Disable automatic Tesla driver installation for OS replacement - Elastic GPU Service

When you create a GPU-accelerated instance, if you select an image and the Auto-install GPU Driver option, a GPU (Tesla) driver is automatically installed after the instance is created. If you need to change the operating system of the GPU-accelerated instance for any reason, such as the current OS not meeting your business requirements, you must also disable the automatic Tesla driver installation feature and manually install a Tesla driver that is compatible with the new OS. This ensures that you can properly use the high-performance features of the GPU-accelerated instance.

Procedure

Connect to the GPU-accelerated instance.

For more information, see Log on to a Linux instance by using Workbench.

Run the nvidia-smi command to check the Tesla driver version.

[ecs-uxxxx@taZ ~]$ nvidia-smi
Tue Aug  6 17:36:22 2024
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.90.07              Driver Version: 550.90.07      CUDA Version: 12.4     |
|----------------------------------------------+------------------------+------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  Tesla T4                       On  |   00000000:00:07.0 Off |                    0 |
| N/A   28C    P8              9W /  70W  |       1MiB /  15360MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI        PID   Type   Process name                          GPU Memory     |
|        ID   ID                                                           Usage          |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+

Stop the GPU-accelerated instance.

When stopping the instance, set Stop Mode to Standard Mode to prevent startup failures after you replace the operating system. For more information, see Stop an instance.
In the instance list, find the stopped GPU-accelerated instance. In the Actions column, choose Instance Settings > Set User Data.
In the User Data section, delete the content and click OK.
Replace the operating system of the GPU-accelerated instance.

This operation essentially replaces the system disk. You can replace the operating system by changing the instance's image. For more information, see Replace the operating system (system disk).
In the instance list, click the ID of the modified instance. On the Instance Details tab, verify that the operating system and image information is updated.
Reconnect to the GPU-accelerated instance and run the nvidia-smi command to verify that automatic Tesla driver installation is disabled.
```
[ecs-user@xxx ~]$ nvidia-smi
-bash: nvidia-smi: command not found
```

Next steps

To use the high-performance features of the GPU-accelerated instance, you must now manually install a compatible driver. For more information, see Manually install a Tesla driver on a GPU-accelerated compute-optimized instance (Linux).