> ## Documentation Index
> Fetch the complete documentation index at: https://www.thundercompute.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Monitoring Instances

> View active Thunder Compute  nstances in the CLI. Audit status, IP, and GPU resources of your active cloud compute resources.

<Columns cols={3}>
  <Card title="VS Code" icon="window" href="/vscode/operations/monitoring-instances" color="#95c5ea" className="platform-card">
    Editor extension
  </Card>

  <Card title="CLI" icon="terminal" color="#95c5ea" className="platform-card platform-current">
    Command line
  </Card>

  <Card title="Console" icon="browser" href="/console/operations/monitoring-instances" color="#95c5ea" className="platform-card">
    Web interface
  </Card>
</Columns>

## View Instance Status

List all your instances with their details:

```bash theme={null}
tnr status
```

This displays:

* **Instance ID**: The identifier used for other commands (e.g., `0`)
* **Status**: `RUNNING`, `CREATING`, `RESTORING`, etc.
* **IP Address**: The public IP for SSH access
* **GPU Type**: A6000, A100, or H100
* **GPU Count**: Number of GPUs (production only)
* **vCPU Count**: Number of virtual CPUs
* **RAM**: Total memory
* **Disk**: Storage capacity
* **Template**: The template or snapshot used

### Disable Auto-Refresh

By default, `tnr status` monitors for status changes. Use the `--no-wait` flag to get a single snapshot:

```bash theme={null}
tnr status --no-wait
```

## Instance States

| Status      | Description                                |
| ----------- | ------------------------------------------ |
| `RUNNING`   | Instance is active and ready to use        |
| `CREATING`  | Instance is being provisioned              |
| `RESTORING` | Instance is being restored from a snapshot |
| `DELETING`  | Instance is being removed                  |

## Benchmarking Notes

When measuring performance in **development mode**, note that hardware-level metrics (temperature, wattage, utilization) may not be accurate due to CUDA-level optimizations. Use application-level metrics like iterations per second for reliable comparisons.

For accurate hardware metrics, use **production mode** instances.