편집

다음을 통해 공유

What is container network metrics?

Advanced Container Networking Services in Azure Kubernetes Service (AKS) facilitates the collection of comprehensive container network metrics to give you valuable insights into the performance of your containerized environment. The capability continuously captures essential metrics at the node level and pod level, including traffic volume, dropped packets, connection states, and Domain Name System (DNS) resolution times for effective monitoring and optimizing network performance.

Capturing these metrics is essential for understanding how containers communicate, how traffic flows between services, and where bottlenecks or disruptions might occur. Advanced Container Networking Services integrates seamlessly with monitoring tools like Prometheus and Grafana to give you a complete view of networking metrics. Use the metrics for in-depth troubleshooting, network optimization, and performance tuning.

In a cloud-native world, maintaining a healthy and efficient network in a dynamic containerized environment is vital to ensuring that applications perform as expected. Without proper visibility into network traffic and its patterns, identifying potential issues or inefficiencies becomes challenging.

Key benefits

  • Deep visibility into network performance

  • Enhanced troubleshooting and optimization

  • Proactive anomaly detection

  • Better resource management and scaling

  • Capacity planning and compliance

  • Simplified metrics storage and visualization options. Choose between:

    • Azure managed service for Prometheus and Azure Managed Grafana: Azure manages the infrastructure and maintenance, so you can focus on configuring metrics and visualizing metrics.
    • Bring your own (BYO) Prometheus and Grafana: You deploy and configure your own instances of Prometheus and Grafana, and you manage the underlying infrastructure.

Metrics captured

Node-level metrics

Understanding the health of your container network at the node-level is crucial for maintaining optimal application performance. These metrics provide insights into traffic volume, dropped packets, number of connections, and other data by node. The metrics are stored in Prometheus format, so, you can view them in Grafana.

The following metrics are aggregated per node. All metrics include one of these labels:

  • cluster
  • instance (node name)

For Cilium data plane scenarios, Container Network Observability provides metrics only for Linux. Windows is currently not supported. Cilium exposes several metrics including the following used by Container Network Observability.

Metric name Description Extra labels Linux Windows
cilium_forward_count_total Total forwarded packet count direction ✅ ❌
cilium_forward_bytes_total Total forwarded byte count direction ✅ ❌
cilium_drop_count_total Total dropped packet count direction, reason ✅ ❌
cilium_drop_bytes_total Total dropped byte count direction, reason ✅ ❌

Pod-level metrics (Hubble metrics)

These Prometheus metrics include source and destination pod information so that you can pinpoint network-related issues at a granular level. Metrics cover information like traffic volume, dropped packets, TCP resets, and Layer 4/Layer 7 packet flows. DNS metrics also are collected (currently only for non-Cilium data planes), including DNS errors and DNS requests missing responses.

The following table describes the metrics that are aggregated per pod (node information is preserved).

All metrics include labels:

  • cluster

  • instance (node name)

  • source or destination

    • For outgoing traffic, a source label that indicates the source pod namespace and name is applied.

    • For incoming traffic, a destination label that indicates the destination pod namespace and name is applied.

Metric name Description Extra Labels Linux Windows
hubble_dns_queries_total Total DNS requests by query source or destination, query, qtypes (query type) ✅ ❌
hubble_dns_responses_total Total DNS responses by query/response source or destination, query, qtypes (query type), rcode (return code), ips_returned (number of IPs) ✅ ❌
hubble_drop_total Total dropped packet count source or destination, protocol, reason ✅ ❌
hubble_tcp_flags_total Total TCP packets count by flag source or destination, flag ✅ ❌
hubble_flows_processed_total Total network flows processed (Layer 4/Layer 7 traffic) source or destination, protocol, verdict, type, subtype ✅ ❌

Limitations

  • Pod-level metrics are available only on Linux.
  • The Cilium data plane is supported in Kubernetes version 1.29 and later.
  • Metric labels have subtle differences between Cilium and non-Cilium clusters.
  • For Cilium-based clusters, DNS metrics are available only for pods that have Cilium Network Policies (CNPs) configured on their clusters.
  • Flow logs currently aren't available in an air-gapped cloud.
  • If a Hubble node agent fails, Hubble Relay might crash. It might cause interruptions in using the Hubble CLI.

Scale

The managed service for Prometheus in Azure Monitor and Azure Managed Grafana impose service-specific scale limitations. For more information, see Scrape Prometheus metrics at scale in Azure Monitor.

Pricing

Important

Advanced Container Networking Services is a paid offering.

For more information about pricing, see Advanced Container Networking Services - Pricing.

Related content