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

kubewharf · Sep 29, 2024 · b808189 · b808189
1 parent 875d630
commit b808189
Show file tree

Hide file tree

Showing 6 changed files with 24 additions and 10 deletions.
diff --git a/go.mod b/go.mod
@@ -161,6 +161,7 @@ require (
 )
 
 replace (
+	github.com/kubewharf/katalyst-api => github.com/sun-yuliang/katalyst-api v0.0.0-20240913040056-a0a800e37eae
 	k8s.io/api => k8s.io/api v0.24.6
 	k8s.io/apiextensions-apiserver => k8s.io/apiextensions-apiserver v0.24.6
 	k8s.io/apimachinery => k8s.io/apimachinery v0.24.6

diff --git a/go.sum b/go.sum
@@ -568,8 +568,6 @@ 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/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=
@@ -878,6 +876,8 @@ github.com/stretchr/testify v1.8.0/go.mod h1:yNjHg4UonilssWZ8iaSj1OCr/vHnekPRkoO
 github.com/stretchr/testify v1.8.1 h1:w7B6lhMri9wdJUVmEZPGGhZzrYTPvgJArz7wNPgYKsk=
 github.com/stretchr/testify v1.8.1/go.mod h1:w2LPCIKwWwSfY2zedu0+kehJoqGctiVI29o6fzry7u4=
 github.com/subosito/gotenv v1.2.0/go.mod h1:N0PQaV/YGNqwC0u51sEeR/aUtSLEXKX9iv69rRypqCw=
+github.com/sun-yuliang/katalyst-api v0.0.0-20240913040056-a0a800e37eae h1:8CyPnE45qnXgVep/PeBy+pdMmpAEUNB7lrgGMER3VGg=
+github.com/sun-yuliang/katalyst-api v0.0.0-20240913040056-a0a800e37eae/go.mod h1:Y2IeIorxQamF2a3oa0+URztl5QCSty6Jj3zD83R8J9k=
 github.com/syndtr/gocapability v0.0.0-20200815063812-42c35b437635/go.mod h1:hkRG7XYTFWNJGYcbNJQlaLq0fg1yr4J4t/NcTQtrfww=
 github.com/tdakkota/asciicheck v0.0.0-20200416190851-d7f85be797a2/go.mod h1:yHp0ai0Z9gUljN3o0xMhYJnH/IcvkdTBOX2fmJ93JEM=
 github.com/tetafro/godot v0.4.8/go.mod h1:/7NLHhv08H1+8DNj0MElpAACw1ajsCuf3TKNQxA5S+0=

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})