Network Automation, Monitoring, and Telemetry
Free NVIDIA-Certified Professional: AI Networking practice — 6 questions on Network Automation, Monitoring, and Telemetry, with explanations. No sign-up. Full 12-question mixed test →
During a large-scale NCCL all-reduce job, RoCEv2 traffic on a specific Spectrum switch port experiences intermittent, sub-millisecond packet drops. The team must identify the exact hardware-level drop reason (e.g., buffer exhaustion vs. ECN marking vs. CRC error) in real time, without adding capture latency or consuming additional switch CPU for deep packet inspection. Which feature should be used?
WJH is implemented directly in the Spectrum ASIC and classifies drop reasons (buffer, ACL, CRC, tail drop, ECN, symbol errors) at line rate, streaming the results without impacting switch performance — exactly the real-time, low-overhead visibility the scenario requires.
An engineer runs `nv set interface swp1 link mtu 9216` on a Cumulus Linux switch managed with NVUE, then runs `nv config apply` and confirms the interface is using the new MTU. After a scheduled maintenance reboot, the switch reverts swp1 to the default MTU. Which additional command was missing from the workflow?
`nv config apply` only activates staged changes in the running configuration. `nv config save` is required to write the applied configuration into the startup configuration so it persists across a reboot.
A network automation team wants to manage NVUE-based configuration on 200 Cumulus Linux switches using Ansible playbooks, interacting directly with NVUE's declarative REST API (rather than scraping CLI text output) so that changes are idempotent and structured as JSON. Which Ansible collection should they use?
The nvidia.nvue collection provides modules that interact with the NVUE REST API directly, allowing idempotent, structured management of Cumulus Linux configuration state without CLI screen-scraping.
A capacity planning team needs six months of historical port counters, PFC pause ratios, and congestion statistics from an InfiniBand fabric, exported continuously into Grafana for long-term trend analysis. Which UFM capability should be configured?
UFM Telemetry supports streaming counters and fabric health metrics via Fluentd/Prometheus-compatible exporters into an external time-series database, which is the intended mechanism for long-term, Grafana-based historical trend analysis.
A security team wants to offload stateful connection tracking and hardware-accelerated flow rule matching (for a zero-trust firewall policy) onto a BlueField DPU, freeing host CPU cycles entirely from packet processing. Which DOCA library is designed for this purpose?
DOCA Flow provides a hardware-accelerated flow processing pipeline API used to program connection tracking, NAT, and firewall/ACL rules directly onto the BlueField DPU's flow tables, offloading this work from the host CPU.
An operations team wants sub-second, push-based delivery of interface counters and queue-depth metrics from Cumulus Linux switches to a central collector, subscribing only to specific YANG data paths of interest, instead of periodically polling the entire MIB tree. Which telemetry approach satisfies this requirement?
gNMI (gRPC Network Management Interface) with OpenConfig YANG models allows a collector to subscribe to specific paths and receive push-based, sub-second streaming updates, avoiding the overhead and latency of periodic full-tree polling.
Ready for the real thing?
The full course has two full-length practice tests, video lessons for every exam domain, hands-on labs and detailed answer explanations.
Start my full course on Udemy →