“lsmod and nvidia-smi show loaded driver and reachable GPUs” addresses driver module was not rebuilt or loaded for the new kernel, but the current hardware or infrastructure issue is mixed installation methods have left inconsistent GPU driver files. The option “package verification and functional tests show one coherent driver stack” is more directly tied to the required outcome.

“DOCA version and DPU communication tests pass on the qualified stack” addresses dOCA host installation ignored the supported compatibility matrix, but the current hardware or infrastructure issue is mixed installation methods have left inconsistent GPU driver files. The option “package verification and functional tests show one coherent driver stack” is more directly tied to the required outcome.

“new driver loads and representative CUDA test passes after jobs resume” addresses driver lifecycle change was executed without workload draining, but the current hardware or infrastructure issue is mixed installation methods have left inconsistent GPU driver files. The option “package verification and functional tests show one coherent driver stack” is more directly tied to the required outcome.

Useful for dependent NVIDIA container packages are not version coherent, but it does not remove the specific condition: container runtime has not been configured to expose GPU devices.

“set the supported GPU device request in the container invocation” addresses container launch configuration does not request GPUs through the runtime, but the current hardware or infrastructure issue is container runtime has not been configured to expose GPU devices. The option “install the NVIDIA container toolkit and configure its supported Docker runtime integration” is more directly tied to the required outcome.

“configure the approved NVIDIA container runtime hook rather than copying host libraries manually” addresses nVIDIA GPU runtime hook is missing from container creation, but the current hardware or infrastructure issue is container runtime has not been configured to expose GPU devices. The option “install the NVIDIA container toolkit and configure its supported Docker runtime integration” is more directly tied to the required outcome.

“container CUDA initialization succeeds against the host driver” addresses nVIDIA GPU runtime hook is missing from container creation, but the current hardware or infrastructure issue is container runtime has not been configured to expose GPU devices. The option “a bounded container sees the assigned GPU through nvidia-smi” is more directly tied to the required outcome.

“package signature and installed release match the approved build manifest” addresses container runtime supply chain provenance was not established, but the current hardware or infrastructure issue is container runtime has not been configured to expose GPU devices. The option “a bounded container sees the assigned GPU through nvidia-smi” is more directly tied to the required outcome.

“an explicit GPU-request run exposes exactly the authorized GPU UUID” addresses container launch configuration does not request GPUs through the runtime, but the current hardware or infrastructure issue is container runtime has not been configured to expose GPU devices. The option “a bounded container sees the assigned GPU through nvidia-smi” is more directly tied to the required outcome.

“verify signed repository configuration and supported toolkit release” addresses container runtime supply chain provenance was not established, but the current hardware or infrastructure issue is service-level runtime settings were not applied to the active daemon. The option “apply the toolkit runtime configuration and reload Docker through supported workflow” is more directly tied to the required outcome.

“reinstall the supported matching toolkit and library set” addresses dependent NVIDIA container packages are not version coherent, but the current hardware or infrastructure issue is service-level runtime settings were not applied to the active daemon. The option “apply the toolkit runtime configuration and reload Docker through supported workflow” is more directly tied to the required outcome.

“set the supported GPU device request in the container invocation” addresses container launch configuration does not request GPUs through the runtime, but the current hardware or infrastructure issue is service-level runtime settings were not applied to the active daemon. The option “apply the toolkit runtime configuration and reload Docker through supported workflow” is more directly tied to the required outcome.

“package inventory and GPU container samples pass without library load errors” addresses dependent NVIDIA container packages are not version coherent, but the current hardware or infrastructure issue is service-level runtime settings were not applied to the active daemon. The option “docker info and a GPU-enabled sample container show configured GPU access” is more directly tied to the required outcome.

“a bounded container sees the assigned GPU through nvidia-smi” addresses container runtime has not been configured to expose GPU devices, but the current hardware or infrastructure issue is service-level runtime settings were not applied to the active daemon. The option “docker info and a GPU-enabled sample container show configured GPU access” is more directly tied to the required outcome.

“package signature and installed release match the approved build manifest” addresses container runtime supply chain provenance was not established, but the current hardware or infrastructure issue is service-level runtime settings were not applied to the active daemon. The option “docker info and a GPU-enabled sample container show configured GPU access” is more directly tied to the required outcome.

Useful for container runtime has not been configured to expose GPU devices, but it does not remove the specific condition: NVIDIA GPU runtime hook is missing from container creation.

“set the supported GPU device request in the container invocation” addresses container launch configuration does not request GPUs through the runtime, but the current hardware or infrastructure issue is nVIDIA GPU runtime hook is missing from container creation. The option “configure the approved NVIDIA container runtime hook rather than copying host libraries manually” is more directly tied to the required outcome.

