Skip to content
  • VM Provisioning

GPU capacity request with approval and driver verification

Accelerated capacity is scarce and expensive, and a build with the wrong driver stack wastes both the allocation and the customer's time.

The workflow

How StackHeal handles it

Each step is a registered action, a condition or an approval. Nothing here is a general-purpose command.

  1. 1

    Validate the requested accelerator against live inventory

  2. 2

    Hold for approval above a cost threshold

  3. 3

    Provision with the correct driver and runtime image

  4. 4

    Confirm the accelerator is visible to the workload

What the run verifies

Deliberate allocation with the approval recorded and the driver stack confirmed working before hand-off.

Where a person stays involved

Any step that changes state can be configured to require an approval, and a run that cannot verify its own outcome escalates rather than closing. The confidence threshold, the blast-radius limit and the approver are all yours to set.

Related products

StackHeal products share one workflow engine, one action catalogue and one run history.

Learn more about VM Provisioning

VM Provisioning

Standardize virtual machine provisioning and infrastructure lifecycle operations.

Learn more

See this workflow built against your systems.

We will walk through the intake, the policy checks, the approval points and the verification step using your own tooling.