NVIDIA/ais-k8s v2.7.0
NVIDIA/ais-k8s
Captured source
source ↗published Oct 20, 2025seen Jun 26captured Aug 31http 200method plain
v2.7.0
Repository: NVIDIA/ais-k8s
Tag: v2.7.0
Published: 2025-10-20T21:00:29Z
Prerelease: no
Release notes:
Added
- Auto-scaling mode: Set
size: -1to automatically scale proxy/target pods based on node selectors and tolerations - Auto-scale status tracking: New
AutoScaleStatusfield in cluster status tracks expected nodes for autoScaling clusters - Host path mounting: Use
useHostPath: truein mount spec to bypass PV/PVC provisioning for direct host storage - Token exchange authentication support to allow operators to exchange tokens (e.g., Kubernetes service account tokens or OIDC tokens) with authentication services for AIS JWT tokens
Changed
- Size validation: Allow
size: -1for autoScaling mode - Update target StatefulSet update strategy changed from RollingUpdate to OnDelete.
- Set maintenance mode before pod deletion during target rollouts.
- Reverse target rollout order to start from lowest ordinal (0 to N-1).