Before an HGX firmware upgrade, what should you confirm?
Why D is the answer
The target version and supported procedure
Use the supported set and take the node out of service.
In detail
- Use the supported version and procedure, with the node out of service.
- Use the supported set.
Use case
An illustrative scenario, not a real incident.
Before an HGX upgrade, a team confirms the target version and the procedure, and takes the node out of service.
See it in different ways
flowchart TD A["Supported release set"] --> B["GPU and HGX"] A --> C["NIC and BlueField"] A --> D["Switches"] A --> E["Transceivers"]
sequenceDiagram participant F as Facility participant B as BMC participant W as Firmware participant O as OS and drivers F->>B: Power and network B->>W: Update through out-of-band access W->>O: Install the OS and drivers O->>O: Confirm GPUs, then run diagnostics
flowchart LR S0["Confirm target version"] --> S1["Confirm procedure"] --> S2["Take node out of service"] --> S3["Upgrade"]
Key point: Confirm version and procedure.
Why the other options are wrong
- A. That the nodes are running a full HPL load
- A load test during an update is unsafe.
- B. That all fans have been removed for access
- Fans are needed.
- C. That Slurm has queued a training job on it
- A job should not run during an update.