May 2026 monthly summary for kata-containers/kata-containers focused on feature delivery and automation-readiness. Implemented RuntimeClass Version Annotations to enable automated version determination and cross-tool compatibility, with RBAC-based version reads. This enables integrations with automation pipelines (e.g., genpolicy) and reduces manual verification, contributing to safer upgrades and consistent runtimes.

February 2026: Delivered GPU runtime enhancements and NVIDIA libraries improvements, extended the NVIDIA GPU Operator with Kata device plugin support, integrated the CC manager into the OLM bundle, and expanded GPU state management test coverage. These efforts improved scheduling consistency across CC and non-CC nodes, enabled sandbox workloads via Kata containers, and increased reliability of device plugin configurations across environments.

January 2026: Delivered Confidential Computing (CC) integration for NVIDIA/gpu-operator, enabling default CC management on kata nodes, with configurable daemonset YAMLs and removal of selective NodeFeatureRules to streamline deployment. Fixed a bug by eliminating hardcoded command names in the CC deployment workflow, improving portability and maintainability. These changes reduce deployment complexity, strengthen security posture for confidential workloads, and accelerate enterprise adoption of CC in GPU workloads.

2025-11 monthly summary for DataDog/kata-containers focused on delivering device cold plug support via the Kubernetes Pod Resources API integration. Implemented changes to enable cold plugging during pod creation, ensured Kubernetes compatibility through vendor updates, and introduced configuration adjustments to activate cold plug in runtime. The work improves resource allocation flexibility and aligns container runtime behavior with Pod Resources API expectations.