In a traditional Kubernetes cluster, pods share the same node and therefore have the same level of access to the host system. This can lead to potential security risks, particularly if a malicious actor gains access to a vulnerable pod. Pod sandboxing in AKS addresses this issue by creating a dedicated container for each pod, which is isolated from other pods and the host system.
AKS pod sandboxing achieves this isolation by running each pod in its own container, using the gVisor sandboxing technology. gVisor is an open-source sandboxing solution that provides a lightweight, container-based isolation mechanism for running untrusted workloads. This approach enables AKS to provide a secure runtime environment for each pod, without sacrificing performance or scalability.
AKS pod sandboxing also provides a number of other security features, including encryption at rest for pod volumes, network isolation through virtual networks, and integrated identity and access management through Azure Active Directory. These features help to protect sensitive data and prevent unauthorized access to Kubernetes resources.
To use AKS pod sandboxing, users can simply enable the feature when creating a new AKS cluster. Once enabled, all pods deployed to the cluster will be automatically sandboxed, providing an added layer of security and isolation.
In summary, pod sandboxing is an important technique for securing Kubernetes workloads, particularly in multi-tenant environments. AKS pod sandboxing provides a powerful and easy-to-use solution for isolating pods from one another and from the host system, using the lightweight gVisor sandboxing technology. By enabling AKS pod sandboxing, users can improve the security and reliability of their Kubernetes deployments, while maintaining performance and scalability.
How it works:
To achieve this functionality on AKS, Kata Containers running on Mariner AKS Container Host (MACH) stack delivers hardware-enforced isolation. Pod Sandboxing extends the benefits of hardware isolation such as a separate kernel for each Kata pod. Hardware isolation allocates resources for each pod and doesn’t share them with other Kata Containers or namespace containers running on the same host.
The solution architecture is based on the following components:
- Mariner AKS Container Host
- Microsoft Hyper-V Hypervisor
- Azure-tuned Dom0 Linux Kernel
- Open-source Cloud-Hypervisor Virtual Machine Monitor (VMM)
- Integration with Kata Container framework
To use this feature with a pod, the only difference is to add runtimeClassName kata-mshv-vm-isolation to the pod spec.
- Kata containers may not reach the IOPS performance limits that traditional containers can reach on Azure Files and high performance local SSD.
- Microsoft Defender for Containers doesn’t support assessing Kata runtime pods.
- Container Insights doesn’t support monitoring of Kata runtime pods in the preview release.
- Kata host-network isn’t supported.
- AKS does not support Container Storage Interface drivers and Secrets Store CSI driver in this preview release.
Enable Pod Sandboxing to an existing AKS cluster:
az aks nodepool add --cluster-name myAKSCluster --resource-group myResourceGroup --name nodepool2 --os-sku mariner --workload-runtime KataMshvVmIsolation --node-vm-size Standard_D4s_v3 az aks update --name myAKSCluster --resource-group myResourceGroup