Skip to main content

Metrics

Fully managed Prometheus metrics available for your cloud environment.

On CoreWeave, a wide range of Prometheus metrics are made available by default. The following guides presume that you already have an active CoreWeave account including a functioning API Access Token.

Once logged in to your CoreWeave Cloud account, visit the Prometheus dashboard at https://prometheus.ord1.coreweave.com.

API key authentication

To authenticate using HTTP, access the Prometheus server by sending your Access Token as the Authorization Header value in the HTTP request.

Note

Replace <TOKEN> with your CoreWeave Access Token.

ParameterValue
MethodsGET or POST
HeaderAuthorization
ValueBearer <TOKEN>

Selecting a Data Source

For the URL parameter, select the appropriate endpoint for the target date range:

URLDate
https://prometheus.ord1.coreweave.com/CoreWeave Metrics (current data)
https://prometheus-legacy.ord1.coreweave.com/Legacy CoreWeave Metrics (older data)
Historical data

For queries that require historical data collected prior to October 21, 2024, we recommend first querying the current Prometheus endpoint, as it is significantly more performant. If the returned data contains significant gaps, validate the same query against the Legacy Prometheus system.

Example Prometheus query using an Access Token

POST https://prometheus.ord1.coreweave.com/api/v1/query

Example usage for sending a request to the Prometheus API using your Access Token.

Query Parameters

NameTypeDescription
timenumberUnix timestamp of current time
querystringThe PromQL Query (kube_pod_container_info)

Headers

NameTypeDescription
AuthorizationstringSet value to Bearer <TOKEN> as the value, replacing <TOKEN> with your CoreWeave Access Token.

Example request

Example
$
curl -X POST 'https://prometheus.ord1.coreweave.com/api/v1/query?time=<UNIX_TIMESTAMP>&query=<PROMQL_QUERY>' \
-H 'Authorization: Bearer <TOKEN>'

Responses

ValueTypeDescription
200numberSuccess

Example Prometheus response data:

Example
{
"status": "success",
"data": {
"resultType": "vector",
"result": [{
"metric": {
"__name__": "kube_pod_container_info",
"container": "example",
"container_id": "docker://1548c7ce5d717cfb9e3e797464d9fe4fbda0cea6f324fda8b9b64b2f5fd44e2f",
"endpoint": "http",
"image": "example/example_image:1.6.4",
"image_id": "docker-pullable://docker-registry-image",
"instance": "10.0.0.1:8080",
"job": "kube-state-metrics",
"namespace": "tenant-example",
"pod": "example-5d796977bc-crk9m",
"service": "tenant-metrics-kube-state-metrics"
},
"value": [1629254349.447, "1"]
}]
}
}

Custom metric scraping

CoreWeave's hosted Prometheus service may be used to scrape custom targets using a ServiceMonitor. Annotations from the ServiceMonitor must match the Service hosting the Prometheus exporter endpoint, usually at the path /metrics.

Below is an example of a Prometheus exporter scraped by the Prometheus service.

Important

The value of .endpoints.port must be the literal string http, and may not be replaced with a port number integer (e.g., 80).

Example
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
name: example
namespace: tenant-example
labels:
app: example
spec:
jobLabel: example-scraping
selector:
matchLabels:
app: example
namespaceSelector:
matchNames:
- tenant-example
endpoints:
- port: http
scheme: http
path: /metrics
interval: 15s
scrapeTimeout: 15s