“use NVIDIA container toolkit with limited device capabilities and least privilege” addresses excessive container privileges substitute for targeted GPU runtime integration, but the current hardware or infrastructure issue is nVIDIA GPU runtime hook is missing from container creation. The option “configure the approved NVIDIA container runtime hook rather than copying host libraries manually” is more directly tied to the required outcome.

“a bounded container sees the assigned GPU through nvidia-smi” addresses container runtime has not been configured to expose GPU devices, but the current hardware or infrastructure issue is nVIDIA GPU runtime hook is missing from container creation. The option “container CUDA initialization succeeds against the host driver” is more directly tied to the required outcome.

“identical GPU container smoke jobs pass on representative nodes in each category” addresses cluster image drift created inconsistent GPU container runtime capability, but the current hardware or infrastructure issue is nVIDIA GPU runtime hook is missing from container creation. The option “container CUDA initialization succeeds against the host driver” is more directly tied to the required outcome.

“package signature and installed release match the approved build manifest” addresses container runtime supply chain provenance was not established, but the current hardware or infrastructure issue is nVIDIA GPU runtime hook is missing from container creation. The option “container CUDA initialization succeeds against the host driver” is more directly tied to the required outcome.

“verify signed repository configuration and supported toolkit release” addresses container runtime supply chain provenance was not established, but the current hardware or infrastructure issue is cluster image drift created inconsistent GPU container runtime capability. The option “deploy toolkit through managed node images and validate every category” is more directly tied to the required outcome.

“apply the toolkit runtime configuration and reload Docker through supported workflow” addresses service-level runtime settings were not applied to the active daemon, but the current hardware or infrastructure issue is cluster image drift created inconsistent GPU container runtime capability. The option “deploy toolkit through managed node images and validate every category” is more directly tied to the required outcome.

“diff preserve and validate approved daemon runtime configuration through change control” addresses runtime upgrade was not reconciled with local daemon settings, but the current hardware or infrastructure issue is cluster image drift created inconsistent GPU container runtime capability. The option “deploy toolkit through managed node images and validate every category” is more directly tied to the required outcome.

“container CUDA initialization succeeds against the host driver” addresses nVIDIA GPU runtime hook is missing from container creation, but the current hardware or infrastructure issue is cluster image drift created inconsistent GPU container runtime capability. The option “identical GPU container smoke jobs pass on representative nodes in each category” is more directly tied to the required outcome.

“package inventory and GPU container samples pass without library load errors” addresses dependent NVIDIA container packages are not version coherent, but the current hardware or infrastructure issue is cluster image drift created inconsistent GPU container runtime capability. The option “identical GPU container smoke jobs pass on representative nodes in each category” is more directly tied to the required outcome.

“a bounded container sees the assigned GPU through nvidia-smi” addresses container runtime has not been configured to expose GPU devices, but the current hardware or infrastructure issue is cluster image drift created inconsistent GPU container runtime capability. The option “identical GPU container smoke jobs pass on representative nodes in each category” is more directly tied to the required outcome.

Useful for container runtime has not been configured to expose GPU devices, but it does not remove the specific condition: excessive container privileges substitute for targeted GPU runtime integration.

“verify signed repository configuration and supported toolkit release” addresses container runtime supply chain provenance was not established, but the current hardware or infrastructure issue is excessive container privileges substitute for targeted GPU runtime integration. The option “use NVIDIA container toolkit with limited device capabilities and least privilege” is more directly tied to the required outcome.

“deploy toolkit through managed node images and validate every category” addresses cluster image drift created inconsistent GPU container runtime capability, but the current hardware or infrastructure issue is excessive container privileges substitute for targeted GPU runtime integration. The option “use NVIDIA container toolkit with limited device capabilities and least privilege” is more directly tied to the required outcome.

“package signature and installed release match the approved build manifest” addresses container runtime supply chain provenance was not established, but the current hardware or infrastructure issue is excessive container privileges substitute for targeted GPU runtime integration. The option “GPU job runs while unrelated host devices and privileged operations remain inaccessible” is more directly tied to the required outcome.

“upgraded production worker passes runtime configuration and GPU smoke test” addresses runtime upgrade was not reconciled with local daemon settings, but the current hardware or infrastructure issue is excessive container privileges substitute for targeted GPU runtime integration. The option “GPU job runs while unrelated host devices and privileged operations remain inaccessible” is more directly tied to the required outcome.

“package inventory and GPU container samples pass without library load errors” addresses dependent NVIDIA container packages are not version coherent, but the current hardware or infrastructure issue is excessive container privileges substitute for targeted GPU runtime integration. The option “GPU job runs while unrelated host devices and privileged operations remain inaccessible” is more directly tied to the required outcome.