Validate that the AMD SEV-SNP + NVIDIA Blackwell confidential compute stack runs end-to-end on a Kubernetes cluster deployed via Kubespray, using the official Akash provider-playbooks install script (akash-network/provider-playbooks). This is the production-relevant deployment path for Akash GPU providers; the K3s path is tracked separately as an alternative.
Scope: stock Ubuntu 26.04 LTS on the same hardware; Kubespray v2.31.0 → Kubernetes v1.35.4, containerd 2.2.3, Calico 3.31.5; the Akash install script driven interactively with only the Kubespray component selected (no GPU drivers, no Akash provider service, no other optional components); kata-deploy v3.29.0; NVIDIA GPU Operator v26.3.1 in sandboxWorkloads.mode=kata. Validation criterion: same as K3s path — both runtime classes pass with SEV-SNP active at VMPL0 inside the guest, CUDA workload completes on the confidential GPU. Aligns with the broader Akash TEE workstream.
Validate that the AMD SEV-SNP + NVIDIA Blackwell confidential compute stack runs end-to-end on a Kubernetes cluster deployed via Kubespray, using the official Akash
provider-playbooksinstall script (akash-network/provider-playbooks). This is the production-relevant deployment path for Akash GPU providers; the K3s path is tracked separately as an alternative.Scope: stock Ubuntu 26.04 LTS on the same hardware; Kubespray v2.31.0 → Kubernetes v1.35.4, containerd 2.2.3, Calico 3.31.5; the Akash install script driven interactively with only the Kubespray component selected (no GPU drivers, no Akash provider service, no other optional components); kata-deploy v3.29.0; NVIDIA GPU Operator v26.3.1 in
sandboxWorkloads.mode=kata. Validation criterion: same as K3s path — both runtime classes pass with SEV-SNP active at VMPL0 inside the guest, CUDA workload completes on the confidential GPU. Aligns with the broader Akash TEE workstream.