Skip to navigation
ReferenceCheck Catalog

Node Checks

View as Markdown

Node covers host-wide GPU-system signals that are not bound to any single device — GPU count, driver version, and similar inventory checks.

GPU Count

gpu_count

Verifies a node enumerated the expected number of GPUs for its architecture.

PropertyValue
Provenancenvidia
Condition scopeNODE
Execution classPASSIVE
Scheduling cadenceCONTINUOUS
Maturitystable
Execution contextsOPERATOR_INFRA, TENANT_GUEST
ArchitecturesGB200, GB300
Required signalsgpu_count

Thresholds

NameValueUnitCategoryDescription
expected_gpu_count4countDEFAULTExpected number of GPUs on a node. Override per fleet.

Rules

  • CRITICAL when gpu_count != expected_gpu_count
    • Message: Expected {expectedgpucount}GPUs,found\{expected_gpu_count\} GPUs, found {gpu_count}.

Recommended action: operational state drain, then reset_gpu → restart_os → service_required