Skip to main content

check-dcgmi

GPU diagnostics and NVLink validation using NVIDIA Data Center GPU Manager (DCGM).

Subcommands​

SubcommandPurposeKey Feature
diagHardware diagnostics across multiple test levelsDeployment, integration, hardware, and stress testing with category exclusion
nvlinkNVLink error and status monitoringError threshold validation and link status detection

Quick Start​

Run GPU Diagnostics​

health_checks check-dcgmi diag \
--diag_level 1 \
[CLUSTER] \
app
health_checks check-dcgmi nvlink \
--check nvlink_errors \
--gpu_num 8 \
[CLUSTER] \
app
health_checks check-dcgmi nvlink \
--check nvlink_status \
--gpu_num 8 \
[CLUSTER] \
app