add cpu burst metrics

Author: sambhav-jain-16

container/libcontainer/handler.go

@@ -773,6 +773,8 @@ func setCPUStats(s *cgroups.Stats, ret *info.ContainerStats, withPerCPU bool) {
 	ret.Cpu.CFS.Periods = s.CpuStats.ThrottlingData.Periods
 	ret.Cpu.CFS.ThrottledPeriods = s.CpuStats.ThrottlingData.ThrottledPeriods
 	ret.Cpu.CFS.ThrottledTime = s.CpuStats.ThrottlingData.ThrottledTime
+	ret.Cpu.CFS.BurstsPeriods = s.CpuStats.BurstData.BurstsPeriods
+	ret.Cpu.CFS.BurstTime = s.CpuStats.BurstData.BurstTime
 	setPSIStats(s.CpuStats.PSI, &ret.Cpu.PSI)
 
 	if !withPerCPU {

info/v1/container.go

@@ -329,6 +329,13 @@ type CpuCFS struct {
 	// Total time duration for which tasks in the cgroup have been throttled.
 	// Unit: nanoseconds.
 	ThrottledTime uint64 `json:"throttled_time"`
+
+	// Total number of periods when CPU burst occurs.
+	BurstsPeriods uint64 `json:"bursts_periods"`
+
+	// Total time duration when CPU burst occurs.
+	// Unit: nanoseconds.
+	BurstTime uint64 `json:"burst_time"`
 }
 
 // Cpu Aggregated scheduler statistics

metrics/prometheus.go

@@ -234,6 +234,30 @@ func NewPrometheusCollector(i infoProvider, f ContainerLabelsFunc, includedMetri
 							timestamp: s.Timestamp,
 						}}
 				},
+			}, {
+				name:      "container_cpu_cfs_burst_periods_total",
+				help:      "Number of periods when burst occurs.",
+				valueType: prometheus.CounterValue,
+				condition: func(s info.ContainerSpec) bool { return s.Cpu.Quota != 0 },
+				getValues: func(s *info.ContainerStats) metricValues {
+					return metricValues{
+						{
+							value:     float64(s.Cpu.CFS.BurstsPeriods),
+							timestamp: s.Timestamp,
+						}}
+				},
+			}, {
+				name:      "container_cpu_cfs_burst_seconds_total",
+				help:      "Total time duration the container has been bursted.",
+				valueType: prometheus.CounterValue,
+				condition: func(s info.ContainerSpec) bool { return s.Cpu.Quota != 0 },
+				getValues: func(s *info.ContainerStats) metricValues {
+					return metricValues{
+						{
+							value:     float64(s.Cpu.CFS.BurstTime) / float64(time.Second),
+							timestamp: s.Timestamp,
+						}}
+				},
 			},
 		}...)
 	}

metrics/prometheus_fake.go

@@ -320,6 +320,8 @@ func (p testSubcontainersInfoProvider) GetRequestedContainersInfo(string, v2.Req
 							Periods:          723,
 							ThrottledPeriods: 18,
 							ThrottledTime:    1724314000,
+							BurstsPeriods:    25,
+							BurstTime:        500000000,
 						},
 						Schedstat: info.CpuSchedstat{
 							RunTime:      53643567,

metrics/prometheus_test.go

@@ -428,3 +428,39 @@ func TestIOCostMetrics(t *testing.T) {
 		})
 	}
 }
+
+func TestCPUBurstMetrics(t *testing.T) {
+	containerStats := &info.ContainerStats{
+		Timestamp: time.Unix(1395066363, 0),
+		Cpu: info.CpuStats{
+			CFS: info.CpuCFS{
+				BurstsPeriods: 25,
+				BurstTime:     500000000,
+			},
+		},
+	}
+
+	testCases := []struct {
+		name          string
+		getValue      func() float64
+		expectedValue float64
+	}{
+		{
+			name:          "BurstsPeriods",
+			getValue:      func() float64 { return float64(containerStats.Cpu.CFS.BurstsPeriods) },
+			expectedValue: 25.0,
+		},
+		{
+			name:          "BurstTime",
+			getValue:      func() float64 { return float64(containerStats.Cpu.CFS.BurstTime) / float64(time.Second) },
+			expectedValue: 0.5,
+		},
+	}
+
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			result := tc.getValue()
+			assert.Equal(t, tc.expectedValue, result)
+		})
+	}
+}

metrics/testdata/prometheus_metrics

@@ -18,6 +18,12 @@ container_cpu_cfs_throttled_periods_total{container_env_foo_env="prod",container
 # HELP container_cpu_cfs_throttled_seconds_total Total time duration the container has been throttled.
 # TYPE container_cpu_cfs_throttled_seconds_total counter
 container_cpu_cfs_throttled_seconds_total{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",zone_name="hello"} 1.724314 1395066363000
+# HELP container_cpu_cfs_burst_periods_total Number of periods when burst occurs.
+# TYPE container_cpu_cfs_burst_periods_total counter
+container_cpu_cfs_burst_periods_total{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",zone_name="hello"} 25 1395066363000
+# HELP container_cpu_cfs_burst_seconds_total Total time duration the container has been bursted.
+# TYPE container_cpu_cfs_burst_seconds_total counter
+container_cpu_cfs_burst_seconds_total{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",zone_name="hello"} 0.5 1395066363000
 # HELP container_cpu_load_average_10s Value of container cpu load average over the last 10 seconds.
 # TYPE container_cpu_load_average_10s gauge
 container_cpu_load_average_10s{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",zone_name="hello"} 2 1395066363000

metrics/testdata/prometheus_metrics_whitelist_filtered

@@ -18,6 +18,12 @@ container_cpu_cfs_throttled_periods_total{container_env_foo_env="prod",id="testc
 # HELP container_cpu_cfs_throttled_seconds_total Total time duration the container has been throttled.
 # TYPE container_cpu_cfs_throttled_seconds_total counter
 container_cpu_cfs_throttled_seconds_total{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",zone_name="hello"} 1.724314 1395066363000
+# HELP container_cpu_cfs_burst_periods_total Number of periods when burst occurs.
+# TYPE container_cpu_cfs_burst_periods_total counter
+container_cpu_cfs_burst_periods_total{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",zone_name="hello"} 25 1395066363000
+# HELP container_cpu_cfs_burst_seconds_total Total time duration the container has been bursted.
+# TYPE container_cpu_cfs_burst_seconds_total counter
+container_cpu_cfs_burst_seconds_total{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",zone_name="hello"} 0.5 1395066363000
 # HELP container_cpu_load_average_10s Value of container cpu load average over the last 10 seconds.
 # TYPE container_cpu_load_average_10s gauge
 container_cpu_load_average_10s{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",zone_name="hello"} 2 1395066363000