In Azure Kubernetes Service (AKS), you can choose between two networking models: kubenet and Azure CNI. Which statement correctly describes how pod IP addresses are allocated in these two models?
Think about whether pods get IPs from the Azure network or from a separate range.
Kubenet assigns pod IPs from a private address space managed by AKS, using network address translation (NAT) to communicate outside. Azure CNI assigns pod IPs directly from the Azure virtual network subnet, making pods first-class citizens on the network.
You plan to deploy an AKS cluster with a very high number of pods per node. Which networking model is better suited for this scenario and why?
Consider how IP address usage affects pod density in each model.
Kubenet uses NAT, so pods share node IPs externally and do not require an IP per pod from the Azure subnet, allowing more pods per node. Azure CNI assigns an IP per pod from the subnet, which can limit pod density due to IP exhaustion.
Which AKS networking model supports Azure Network Policies natively for controlling pod traffic?
Think about which model integrates better with Azure's native network security features.
Azure CNI integrates with Azure Network Policies, allowing native enforcement of traffic rules. Kubenet does not support Azure Network Policies directly and requires other solutions like Calico for network policy enforcement.
In an AKS cluster using kubenet networking, how is pod-to-pod communication across different nodes handled?
Consider how kubenet uses NAT and routing for pod traffic.
Kubenet assigns pod IPs from a private range and uses NAT on nodes to translate pod IPs to node IPs for cross-node communication. Routing rules on nodes ensure traffic reaches the correct pod.
You have an AKS cluster using Azure CNI networking. You want to scale the cluster to 100 nodes, each running 30 pods. What is the best practice to avoid IP address exhaustion in the Azure virtual network subnet?
Think about how Azure CNI assigns IPs and what subnet size is needed.
Azure CNI assigns an IP per pod from the Azure subnet. To support 100 nodes with 30 pods each plus node IPs, the subnet must be large enough to provide at least 3000 IP addresses. Increasing subnet size is the best practice.