Kubernetes 1.36 expands device resource allocation with seven Beta features
Released May 7, the update brings partitionable GPUs, device health reporting, and taints to production readiness while removing limits on shared resource claims.
Disclaimer
This article was produced by Scroll Insights News Desk using automated systems and published under our standing editorial policy. It is compiled from the primary sources linked above and is provided for general information only — it is not legal, financial, investment, tax or professional advice, and no decision should be taken on it without independent verification against those sources. Errors can be reported to corrections@scrollinsights.com and are corrected on the record.
Kubernetes v1.36, released on May 7, 2026, advances Dynamic Resource Allocation (DRA) toward production use with seven features graduating to Beta status and one reaching Stable.
Device partitioning and reservation
Partitionable devices now support native DRA for hardware that can be carved into smaller instances, such as Multi-Instance GPUs. Device taints and tolerations allow cluster administrators to mark faulty hardware offline or reserve specific devices for particular teams and workloads. Both features reached Beta in this release.
Health visibility and constraint logic
Resource health status exposes device health directly in Pod status using human-readable messages, graduating to Beta. ResourceClaim constraints now use matchAttribute to detect non-empty intersections and distinctAttribute to enforce pairwise disjoint values across claims. Extended resource support lets workloads transition gradually from traditional extended resources to DRA, also reaching Beta.
Removing bottlenecks for shared access
ResourceClaim support for workloads eliminates prior limits on how many pods can share a single claim, addressing a constraint that blocked multi-tenant device use. Device binding conditions and Resource pool status both reached Beta, allowing administrators to query device availability in DRA pools and track binding outcomes.
Supporting hardware discovery
Kubernetes 1.36 defines a standard protocol for DRA drivers to expose device attributes as versioned JSON at well-known paths, enabling containers to discover hardware capabilities. Prioritized list, which controls device selection order, reached Stable status. The scheduler was updated to evaluate devices using lexicographical ordering based on resource pool and ResourceSlice names. A CEL includes() function was added for DRA contexts to support membership testing in device constraints.
Node allocatable resources for CPU and memory entered alpha, and a DRA kubelet plugin library handles file layout, CDI bind-mounts, versioning, and device metadata lifecycle.