# Send Render Metrics

> Stream the metrics of your Render services, databases, and Key Value instances to a Telemetry Machine metrics dataset.

Render can stream the metrics of a workspace over OpenTelemetry. Telemetry Machine stores them in a metrics dataset.

## Before you start

- Your Render workspace must be on the Pro plan or higher.
- You must be an admin of the Render workspace.
- To create a dataset or an API key in Telemetry Machine, you must be an owner or an admin of the organization.

## Create a metrics dataset

### In the application

1. Open the **Datasets** page.
1. Select **New dataset**.
1. Select **Metrics**.
1. Select **Next**.
1. In **Dataset name**, type a name for the dataset, such as `render-metrics`.
1. Select **Next**.
1. Select **Create & Setup**.
1. Copy the API key from **Your API Key**. Keep it in a safe location.

### With tm-cli

```console
$ tm-cli datasets create --name render-metrics --type metrics
$ tm-cli api-keys create --name render --dataset render-metrics
```

The output shows the full key in `api_key.api_key`. It shows the full key one time only.

## Add the destination in Render

1. In the Render Dashboard, open your workspace, then select **Observability**.
1. Under **Metrics Stream**, select **+ Add destination**.
1. In the provider list, select **Custom**.
1. In the endpoint field, type the URL below, with the name of your dataset.
1. In the API key field, paste the Telemetry Machine API key.
1. Select **Add destination**.

```text
https://api.telemetrymachine.com/datasets/render-metrics
```

Render sends the key as `Authorization: Bearer <key>`. The dataset name in the URL takes the place of the `X-Dataset-Name` header.

## Query the metrics

Render names its metrics `render.service.*`, `render.postgres.*`, and `render.keyvalue.*`, such as `render.service.cpu.time` and `render.service.memory.usage`. The `service.name` and `service.id` attributes tell the services apart, and `service.instance.id` tells their instances apart. See Render's [metrics reference](https://render.com/docs/metrics-streams-reference).

`render.service.http.requests.total` is a cumulative count of requests. `render.service.http.requests.latency` has the p50, p95, and p99 latency in seconds, with the `quantile` attribute.

Static sites, cron jobs, and one-off jobs have no metrics.

---

All pages: https://telemetrymachine.com/docs/llms.txt