 	kubeletStatsCache := &kubeletStatsCache{
 		cacheMap: make(map[uint64]*statsCache),
 	}
-	perfMetrics := util.NewPerfMetricsCache(0)
+	metricsRepo := util.NewMetricsRepo()
 	return func(base mb.BaseModule) (mb.Module, error) {
 		hash, err := generateCacheHash(base.Config().Hosts)
 		if err != nil {
 			return nil, errors.Wrap(err, "error generating cache hash for kubeStateMetricsCache")
 		}
 
-		// NOTE: `Period * 2` is an arbitrary value to make the cache NEVER to expire before the next scraping run
-		// if different metricsets have different periods, we will effectively set (timeout = max(Period) * 2)
-		minCacheExpirationTime := base.Config().Period * 2
-		if perfMetrics.GetTimeout() < minCacheExpirationTime {
-			perfMetrics.SetOrUpdateTimeout(minCacheExpirationTime)
-		}
-
 		m := module{
 			BaseModule:            base,
 			kubeStateMetricsCache: kubeStateMetricsCache,
 			kubeletStatsCache:     kubeletStatsCache,
-			perfMetrics:           perfMetrics,
+			metricsRepo:           metricsRepo,
 			cacheHash:             hash,
 		}
 		return &m, nil
@@ -167,6 +160,6 @@ func generateCacheHash(host []string) (uint64, error) {
 	return id, nil
 }
 
-func (m *module) GetPerfMetricsCache() *util.PerfMetricsCache {
-	return m.perfMetrics
+func (m *module) GetMetricsRepo() *util.MetricsRepo {
+	return m.metricsRepo
 }