feat(advisor): add cpu usage ratio for pod and node custom metrics

kubewharf · Sep 29, 2024 · 26a9a5b · 26a9a5b
1 parent da33769
commit 26a9a5b
Show file tree

Hide file tree

Showing 6 changed files with 24 additions and 11 deletions.
diff --git a/go.mod b/go.mod
@@ -18,7 +18,7 @@ require (
 	github.com/google/uuid v1.3.0
 	github.com/h2non/gock v1.2.0
 	github.com/klauspost/cpuid/v2 v2.2.6
-	github.com/kubewharf/katalyst-api v0.5.1-0.20240911051124-d5c09c115a19
+	github.com/kubewharf/katalyst-api v0.5.1-0.20240929080014-ae613a8935d8
 	github.com/montanaflynn/stats v0.7.1
 	github.com/opencontainers/runc v1.1.6
 	github.com/opencontainers/selinux v1.10.0

diff --git a/go.sum b/go.sum
@@ -568,8 +568,8 @@ github.com/kr/pty v1.1.1/go.mod h1:pFQYn66WHrOpPYNljwOMqo10TkYh1fy3cYio2l3bCsQ=
 github.com/kr/text v0.1.0/go.mod h1:4Jbv+DJW3UT/LiOwJeYQe1efqtUx/iVham/4vfdArNI=
 github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
 github.com/kr/text v0.2.0/go.mod h1:eLer722TekiGuMkidMxC/pM04lWEeraHUUmBw8l2grE=
-github.com/kubewharf/katalyst-api v0.5.1-0.20240911051124-d5c09c115a19 h1:5kz2u+tLNzW94ByL9yfsfkvE4N8n/r6C1TKDiFWK8rw=
-github.com/kubewharf/katalyst-api v0.5.1-0.20240911051124-d5c09c115a19/go.mod h1:Y2IeIorxQamF2a3oa0+URztl5QCSty6Jj3zD83R8J9k=
+github.com/kubewharf/katalyst-api v0.5.1-0.20240929080014-ae613a8935d8 h1:rIwZVD4iia7pTiB6h1xR8muc4jx4GgHWzhVCPKlEhXw=
+github.com/kubewharf/katalyst-api v0.5.1-0.20240929080014-ae613a8935d8/go.mod h1:Y2IeIorxQamF2a3oa0+URztl5QCSty6Jj3zD83R8J9k=
 github.com/kubewharf/kubelet v1.24.6-kubewharf.9 h1:jOTYZt7h/J7I8xQMKMUcJjKf5UFBv37jHWvNp5VRFGc=
 github.com/kubewharf/kubelet v1.24.6-kubewharf.9/go.mod h1:MxbSZUx3wXztFneeelwWWlX7NAAStJ6expqq7gY2J3c=
 github.com/kyoh86/exportloopref v0.1.7/go.mod h1:h1rDl2Kdj97+Kwh4gdz3ujE7XHmH51Q0lUiZ1z4NLj8=

diff --git a/pkg/agent/sysadvisor/plugin/metric-emitter/syncer/node/node.go b/pkg/agent/sysadvisor/plugin/metric-emitter/syncer/node/node.go
@@ -43,7 +43,8 @@ import (
 // nodeRawMetricNameMapping maps the raw metricName (collected from agent.MetricsFetcher)
 // to the standard metricName (used by custom-metric-api-server)
 var nodeRawMetricNameMapping = map[string]string{
-	consts.MetricLoad1MinSystem: apimetricnode.CustomMetricNodeCPULoad1Min,
+	consts.MetricLoad1MinSystem:      apimetricnode.CustomMetricNodeCPULoad1Min,
+	consts.MetricCPUUsageRatioSystem: apimetricnode.CustomMetricNodeCPUUsageRatio,
 
 	consts.MetricMemFreeSystem:      apimetricnode.CustomMetricNodeMemoryFree,
 	consts.MetricMemAvailableSystem: apimetricnode.CustomMetricNodeMemoryAvailable,

diff --git a/pkg/agent/sysadvisor/plugin/metric-emitter/syncer/pod/pod.go b/pkg/agent/sysadvisor/plugin/metric-emitter/syncer/pod/pod.go
@@ -53,9 +53,10 @@ const (
 // podRawMetricNameMapping maps the raw metricName (collected from agent.MetricsFetcher)
 // to the standard metricName (used by custom-metric-api-server)
 var podRawMetricNameMapping = map[string]string{
-	consts.MetricLoad1MinContainer: apimetricpod.CustomMetricPodCPULoad1Min,
-	consts.MetricCPUUsageContainer: apimetricpod.CustomMetricPodCPUUsage,
-	consts.MetricCPUCPIContainer:   apimetricpod.CustomMetricPodCPUCPI,
+	consts.MetricLoad1MinContainer:      apimetricpod.CustomMetricPodCPULoad1Min,
+	consts.MetricCPUUsageContainer:      apimetricpod.CustomMetricPodCPUUsage,
+	consts.MetricCPUUsageRatioContainer: apimetricpod.CustomMetricPodCPUUsageRatio,
+	consts.MetricCPUCPIContainer:        apimetricpod.CustomMetricPodCPUCPI,
 
 	consts.MetricMemRssContainer:   apimetricpod.CustomMetricPodMemoryRSS,
 	consts.MetricMemUsageContainer: apimetricpod.CustomMetricPodMemoryUsage,

diff --git a/pkg/consts/metric.go b/pkg/consts/metric.go
@@ -30,6 +30,8 @@ const (
 
