Merge pull request #859 from forta-network/kisel/improve-docker-cpu-m…

…etric Calculate docker cpu percentage
forta-network · Apr 12, 2024 · c34c465 · c34c465
2 parents df3194e + 05bcb9a
commit c34c465
Show file tree

Hide file tree

Showing 4 changed files with 28 additions and 3 deletions.
diff --git a/clients/docker/client.go b/clients/docker/client.go
@@ -901,6 +901,8 @@ func (d *dockerClient) ContainerStats(ctx context.Context, id string) (*Containe
 			CPUUsage: CPUUsage{
 				TotalUsage: containerResources.CPUStats.CPUUsage.TotalUsage,
 			},
+			SystemUsage: containerResources.CPUStats.SystemUsage,
+			OnlineCPUs:  containerResources.CPUStats.OnlineCPUs,
 		},
 		MemoryStats: MemoryStats{
 			Usage: containerResources.MemoryStats.Usage,

diff --git a/clients/docker/types.go b/clients/docker/types.go
@@ -7,7 +7,9 @@ type ContainerResources struct {
 }
 
 type CPUStats struct {
-	CPUUsage CPUUsage `json:"cpu_usage"`
+	CPUUsage    CPUUsage `json:"cpu_usage"`
+	SystemUsage uint64   `json:"system_cpu_usage"`
+	OnlineCPUs  uint32   `json:"online_cpus"`
 }
 
 type CPUUsage struct {

diff --git a/services/components/containers/bot_client.go b/services/components/containers/bot_client.go
@@ -130,6 +130,8 @@ func (bc *botClient) pollDockerResources(containerID string, agentConfig config.
 	ticker := initTicker(DockerResourcesPollingInterval)
 	defer ticker.Stop()
 
+	var previousResources docker.ContainerResources
+
 	for t := range ticker.C {
 		logrus.WithField("containerID", containerID).Debug("polling docker resources")
 		// request docker stats
@@ -159,17 +161,36 @@ func (bc *botClient) pollDockerResources(containerID string, agentConfig config.
 			WithField("resources", resources).
 			Debug("sending docker resources metrics")
 
+		cpuPercent := calculateCPUPercentUnix(&previousResources, resources)
+
 		metrics.SendAgentMetrics(bc.msgClient, []*protocol.AgentMetric{
 			metrics.CreateAgentResourcesMetric(
-				agentConfig, t, domain.MetricDockerResourcesCPU, float64(resources.CPUStats.CPUUsage.TotalUsage)),
+				agentConfig, t, domain.MetricDockerResourcesCPU, cpuPercent),
 			metrics.CreateAgentResourcesMetric(
 				agentConfig, t, domain.MetricDockerResourcesMemory, float64(resources.MemoryStats.Usage)),
 			metrics.CreateAgentResourcesMetric(
 				agentConfig, t, domain.MetricDockerResourcesNetworkSent, float64(bytesSent)),
 			metrics.CreateAgentResourcesMetric(
 				agentConfig, t, domain.MetricDockerResourcesNetworkReceive, float64(bytesRecv)),
 		})
+
+		previousResources = *resources
+	}
+}
+
+func calculateCPUPercentUnix(previousResources, resources *docker.ContainerResources) float64 {
+	var (
+		cpuPercent = 0.0
+		// calculate the change for the cpu usage of the container in between readings
+		cpuDelta = float64(resources.CPUStats.CPUUsage.TotalUsage) - float64(previousResources.CPUStats.CPUUsage.TotalUsage)
+		// calculate the change for the entire system between readings
+		systemDelta = float64(resources.CPUStats.SystemUsage) - float64(previousResources.CPUStats.SystemUsage)
+	)
+
+	if systemDelta > 0.0 && cpuDelta > 0.0 {
+		cpuPercent = (cpuDelta / systemDelta) * float64(resources.CPUStats.OnlineCPUs) * 100.0
 	}
+	return cpuPercent
 }
 
 func (bc *botClient) attachServiceContainers(ctx context.Context, botNetworkID string) error {

diff --git a/services/components/containers/bot_client_test.go b/services/components/containers/bot_client_test.go
@@ -129,7 +129,7 @@ func (s *BotClientTestSuite) TestLaunchBot_Exists() {
 		// CPU metric
 		assert.Equal(s.T(), botConfig.ID, metrics.Metrics[0].AgentId)
 		assert.Equal(s.T(), domain.MetricDockerResourcesCPU, metrics.Metrics[0].Name)
-		assert.Equal(s.T(), float64(33), metrics.Metrics[0].Value)
+		assert.Equal(s.T(), float64(0), metrics.Metrics[0].Value)
 
 		// Memory metric
 		assert.Equal(s.T(), botConfig.ID, metrics.Metrics[1].AgentId)