Our serverless monitoring for AWS Lambda offers in-depth performance monitoring for your Lambda functions. This document will explain the source, structure, and storage of your Lambda data.
Sources of Lambda data
Our Lambda monitoring data comes from these two sources:
- Our APM agent instrumentation (or similar customer-created instrumentation)
- AWS CloudWatch metrics
For details on how this data is configured and how it flows to New Relic, see the enablement procedures.
The data displayed in the UI is a combination of these data sources. For example, the Overview page displays data reported by instrumentation, while the Metrics page displays CloudWatch data.
Event definitions and attributes
Lambda data is stored in our database (NRDB) as events (data objects with associated attributes).
Lambda data is attached to the following event types. Select an event name to see its attributes.
AwsLambdaInvocationevent: Captures overall timing and associated metadata. A Lambda invocation generates a single
AwsLambdaInvocationErrorevent: If an error occurs during a Lambda, this event will be generated.
Span: This includes detail about a segment of a Lambda function. Spans are used for distributed tracing. Distributed tracing relies on data sampling; 10% of invocations are sampled to generate spans.
- Custom event types: With some agent APIs, custom events can be created and associated with a particular Lambda invocation, and then queried with NRQL.
For more about limits on event storage, see Access and requirements.