 // System compute metrics
 const (
+	MetricCPUUsageRatioSystem = "cpu.usage.ratio.system"
+
 	MetricLoad1MinSystem  = "cpu.load.1min.system"
 	MetricLoad5MinSystem  = "cpu.load.5min.system"
 	MetricLoad15MinSystem = "cpu.load.15min.system"
@@ -170,10 +172,11 @@ const (
 
 // container cpu metrics
 const (
-	MetricCPULimitContainer     = "cpu.limit.container"
-	MetricCPUUsageContainer     = "cpu.usage.container"
-	MetricCPUUsageUserContainer = "cpu.usage.user.container"
-	MetricCPUUsageSysContainer  = "cpu.usage.sys.container"
+	MetricCPULimitContainer      = "cpu.limit.container"
+	MetricCPUUsageContainer      = "cpu.usage.container"
+	MetricCPUUsageUserContainer  = "cpu.usage.user.container"
+	MetricCPUUsageSysContainer   = "cpu.usage.sys.container"
+	MetricCPUUsageRatioContainer = "cpu.usage.ratio.container"
 
 	MetricCPUShareContainer         = "cpu.share.container"
 	MetricCPUQuotaContainer         = "cpu.quota.container"

diff --git a/pkg/metaserver/agent/metric/provisioner/malachite/provisioner.go b/pkg/metaserver/agent/metric/provisioner/malachite/provisioner.go
@@ -234,6 +234,10 @@ func (m *MalachiteMetricsProvisioner) processSystemComputeData(systemComputeData
 	// todo, currently we only get a unified data for the whole system compute data
 	updateTime := time.Unix(systemComputeData.UpdateTime, 0)
 
+	globalCPU := systemComputeData.GlobalCPU
+	m.metricStore.SetNodeMetric(consts.MetricCPUUsageRatioSystem,
+		utilmetric.MetricData{Value: globalCPU.CPUUsage / 100, Time: &updateTime})
+
 	load := systemComputeData.Load
 	m.metricStore.SetNodeMetric(consts.MetricLoad1MinSystem,
 		utilmetric.MetricData{Value: load.One, Time: &updateTime})
@@ -742,6 +746,8 @@ func (m *MalachiteMetricsProvisioner) processContainerCPUData(podUID, containerN
 			utilmetric.MetricData{Value: cpu.CPUUserUsageRatio, Time: &updateTime})
 		m.metricStore.SetContainerMetric(podUID, containerName, consts.MetricCPUUsageSysContainer,
 			utilmetric.MetricData{Value: cpu.CPUSysUsageRatio, Time: &updateTime})
+		m.metricStore.SetContainerMetric(podUID, containerName, consts.MetricCPUUsageRatioContainer,
+			utilmetric.MetricData{Value: cpu.CPUUsageRatio / (float64(cpu.CfsQuotaUs) / float64(cpu.CfsPeriodUs)), Time: &updateTime})
 		m.metricStore.SetContainerMetric(podUID, containerName, consts.MetricCPUShareContainer,
 			utilmetric.MetricData{Value: float64(cpu.CPUShares), Time: &updateTime})
 		m.metricStore.SetContainerMetric(podUID, containerName, consts.MetricCPUQuotaContainer,
@@ -811,6 +817,8 @@ func (m *MalachiteMetricsProvisioner) processContainerCPUData(podUID, containerN
 			utilmetric.MetricData{Value: cpu.CPUUserUsageRatio, Time: &updateTime})
 		m.metricStore.SetContainerMetric(podUID, containerName, consts.MetricCPUUsageSysContainer,
 			utilmetric.MetricData{Value: cpu.CPUSysUsageRatio, Time: &updateTime})
+		m.metricStore.SetContainerMetric(podUID, containerName, consts.MetricCPUUsageRatioContainer,
+			utilmetric.MetricData{Value: cpu.CPUUsageRatio / (float64(cpu.Max) / float64(cpu.MaxPeriod)), Time: &updateTime})
 
 		m.metricStore.SetContainerMetric(podUID, containerName, consts.MetricCPUNrRunnableContainer,
 			utilmetric.MetricData{Value: float64(cpu.TaskNrRunning), Time: &updateTime})