What problem does it solve? Upgrading the NVIDIA Network Operator in a Kubernetes cluster involves coordinated steps—cordoning nodes, regenerating configs, applying CRDs, and running helm upgrade—where mistakes can disrupt RDMA networking for AI/ML workloads. This Skill guides the full upgrade workflow safely, including version compatibility checks and rollback. ## Core Features & Use Cases - Guided Upgrade Procedure: Step-by-step workflow covering pre-upgrade checks, dry-run config generation with CREATE_CONFIG_ONLY, and execution of upgrade-network-operator.sh. - Version-Specific Guidance: Documents behavioral differences across operator versions 24.7.0 through 26.1.0, including MAINTENANCE_OPERATOR_ENABLE defaults and API parameter changes. - Rollback and Troubleshooting: Provides rollback instructions and a failure table covering helm chart errors, CRD validation issues, stuck cordons, and finalizer blocks. - Use Case: A cluster administrator needs to move from Network Operator 25.7.0 to 26.1.0 on a DGX cluster; the Skill walks them through previewing config diffs, running the upgrade script, and verifying SR-IOV and IPAM state afterward. ## Quick Start Ask the AI to upgrade the NVIDIA Network Operator to version 26.1.0 and verify the cluster networking state afterward.
