The ability for CSI drivers to update attachable volume counts on nodes has moved from Alpha to Beta in Kubernetes v1.34. This feature aims to improve the accuracy of stateful pod scheduling. Previously, CSI drivers reported static volume limits, which could become outdated. Reasons for outdated information include external volume operations, hardware consumption of slots, and multi-driver interactions. This inaccuracy could lead to pods failing to start and getting stuck in a ContainerCreating state. The new functionality allows CSI drivers to dynamically report node attachment capacities at runtime. Kubernetes now supports two update mechanisms: periodic refreshes and immediate updates upon attachment failures. To enable this beta feature, the MutableCSINodeAllocatableCount feature gate must be activated on kube-apiserver and kubelet. CSI drivers can be configured with a nodeAllocatableUpdatePeriodSeconds for periodic updates, with a minimum interval enforced. Immediate updates are triggered by ResourceExhausted errors during volume attachments, preventing pods from permanent failures. Users are encouraged to enable and test this feature in v1.34 and provide feedback for its progression to General Availability.