Local Observability Stack

In this tutorial, you will run the local query-only flow with Docker:

OpenTelemetry Demo -> duckdb-otlp OTLP writer -> Quack DuckLake catalog -> canardstack query APIs -> Grafana

You will send OpenTelemetry Demo telemetry to duckdb-otlp, query it through canardstack, and open a provisioned Grafana dashboard.

For duckdb-otlp writer configuration outside this Compose stack, use the duckdb-otlp documentation.

Prerequisites

• Docker Compose
• A local OpenTelemetry Demo checkout

Start the Stack

From the canardstack repository root:

docker compose up

This starts:

• DuckDB serving a DuckLake catalog over Quack
• the published duckdb-otlp image running OTLP/HTTP ingest on localhost:4318
• canardstack on localhost:9090
• Grafana on localhost:3000

Grafana and canardstack are ready when the containers finish starting.

To test local canardstack image changes, add the build override:

docker compose -f compose.yaml -f compose.build.yaml up --build

Connect the OpenTelemetry Demo

In your OpenTelemetry Demo checkout, add canardstack as an OTLP/HTTP backend in src/otel-collector/otelcol-config-extras.yml:

exporters:
  otlp_http/canardstack:
    endpoint: http://host.docker.internal:4318
    headers:
      Authorization: Bearer dev-otlp-token-123456

service:
  pipelines:
    traces:
      exporters: [debug, span_metrics, otlp_http/canardstack]
    metrics:
      exporters: [debug, otlp_http/canardstack]
    logs:
      exporters: [debug, otlp_http/canardstack]

Start the full demo with its extras layer:

docker compose \
  -f compose.yaml \
  -f compose.full.yaml \
  -f compose.extras.yaml \
  up --force-recreate --remove-orphans --detach

Open the demo store and let the load generator run:

http://localhost:8080/

The demo collector exports OTLP/HTTP to duckdb-otlp on localhost:4318.

Flush buffered telemetry before opening Grafana:

docker compose exec ingest sh -c \
  "printf '%s\n' \"SELECT * FROM otlp_flush('otlp:0.0.0.0:4318');\" > /tmp/duckdb-otlp-ingest.sql"

Open the Dashboard

Grafana is provisioned with Prometheus, Loki, and Tempo datasources pointing at canardstack, plus a Canardstack OTel Demo dashboard:

http://localhost:3000/d/canardstack-otel-demo/canardstack-otel-demo

The dashboard includes OTel demo activity metrics, service memory, recent logs, and frontend trace search results.

Query Directly

You can also call the Loki API directly:

curl -sS -G http://localhost:9090/loki/api/v1/query_range \
  -H 'Authorization: Bearer dev-canardstack-key' \
  --data-urlencode 'query={}' \
  --data-urlencode "start=$(python3 -c 'import time; print(int(time.time()) - 900)')" \
  --data-urlencode "end=$(python3 -c 'import time; print(int(time.time()))')" \
  --data-urlencode 'limit=10'

You should receive a Loki success response once demo logs have been flushed into DuckLake.

Next

• Serve an existing DuckLake catalog shows the smallest server command.
• Connect Grafana shows the datasource settings.
• Storage schema reference lists the table contract.
• duckdb-otlp documentation covers writer setup and OTLP ingest details.