OpenTelemetry Collector exporter

SketchLog already includes an in-process Python OpenTelemetry integration. The Collector exporter is different: it lets an existing OpenTelemetry Collector pipeline send selected metrics, traces, and logs into SketchLog without changing application code.

What it does

The exporter converts OpenTelemetry signals into SketchLog event batches:

Signal	Conversion
Selected metrics	mapped to latency, event, or unique streams
Traces	span durations in milliseconds
Logs	event counts from EventName, severity, or log_record fallback

It writes to the existing SketchLog ingestion API:

POST /v1/namespaces/{namespace}/streams/{stream}/events

Example Collector config

receivers:
  otlp:
    protocols:
      grpc:
      http:

exporters:
  sketchlog:
    endpoint: http://sketchlog:8000
    namespace: production
    auth_token: ${env:SKETCHLOG_AUTH_TOKEN}
    timeout: 10s
    max_batch_items: 1000
    metrics:
      - name: http.server.duration
        stream: api.latency
        kind: latency
        scale: 1000
      - name: app.errors
        stream: app.events
        kind: event
        event_name: error
    span_duration_stream: traces.duration
    log_event_stream: logs.events

service:
  pipelines:
    metrics:
      receivers: [otlp]
      exporters: [sketchlog]
    traces:
      receivers: [otlp]
      exporters: [sketchlog]
    logs:
      receivers: [otlp]
      exporters: [sketchlog]

Metric mapping

Each metric mapping has:

• name: OTEL metric name to match.
• stream: SketchLog stream to write.
• kind: latency, event, or unique.
• event_name: required for event mappings.
• scale: optional multiplier, useful for converting seconds to milliseconds.

Authentication

If the SketchLog server uses SKETCHLOG_AUTH_TOKEN, set auth_token from an environment variable. The exporter sends it as X-SketchLog-Auth-Token.

Module location

The Collector module lives in:

contrib/otelcol-sketchlog-exporter

It is intended for use with a custom Collector build.