Telemetry

OpenTelemetry → Grafana Cloud. Every Lambda runs two AWS Lambda extension layers that ship signals to a shared OTLP collector.

Signals

Signal	Destination	Source
Traces	Tempo (`grafanacloud-traces`)	X-Ray + OTel instrumentation
Logs	Loki	Lambda structured logs (JSON)
Metrics	Prometheus	OTel + CloudWatch

Architecture

flowchart LR
  L[Lambda] -->|AWS_LAMBDA_EXEC_WRAPPER| INST[OTel Node instrumentation]
  L -->|telemetryapi| COL[Collector layer - Grafana]
  INST -->|otlp http/protobuf localhost:4318| COL
  COL -->|otlphttp/drive_otlp| OTLP["otlp.&lt;root-domain&gt;"]
  OTLP --> GC[Grafana Cloud]

Two layers per Lambda (from serverless.common.yml custom.otelCollect + custom.otelInstrument):

Collector layer (Grafana collector-lambda-extension) — receives OTLP on localhost:4318, exports to DRIVE_OTLP_ENDPOINT with DRIVE_OTLP_AUTH_HEADER.
Instrumentation layer (aws-otel-nodejs-arm64) — auto-instruments Node (aws-sdk, http/https, dns, host) via AWS_LAMBDA_EXEC_WRAPPER.

Resource attributes

Set via OTEL_RESOURCE_ATTRIBUTES:

service.name=${service}
service.namespace=aws.lambda
deployment.environment=${stage}
cloud.provider=aws
cloud.platform=aws_lambda
cloud.region=${region}

Collector backstops service.namespace, deployment.environment(.name) (X-Ray receiver path doesn't see env vars).

Pages

Topic	Page
Collector config, layers, OTLP export	OpenTelemetry
TraceQL queries in Tempo	Tracing
Structured JSON logs in Loki	Logs
CloudWatch + Prometheus metrics	Metrics
`Logger`/`TraceUtils` lib + drive-ai attrs	drive-ai integration