Release notes for NVIDIA Base Command™ Manager (BCM) 10.30.0
Released: 3 November 2025
General
New Features
Added CUDA 13.0 packages
Added support for Slurm 25.05
Added cm-nccl2-cuda13.0-gcc and cm-cudnn9.13-cuda13.0 packages
Updated Slurm 24.11 to 24.11.6
Updated etcd to 3.5.22
Updated cm-nvhpc to 25.5
Updated cuda-driver to 580.82.07
Updated cuda-driver-550 to 550.163.01
Updated cuda-driver-535 to 535.261.03
Updated mlnx-ofed23.10 to 23.10-5.1.4.0
Updated mlnx-ofed24.10 to 24.10-3.2.5.0
Updated mlnx-ofed58 to 5.8-7.0.6.1
Fixed Issues
An issue with the enroot cache path generated by bcm-post-install
An issue with sorting in the internal BracketParser::collapseSyntax function, which can result in duplicated items in the slurm.conf file
CMDaemon
New Features
Allow the option to disable writing the root kubeconfig configuration files on Kubernetes worker nodes by setting a GlobalConfig option
Invoke kubeadm kubeconfig user commands only when actually needed
Fixed Issues
An issue with enum metrics value translations when upper and lower cases are mixed for the values
A regression in KubernetesChildNode health check
An issue with the metrics collection scripts in Kubernetes
An issue with duplicated entries in merged Kubernetes configuration files
cluster-tools
New Features
Added DOCA OFED installation helper script
cm-kubernetes-setup
New Features
Added Run:ai self-hosted operator support
Added cm-kubeadm-manage-joins helper script
Updated run:ai to 2.22
Updated Kubeflow training operator to 1.9.2
Fixed Issues
An issue where it was possible to uninstall CAPI templates (stored in the kubernetes submode) by using cm-kubernetes-setup
An issue where long hostnames could visually render incorrect TUI dialogs in the Run:ai configuration dialog
cm-scale
Fixed Issues
An issue with powering off the nodes after a shut down has timed out
cmsh
Fixed Issues
An issue where the flag -m is shown as the short version for both –message and –image in some commands
An issue in cmsh that can result in a crash when a fsexport is cloned