Learn how to scope Amazon ECR image pull permissions to individual Kubernetes pods on a multi-tenant Amazon EKS cluster using KEP 4412 credential providers and ECR repository deny policies, so teams sharing the same nodes can pull only their own container images.

Discover Amazon EKS Auto Mode: Simplifies Kubernetes with automation, reducing management duties for efficient deployments

Amazon EMR on EKS now supports interactive Apache Spark sessions with Spark Connect. Data engineers and data scientists can develop and debug Apache Spark applications interactively from managed notebooks in Amazon SageMaker Unified Studio and their own IDEs, such as Jupyter and Visual Studio Code, with Spark running on the Amazon EKS clusters they already operate.   An interactive session provides a persistent Spark context that spans across cells and scripts, letting you blend local Python code execution with remote Spark operations. Spark Connect's client-server architecture decouples your application client from the Spark driver and allows you to maintain your preferred development environment and tooling while Spark runs on your Amazon EKS cluster. This architecture supports workflows including ad hoc data exploration and incremental PySpark job development before deploying to production. Each session runs as pods on a virtual cluster, secured with your AWS Identity and Access Management (IAM) execution role and tagged by project and user. Spark Connect on Amazon EMR on EKS is available with EMR release 7.14 (Apache Spark 3.5) and emr-spark-8.1.0 (Apache Spark 4.1), in all AWS Commercial Regions. The Amazon SageMaker Unified Studio experience is available in https://docs.aws.amazon.com/sagemaker-unified-studio/latest/adminguide/supported-regions.html. To get started, visit the https://docs.aws.amazon.com/emr/latest/EMR-on-EKS-DevelopmentGuide/emr-eks-spark-connect.html or the https://docs.aws.amazon.com/sagemaker-unified-studio/latest/userguide/notebooks-spark-connect.html#spark-connect-emr-eks

Amazon SageMaker HyperPod Inference Gateway is a Kubernetes-native, GPU-aware routing system that deploys as a single EKS managed add-on on existing SageMaker HyperPod infrastructure with zero application changes. By replacing unintelligent round-robin load balancing with real-time inference-signal-driven routing, it reduces first-token latency by up to 82% and p99 TTFT reductions of 97–98% in mixed-hardware and burst traffic scenarios. The Gateway is built around 3 core components. The Envoy Endpoint terminates HTTPS traffic and exposes a single private endpoint per cluster. The Body-Based Router reads the model name directly from each incoming request and routes it to the correct GPU pool - enabling one gateway to serve many models from a single endpoint URL with no client-side changes required. The Endpoint Picker continuously scores every model server pod in real time across 6 inference-level signals - KV cache utilization, queue depth, LoRA adapter residency, prefix cache hit rate, predicted latency and running requests - selecting the optimal pod for each individual request. The gateway works with any OpenAI-compatible model server, including vLLM and SGLang, requiring no application code changes. Per-cluster routing is available today in all AWS Regions where the SageMaker HyperPod inference add-on is supported. Coming soon - cross-cluster and cross-region routing with a centralized fleet gateway, global rate limiting, and cost-tier-aware traffic shaping. To learn more, read the https://aws.amazon.com/blogs/machine-learning/introducing-amazon-sagemaker-hyperpod-inference-gateway/ and explore the https://docs.aws.amazon.com/sagemaker/latest/dg/sagemaker-hyperpod-model-deployment-inference-gateway.html

Amazon EMR on EKS now supports interactive Apache Spark sessions with Spark Connect. Data engineers and data scientists can develop and debug Apache Spark applications interactively from managed notebooks in Amazon SageMaker Unified Studio and their own IDEs, such as Jupyter and Visual Studio Code, with Spark running on the Amazon EKS clusters they already operate.   An interactive session provides a persistent Spark context that spans across cells and scripts, letting you blend local Python code execution with remote Spark operations. Spark Connect's client-server architecture decouples your application client from the Spark driver and allows you to maintain your preferred development environment and tooling while Spark runs on your Amazon EKS cluster. This architecture supports workflows including ad hoc data exploration and incremental PySpark job development before deploying to production. Each session runs as pods on a virtual cluster, secured with your AWS Identity and Access Management (IAM) execution role and tagged by project and user. Spark Connect on Amazon EMR on EKS is available with EMR release 7.14 (Apache Spark 3.5) and emr-spark-8.1.0 (Apache Spark 4.1), in all AWS Commercial Regions. The Amazon SageMaker Unified Studio experience is available in supported AWS Regions. To get started, visit the Spark Connect on Amazon EMR on EKS documentation or the Amazon SageMaker Unified Studio Getting Started guide.

