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

# Inspect sandbox metrics

> Track a sandbox's requests, error rate, CPU, and memory in the dashboard to investigate workload behavior.

Use metrics when an agent's task runs longer than expected or fails under load.
The charts show request traffic and resource usage for a sandbox until it's
deleted.

## Open a sandbox's metrics

1. In the [Baseten dashboard](https://app.baseten.co), choose **Sandboxes**,
   then choose the sandbox.
2. On **Overview**, check the sandbox's
   [deployment status](/sandboxes/lifecycle#deployment-status), then select
   the time range that contains the operation you're investigating.

## Read the charts

The **Overview** includes these charts:

| Chart | What it shows | What to investigate |
| - | - | - |
| API requests | Requests per second. | Compare traffic with the work your application submitted. |
| Error rate | The percentage of requests with a non-2xx status. | Check the failed requests in the [logs](/sandboxes/logs). |
| CPU usage | CPU utilization. | Check CPU-intensive commands and concurrent work. |
| Memory usage | Memory used, with the allocated memory shown as a limit. | Check large inputs and processes that retain data. |

**Error rate** counts HTTP failures. The execution API can return HTTP `200`
when a command exits with a nonzero code, so check the command's exit status
to detect process failures.

A sandbox's memory also sets its CPU allocation. If a workload reaches the
memory limit or needs more CPU, [create a sandbox](/sandboxes/manage#create-a-sandbox)
with more memory. Memory can't change after creation.