This is a OneAgent extension that can run on any OneAgent-monitored host (whether full stack or infra-only). The extension automatically starts monitoring NVIDIA GPUs detected on the host.

Requirements

• An NVIDIA card supported by gpustat==1.0.0b1
• More details: https://github.com/wookayin/gpustat

Metrics collected

• Temperature
• Memory Used
• Memory Total
• Utilization
• Processes

1. Deploy to OneAgent hosts

This extension is developed with the Dynatrace Extensions 1.0 Framework. This means that the .zip archive you have been provided with must be deployed on every OneAgent host that should run it.

Upload the .zip file and then extract it in the extension deployment directory on the OneAgent host. By default, this is found at:

• On Windows: %PROGRAMFILES%\dynatrace\oneagent\plugin_deployment
• On Linux: /opt/dynatrace/oneagent/plugin_deployment

2. Upload to Dynatrace

Once the extension is deployed to each OneAgent, this must also be uploaded to Dynatrace. Open the Dynatrace UI and navigate to Settings > Monitored technologies > Custom extensions tab, then click Upload extension and upload the same zip archive you have been provided with.

Once the extension appears in the list below you can move to the next step:

This extension does not require any configuration. Once it is uploaded to Dynatrace, it will be automatically activated on any OneAgent host where it is found.

If this is not what you intended, you can disable the global configuration, and enable individual hosts that should run this extension.

Global configuration

Go to Settings > Monitored technologies > Custom extensions tab and click on the extension name from the list. On the page that opens you can enable or disable the global behavior of the extension with the switches called Monitor the environment and Monitor the environment for hosts in infrastructure-only monitoring mode

Host level configuration

From the same page as above, click on the name of a host from the list of hosts. The host settings page opens. Here, expand the row containing NVIDIA. First, enable the switch called Use host configuration. Then, choose whether to enable the extension for this host or not.

To start exploring metrics, open the details page of one of the hosts running the extension.

You will see a specialised tile displaying two of the metrics. Click on it.

You can click on Further details from the below tabs to get detailed metrics per GPU.