NVIDIA/ais-k8s v3.1.0

v3.1.0

Repository: NVIDIA/ais-k8s

Tag: v3.1.0

Published: 2026-06-22T22:48:00Z

Prerelease: no

Release notes:

Added

• Added maxUnavailable to proxySpec.autoScale and targetSpec.autoScale to prevent autoscale scale-down while a cluster has unavailable pods within the configured budget.
• AIStoreAuth custom resource (auth.ais.nvidia.com/v1alpha1), served by the existing operator. This addition also introduces API types, a validating webhook, and a no-op reconciler for AIStoreAuth as the foundation for managing AuthN deployments via the operator.
• Added spec.proxySpec.externalAccess and spec.targetSpec.externalAccess for additional control of external services to expose pod access; currently limited to LoadBalancers.

Changed

• Deprecated enableExternalLB. Use spec.proxySpec.externalAccess and spec.targetSpec.externalAccess for equivalent and extended config.