Passive: GPU Health Monitoring
- ✓DCGM background health checks enabledDCGM diagnostics
- ✓GPUs falling off the bus monitoring (XID 79,
NVML_ERROR_GPU_IS_LOST)NVIDIA XID errors - ✓GPU and CPU memory ECC errors (SBE/DBE volatile and aggregate)
Passive: Network and Hardware Monitoring
- ✓PCIe errors via NVML and DCGM counters (replay/error thresholds)
- ✓Ethernet and InfiniBand link flaps (
ethtool,ibdiagnet,ibportstate) - ✓GPU temperature monitoring (
DCGM_FI_DEV_GPU_TEMP)DCGM field IDs
Passive: Error Detection and Performance
- ✓Uncorrectable NVIDIA XID and SXID error code monitoringNVIDIA XID errors
- ✓Stalled NCCL/RCCL job detection (
GPU_UTILvs power consumption) - ✓InfiniBand health (PKey consistency, link specs, error counters)
Active: GPU Diagnostics and Performance
- ✓NVIDIA DCGM diag level 3 with Extensive Testing (EUD)DCGM diagnostics
- ✓DtoH and HtoD bandwidth testing for PCIe performance validationnvbandwidth
- ✓gpu-burn/gpu-fryer for validating GPU under loadgpu-burn
Active: Communication and Network Testing
- ✓Local NCCL all reduce tests for NVLink/NVSwitch/NVLS performancenccl-tests
- ✓Local InfiniBand all reduce test (with
NCCL_P2P_DISABLE=1)nccl-tests - ✓Pairwise GPU
ib_write_bwandib_write_latencybidirectional testsRDMA perftest
Active: Hardware Validation and AI Workload Testing
- ✓NVIDIA TinyMeg2 for hardware correctness and SDC-free validation
- ✓Megatron or TorchTitan tests for TFLOP/s/GPU performance and loss convergenceTorchTitan
Automation
- ✓Weekly scheduled active health checks on idle nodes
- ✓NCCL and scheduler topology health validation (Slurm
topology.yaml/topology.conf; K8s topology-aware scheduling, gang scheduling, bin packing) - ✓GPU/node health detection, e.g. Node Problem Detector, GPUd, or equivalent, with automated cordon/drain and repair/replace, e.g. Draino, NVSentinel, or equivalentNode Problem DetectorGPUdDrainoNVSentinel
- ✓NVLink connectivity and error tracking (critical for NVL72)
- ✓AI/ML-based prediction of failures
General Expectations
- ✓Console, dashboard, CLI and/or API available to manage resources
- ✓24x7 support availability
- ✓Process for security fixes and upgrades exists, proactive notifications are clear
- ✓Integration with comprehensive monitoring and alerting systems