Volcano v1.16 Released: NamespaceQueue, fairshare, and a broader HyperNode ecosystem

Volcano v1.16.0 ships Namespace Queue and the fairshare plugin, a broader HyperNode ecosystem, and scheduler performance work. Tenants can build hierarchical queues inside their own namespaces within administrator-authorized resource pools, and fairshare folds historical resource consumption into scheduling decisions among namespaces sharing a queue.

Main capabilities

• Pod Group topology anti-affinity: required and preferred rules spread inference instances across HyperNode domains to reduce exposure to a shared failure domain
• Load estimation for bulk Pod dispatch, DRA scoring, Kubernetes 1.36 support, and accelerator and stability improvements

Technical specifications

• A Job is cached only when every rejecting plugin provides a safe hint; otherwise it continues through normal per-session evaluation
• Cached Jobs remain visible to pending demand, queue accounting, and gang requirements, and normal checks run again before placement after a wakeup

Limitations

• Hierarchical capacity rejections and rejections from plugins without hint providers are still evaluated every session
• Different namespaces can use the same queue names without sharing queue state (source text truncated at this point)

Key changes

• Namespace Queue and the fairshare plugin
• Pod Group topology anti-affinity rules
• Kubernetes 1.36 support, load estimation for bulk Pod dispatch, and DRA scoring

Blogs · CNCF Blog · Publisher website: cncf.io · Last verified 31 October 2026 · Catalog ID 127994