Stage 5 · Platform

Azure Monitor & Observability

AKS Monitoring with Container Insights

Node/pod metrics, KEDA scaling metrics, and multi-cluster views.

7 min readAzure Cloud EngineeringPlatform

Container Insights Overview

Container Insights is a feature of Azure Monitor that collects performance metrics, container logs, and inventory data from AKS clusters. It provides a pre-built monitoring experience with maps, metrics explorer, and analytics queries.

Enable Container Insights


# Enable during cluster creation
az aks create \
  --resource-group rg-aks-prod \
  --name aks-payments-prod \
  --enable-addons monitoring \
  --workspace-resource-id "/subscriptions/{sub-id}/resourceGroups/rg-monitoring/providers/Microsoft.OperationalInsights/workspaces/law-aks" \
  --node-count 3

# Enable on existing cluster
az aks enable-addons \
  --resource-group rg-aks-prod \
  --name aks-payments-prod \
  --addons monitoring \
  --workspace-resource-id "/subscriptions/{sub-id}/resourceGroups/rg-monitoring/providers/Microsoft.OperationalInsights/workspaces/law-aks"

Container Insights installs a DaemonSet (one pod per node) that collects metrics and logs. The data flows to the specified Log Analytics workspace.

Node Metrics

Query node metrics


// Node CPU and memory usage over time
InsightsMetrics
| where TimeGenerated > ago(1h)
| where Namespace == "K8SNode"
| where Name in ("cpuUsagePercentage", "memoryWorkingSetPercentage")
| summarize avg(Val) by Computer, Name, bin(TimeGenerated, 5m)
| render timechart

// Node with highest CPU
InsightsMetrics
| where TimeGenerated > ago(5m)
| where Namespace == "K8SNode"
| where Name == "cpuUsagePercentage"
| top 5 by Val desc
| project Computer, NodeCPU = Val

InsightsMetrics is the table populated by Container Insights. Namespace identifies the metric source (K8SNode, K8SContainer, K8SPod).

Pod and Container Metrics

Pod resource consumption


// Top 10 pods by CPU usage
InsightsMetrics
| where TimeGenerated > ago(5m)
| where Namespace == "K8SContainer"
| where Name == "cpuUsagePercentage"
| top 10 by Val desc
| project Pod = Computer, Container = InstanceName, CPU = Val

// Pods exceeding memory limits
InsightsMetrics
| where TimeGenerated > ago(5m)
| where Namespace == "K8SContainer"
| where Name == "memoryRSS"
| where Val > 0
| join kind=leftouter (
  KubePodInventory
  | project PodName, Namespace, MemoryLimit
) on PodName, Namespace
| where MemoryLimit > 0 and Val > MemoryLimit * 0.9
| project Pod = PodName, Namespace, MemoryMB = Val / 1024 / 1024, LimitMB = MemoryLimit / 1024 / 1024

KubePodInventory provides Kubernetes metadata (labels, status, owner) that you can join with InsightsMetrics for rich pod-level analysis.

Use the Container Insights workbook

Container Insights includes a pre-built Azure Monitor Workbook with node health, pod status, and resource consumption charts. Navigate to your AKS cluster in the Portal and click 'Insights'.

Container Logs

Query container logs


// All logs from a specific pod
ContainerLogV2
| where TimeGenerated > ago(1h)
| where PodName == "payments-api-7d4b8c6f5-2xk9j"
| project TimeGenerated, LogEntry

// Error logs across all pods
ContainerLogV2
| where TimeGenerated > ago(24h)
| where LogEntry contains "ERROR"
| summarize count() by PodName, Namespace
| order by count_ desc

// Logs from a specific container in a pod
ContainerLogV2
| where TimeGenerated > ago(1h)
| where PodName startswith "payments"
| where ContainerName == "payments"
| where LogEntry contains "exception"
| project TimeGenerated, LogEntry

ContainerLogV2 provides structured container logs with PodName, ContainerName, and Namespace fields. Use LogEntry for the actual log message.

KEDA Scaling Metrics

KEDA (Kubernetes Event-Driven Autoscaler) publishes custom metrics for pod autoscaling. These metrics appear in Container Insights and can trigger HPA scaling.

Query KEDA scaling metrics


// View KEDA scaler metrics
InsightsMetrics
| where TimeGenerated > ago(1h)
| where Namespace == "keda"
| where Name == "scaler_active"
| summarize avg(Val) by Tag, bin(TimeGenerated, 1m)
| render timechart

// Current queue length for scaled objects
InsightsMetrics
| where TimeGenerated > ago(5m)
| where Namespace == "keda"
| where Name == "scaler_metrics_value"
| project TimeGenerated, Metric = Tag, Value = Val

KEDA metrics are published to the Kubernetes metrics API. Container Insights captures them automatically when installed.

Multi-Cluster Monitoring

When running multiple AKS clusters (dev, staging, prod, or multi-region), you need a unified monitoring view. Use a single Log Analytics workspace to aggregate data from all clusters.

Multi-cluster queries


// Compare CPU across all clusters
InsightsMetrics
| where TimeGenerated > ago(5m)
| where Namespace == "K8SNode"
| where Name == "cpuUsagePercentage"
| summarize avgCPU = avg(Val) by Computer
| order by avgCPU desc

// Aggregate pod restarts across clusters
ContainerInsights
| where TimeGenerated > ago(24h)
| where PodRestartCount > 0
| summarize maxRestarts = max(PodRestartCount) by ClusterId, Namespace, PodName
| order by maxRestarts desc

The ClusterId field identifies which cluster a metric came from. Use it to filter or group data across your fleet of clusters.

Use Azure Monitor for containers across clusters

Point all AKS clusters to the same Log Analytics workspace. This enables cross-cluster queries, unified alerting, and a single pane of glass for your container fleet.

Mark this lesson complete to store local progress and unlock a cleaner resume path the next time you visit.