System and Server Bring-up

NCP-AII · 62 questions

• A city AI factory racks new HGX nodes for a permitting vision model. Which day-0 order should the rack team follow before any workload starts?
• County public-health IT is writing a checklist for new inference nodes. Which validation gates belong before control-plane install and Slurm jobs?
• A transit authority signs off a SuperPOD-style row and asks where cable and transceiver verification sits relative to BMC setup and GPU install. What is the correct placement?
• An emergency-management compute lead debates when third-party storage initial parameters should be set during AI node bring-up. What ordering is appropriate?
• A school-district AI lab's first rack failed because teams installed the OS before power and cooling validation. What risk does that out-of-order sequence create?
• A housing-authority analytics cluster runbook validates nvidia-smi health only after firmware upgrades finish. Why is that sequencing correct?
• City data-center architects compare fabrics for multi-node training of 311 NLP models with heavy east-west GPU traffic. Which topology choice best fits an AI factory pattern?
• A county assessor training cluster needs non-blocking paths between HGX nodes for all-reduce heavy jobs. Which topology property should planners prioritize?
• Transit operations separates management out-of-band networking from the GPU compute fabric in a municipal AI row. What distinction should the design preserve?
• A public library consortium asks why AI factories often use rail or fat-tree patterns instead of only a classic enterprise three-tier LAN. What is the best explanation?
• A water-utility digital-twin rack plan must relate storage and GPU networks. Which description fits AI factory bring-up design?
• Courts e-filing OCR GPUs scale from one rack to multiple. When should planners expect an AI-factory topology redesign?
• A municipal rack team is bringing up new DGX-class nodes for city services. What should they do with BMC before any in-band OS install?
• County IT insists TPM is initialized and policy-aligned before public-health GPUs enter production. What bring-up outcome does that support?
• The city NOC cannot reach a newly racked GPU node in-band and needs first console access. Which approach is correct?
• A school-district audit finds AI lab BMCs still on default passwords. What initial BMC hardening step should operators take?
• A transit authority maps dedicated OOB switches so GPU nodes stay manageable during fabric outages. What configuration goal does that serve?
• A housing authority verifies TPM enrollment status through BMC tools before accepting a node into the municipal cluster. What gate is being enforced?
• A city upgrades HGX firmware on a permitting-model rack. What should operators do regarding faults before returning nodes to Slurm?
• County public-health operators stage firmware bundles for GPU servers. What coordination practice is most appropriate?
• Transit ops detects a failed firmware step on one HGX tray during a row upgrade. What is the safest immediate response?
• Emergency management needs 911 NLP inference capacity to stay N+1 during firmware windows. How should upgrades be planned?
• A school-district AI lab follows vendor guidance on HGX firmware versus driver updates. Which ordering principle is appropriate?
• A courts OCR cluster re-runs health checks after firmware upgrades. What problem does that practice catch?
• City facilities must clear a dense municipal GPU row for power-on. Which validation should happen first against the AI server requirements?
• A county AI room alarms on inlet temperature during a burn-in rehearsal for health-inference GPUs. What should operators do first for power and cooling validation?
• A transit authority must power multi-PSU HGX nodes for a municipal training row. Which power validation best supports redundancy policy?
• A water-utility digital-twin rack is about to accept new AI nodes. What cooling check belongs in bring-up before sustained load?
• A housing-authority analytics cluster measures rack draw above designed PDU headroom during AI bring-up. What is the correct response?
• A library consortium installs grant-funded GPU servers and must document power for facilities. Which practice validates power parameters at install?
• City administrators finish OS and driver install on new permitting GPU nodes. What SMI-centered check confirms the servers are ready?
• County public-health operators must accept new inference hosts into the queue. Which SMI inventory check is an appropriate install gate?
• Transit operations wants every new HGX host checked the same way during municipal install. How should nvidia-smi fit the workflow?
• A school-district AI lab racks a new GPU server but nvidia-smi reports zero GPUs. What is the best next diagnostic move at install?
• Emergency-management standby inference hosts need known GPU operational settings after install. What should operators validate with nvidia-smi?
• A courts OCR farm accepts new GPU servers into municipal inventory. How should nvidia-smi support asset tracking at install?
• City AI-factory operators finish racking a municipal row and must prove hardware completeness. Which installed-hardware validation fits day-0?
• County public-health IT compares a GPU server’s bill of materials to live discovery before sign-off. What outcome should block acceptance?
• Transit authority hardware validation on a multi-rail GPU node finds a missing NIC. What should operators do?
• A school-district AI lab performs post-install hardware checks on new GPU servers. Which link validation belongs in that gate?
• A housing-authority analytics team reviews day-0 health status on a new GPU server. What indicates the node should not be accepted yet?
• City cabling crews build the GPU fabric for a municipal AI row. How should they choose interconnect media?
• County health-cluster switches are ready for InfiniBand or Ethernet GPU links. What transceiver step belongs in bring-up validation?
• Transit operations finishes pulling GPU fabric cables for a municipal AI factory. What documentation check validates cable types?
• A water-utility AI fabric links with mismatched optics that come up briefly but fail expected signaling margins. What is the correct bring-up conclusion?
• A library consortium AI rack must keep compute and storage cables distinct during bring-up. Which cabling discipline helps?
• A courts OCR cluster designs short-reach GPU links versus longer runs in the same municipal row. When do different cable or transceiver classes apply?
• A city field technician seats a replacement GPU in a municipal HGX server during bring-up. Which physical install practice is required?
• County health racks install additional GPU boards on an AI server. What must be completed before electrical bring-up?
• Transit authority adds GPU capacity and must confirm the physical install before returning a node to service. What integrity check is appropriate?
• A school-district AI lab installs new GPUs and must track physical identity against the purchase order. What install practice is appropriate?
• Emergency-management compute finishes seating GPUs on a standby inference host. When should the chassis be powered on?
• A city permitting cluster must prove new GPUs operate under load before production. Which validation approach fits?
• County public-health nodes show GPUs in idle inventory. What additional gate is required before accepting PHI inference?
• A transit authority runs a municipal train job on new GPUs and wants proof the hardware behaves as expected. What should operators compare?
• Housing-authority analytics sees a node throttle immediately under a known-good test job. What is the correct bring-up response?
• A library consortium completes day-0 workload validation on NLP GPUs. What documentation practice supports later regression compares?
• A city AI factory must reach municipal NFS or parallel storage for training data during bring-up. Which storage step belongs here?
• County health needs datasets reachable on new GPU nodes before BCM imaging finishes. How should storage initial config be treated?
• Transit ops must confirm GPU nodes can use the shared training volume on third-party storage. What should be verified?
• A water-utility digital twin points new GPU servers at storage for checkpoint writes. What endpoint choice is correct at initial config?
• A courts OCR farm configures third-party storage for new GPU servers. Which initial parameters should follow the vendor guide?