Can customers request hardware-level diagnostics?
Here is how you can access hardware-level insights in OCI today:
For Bare Metal instances and GPU clusters, OCI exposes several hardware-level data points directly through the OCI Monitoring service. You don't need to ask for these; they are streamed to your console:
Infrastructure Health Metrics: These indicate the status of the underlying physical host, including power supply health and cooling status within your Fault Domain.
Local NVMe Health: For "Dense I/O" shapes, you can see SMART-equivalent data for your local drives to track wear-and-tear or impending failures.
GPU Diagnostics: Real-time visibility into GPU temperature, power draw, and ECC (Error Correction Code) memory errors.
If your server is crashing and the standard metrics don't show why, you can request a Diagnostic Reboot or a Support-Led Hardware Audit.
Diagnostic Reboot: When you trigger this from the console, OCI doesn't just "turn it off and on." It performs a hardware-level check of the motherboard and NICs. If a fault is found, the system automatically migrates your VM or alerts a technician to swap your Bare Metal host.
Support Requests (SRs): You can open a "Severity 2" ticket requesting a hardware diagnostic report. Oracle Support can access low-level logs (such as the System Event Log or IPMI data) that are not exposed to the public console for security reasons.
If you are renting an Exadata Cloud Service or a Database Bare Metal instance, you have access to the Autonomous Health Framework.
This tool acts as a "flight recorder" for your hardware and software.
It automatically collects diagnostic bundles when a hardware anomaly is detected.
User Action: You can run a simple command (e.g., ahf generate-report) to get a comprehensive health check that includes OS anomalies and hardware configuration drift.
| Feature | Accessible via Console? | Requires Support Request? |
| CPU/Memory Usage | Yes | No |
| GPU Temperature/Power | Yes | No |
| Disk SMART Status | Yes (on select shapes) | No |
| Chassis Event Logs | No | Yes |
| Physical Memory Maps | No | Yes (for security) |
| Power Supply Failures | Yes (Infrastructure Health) | No |
New in 2026, OCI has introduced an Audit Hardware button for Bare Metal users.
How it works: This triggers a "Hardware Root of Trust" verification.
The Result: It generates a signed report confirming that the firmware has not been tampered with and that the physical components match the advertised specifications of your "rental" shape. This is particularly popular for high-security and defense customers.
In the event of a security incident, Oracle offers Cloud Forensics support. Under specific enterprise contracts, you can request hardware-level forensic images to prove that a breach did not occur at the physical "hypervisor" or "firmware" level.
To see the most hardware-level data in your console, ensure the Oracle Cloud Agent is installed and the "Compute Instance Monitoring" and "Compute Instance Run Command" plugins are Enabled. Without these, OCI can see the physical rack health, but it cannot "reach inside" the server to tell you what your internal sensors are seeing.