Amazon SageMaker HyperPod Inference Gateway is a Kubernetes-native, GPU-aware routing system that deploys as a single EKS managed add-on on existing SageMaker HyperPod infrastructure with zero application changes. By replacing unintelligent round-robin load balancing with real-time inference-signal-driven routing, it reduces first-token latency by up to 82% and p99 TTFT reductions of 97–98% in mixed-hardware and burst traffic scenarios. The Gateway is built around 3 core components. The Envoy Endpoint terminates HTTPS traffic and exposes a single private endpoint per cluster. The Body-Based Router reads the model name directly from each incoming request and routes it to the correct GPU pool - enabling one gateway to serve many models from a single endpoint URL with no client-side changes required. The Endpoint Picker continuously scores every model server pod in real time across 6 inference-level signals - KV cache utilization, queue depth, LoRA adapter residency, prefix cache hit rate, predicted latency and running requests - selecting the optimal pod for each individual request. The gateway works with any OpenAI-compatible model server, including vLLM and SGLang, requiring no application code changes. Per-cluster routing is available today in all AWS Regions where the SageMaker HyperPod inference add-on is supported. Coming soon - cross-cluster and cross-region routing with a centralized fleet gateway, global rate limiting, and cost-tier-aware traffic shaping. To learn more, read the launch blog and explore the documentation

YouTube video by bwer432

AWS CodePipeline now natively deploys to Amazon EKS, including private cluster endpoints

The Amazon SageMaker AI Spaces add-on for Amazon EKS runs managed JupyterLab and Code Editor environments on the cluster your ML team already operates. This post shows how to install and configure the add-on, connect from the browser and from VS Code over SSH-over-SSM, and move your team to OpenID Connect sign-in with Amazon Cognito.

Amazon EKS enhances visibility into the Kubernetes control plane by offering new intuitive dashboards in EKS console and providing a broader set of Kubernetes control plane metrics. This enables cluster administrators to quickly detect, troubleshoot, and remediate issues. All EKS clusters on Kubernetes version 1.28 and above will now automatically display a curated set of dashboards visualizing key control plane metrics within the EKS console, making it easy to observe the health and performance of the control plane. Additionally, a broader set of control plane metrics are made available in Amazon CloudWatch and in a Prometheus endpoint, providing customers with the flexibility to utilize their preferred monitoring solution — be it Amazon CloudWatch, Amazon Managed Service for Prometheus, or third-party monitoring tools. Newly introduced pre-configured dashboards in the EKS console provide cluster administrators with visual representations of key control plane metrics, enabling rapid assessment of control plane health and performance. Additionally, the EKS console dashboards now integrate with Amazon CloudWatch Log Insights queries, surfacing critical insights from control plane logs directly within the console. Finally, customers now get access to Kubernetes control plane metrics from kube-scheduler and kube-controller-manager, in addition to the existing API server metrics. The new set of dashboards and metrics are available at no additional charge in all AWS commercial regions and AWS GovCloud (US) Regions. To learn more, visit the launch blog post or EKS user guide.

AWS Backup now supports Amazon EKS, providing a fully managed, centralized solution to back up and restore Kubernetes clusters and application data without requiring custom scripts or third-party tools.

Starting today, Split Cost Allocation Data now adds support for accelerated-computing workloads running in the Amazon Elastic Kubernetes Service (EKS). The new feature in Split Cost Allocation Data for EKS allows customers to track the costs associated with accelerator-powered (https://aws.amazon.com/ai/machine-learning/trainium/, https://aws.amazon.com/ai/machine-learning/inferentia/, NVIDIA and AMD GPUs) container-level resources within their EKS clusters, in addition to the costs for CPU and Memory. This cost data is available in the AWS Cost and Usage Report, including CUR 2.0. With this new capability, customers get greater visibility over their AI/ML cloud infrastructure expenses. Customers can now allocate application costs to individual business units and teams based on the CPU, memory and accelerator resource reservations of their containerized accelerated-computing workloads. New Split Cost Allocation Data customers can enable this feature in the AWS Billing and Cost Management console. This feature is automatically enabled for existing Split Cost Allocation Data customers. You can use the https://docs.aws.amazon.com/guidance/latest/cloud-intelligence-dashboards/scad-containers-dashboard.html to visualize the costs in Amazon QuickSight and the https://catalog.workshops.aws/cur-query-library/en-US/queries/container to query the costs using Amazon Athena. This feature is available in all AWS Regions where Split Cost Allocation Data for Amazon EKS is available. To get started, visit https://docs.aws.amazon.com/cur/latest/userguide/split-cost-allocation-data.html andhttps://aws.amazon.com/blogs/aws-cloud-financial-management/improve-cost-visibility-of-machine-learning-workloads-on-amazon-eks-with-aws-split-cost-allocation-data/