Trace Drains Reference
Trace Drains forward distributed tracing data from your deployments to external endpoints for storage and analysis. You can configure Trace Drains in two ways:
- Custom endpoint: Send traces to any HTTP endpoint you configure
- Native integration: Use integrations from the Vercel Marketplace like Braintrust
Vercel sends traces to endpoints over HTTPS following the OpenTelemetry Protocol (OTLP) specification.
Trace Drains follow the OpenTelemetry traces specification. Vercel automatically adds these specific resource attributes to all traces:
| Name | Type | Description | Example |
|---|---|---|---|
vercel.projectId | string | Identifier for the Vercel project | "Qmc52npNy86S8VV4Mt8a8dP1LEkRNbgosW3pBCQytkcgf2" |
vercel.deploymentId | string | Identifier for the Vercel deployment | "dpl_2YZzo1cJAjijSf1hwDFK5ayu2Pid" |
Vercel supports the following formats for Trace Drains. You can configure the format when configuring the Drain destination:
Vercel sends traces as JSON objects following the OpenTelemetry specification:
{ "resourceSpans": [{ "resource": { "attributes": [{ "key": "service.name", "value": { "stringValue": "vercel-function" } }] }, "scopeSpans": [{ "scope": { "name": "vercel" }, "spans": [{ "traceId": "7bba9f33312b3dbb8b2c2c62bb7abe2d", "spanId": "086e83747d0e381e", "name": "GET /api/users", "kind": "server", "startTimeUnixNano": "1694723400000000000", "endTimeUnixNano": "1694723400150000000" }] }] }] }Vercel sends traces in binary protobuf format following the OTLP/HTTP specification. This format is more efficient for high-volume trace data transmission.
Sampling rules control how much trace data each drain forwards so you can manage observability depth and spend. Add sampling rules to define how much data reaches your destination:
- If no rules exist, click Add sampling rule.
- Choose the environment you want to sample from.
- Set the sampling percentage.
- (Optional) Specify a request path prefix. Leave it blank to apply the rule to every path.
Example workflows:
- Launch-day monitoring: sample 100% of production traffic when you launch a new feature, then decrease to 10% once traffic stabilizes.
- Static coverage: always collect 5% from
/docsso you can spot regressions on a static documentation site.
Rules run from top to bottom. Requests that match a rule use that rule’s sampling rate, and any other requests are dropped. If you do not add rules, the drain forwards 100% of data to the destination.
Custom spans from functions using the Edge runtime are not forwarded via the Trace Drain.
For more information on Trace Drains and how to use them, check out the following resources:
Was this helpful?