Créer un graphe de propriétés multicloud sur un lakehouse ouvert et sans limites

Ce tutoriel vous explique comment créer un graphe BigQuery unique qui unifie les silos de données dans deux clouds différents à l'aide d'un lakehouse ouvert et sans limites, et de points de terminaison de catalogue REST Apache Iceberg sans déplacer les données.

Avant de commencer

Avant de commencer, configurez votre environnement et activez les API requises.

  1. Définissez votre projet et votre région, puis activez les API :

    export PROJECT_ID="your-gcp-project-id"
export REGION="us-east4"

gcloud config set project "$PROJECT_ID"

gcloud services enable \
  biglake.googleapis.com \
  bigquery.googleapis.com \
  secretmanager.googleapis.com \
  storage.googleapis.com

  2. Créez un environnement virtuel Python pour le chargeur :

    python3 -m venv iceberg-venv
source iceberg-venv/bin/activate
pip install --quiet "pyiceberg[pyarrow]"

Créer le Google Cloud spoke

Configurez un catalogue REST Apache Iceberg ouvert, soutenu par un bucket Cloud Storage, et chargez-y trois tables Iceberg.

  1. Créez le bucket et le catalogue :

    export GCS_BUCKET="gs://${PROJECT_ID}-xcloud-lake"
export GCS_CATALOG="gcs_lake"

gcloud storage buckets create "$GCS_BUCKET" \
  --project="$PROJECT_ID" \
  --location="$REGION"

gcloud biglake iceberg catalogs create "$GCS_CATALOG" \
  --project="$PROJECT_ID" \
  --catalog-type=biglake \
  --primary-location="$REGION" \
  --default-location="$GCS_BUCKET"

  2. Enregistrez le script Python suivant sous le nom load_gcs.py pour amorcer les tables :

    import subprocess, pyarrow as pa
from pyiceberg.catalog.rest import RestCatalog
from pyiceberg.schema import Schema
from pyiceberg.types import NestedField, StringType, LongType, DoubleType
import os

PROJECT = os.environ["PROJECT_ID"]
CATALOG = os.environ["GCS_CATALOG"]
TOKEN = subprocess.check_output(
    ["gcloud", "auth", "application-default", "print-access-token"], text=True
).strip()

cat = RestCatalog(
    name=CATALOG,
    uri="https://biglake.googleapis.com/iceberg/v1/restcatalog",
    warehouse=f"bl://projects/{PROJECT}/catalogs/{CATALOG}",
    token=TOKEN,
    **{"header.x-goog-user-project": PROJECT,
       "header.X-Iceberg-Access-Delegation": "vended-credentials"},
)

cat.create_namespace_if_not_exists("retail")

def mk(name, schema, table):
    ident = ("retail", name)
    try: cat.drop_table(ident)
    except Exception: pass
    t = cat.create_table(ident, schema=schema)
    t.append(table)
    print(f"  {name}: {table.num_rows} rows")

# customers
mk("customers",
   Schema(NestedField(1, "customer_id", StringType()),
          NestedField(2, "name", StringType()),
          NestedField(3, "region", StringType())),
   pa.table({
       "customer_id": ["C1", "C2", "C3", "C4", "C5", "C6"],
       "name": ["Ana", "Ben", "Cara", "Dan", "Eve", "Finn"],
       "region": ["west", "west", "east", "east", "west", "south"],
   }))

# orders
mk("orders",
   Schema(NestedField(1, "order_id", StringType()),
          NestedField(2, "customer_id", StringType()),
          NestedField(3, "status", StringType()),
          NestedField(4, "amount", DoubleType())),
   pa.table({
       "order_id": ["O1","O2","O3","O4","O5","O6","O7","O8","O9","O10"],
       "customer_id": ["C1","C1","C2","C3","C3","C4","C5","C5","C6","C2"],
       "status": ["shipped"]*8 + ["pending","shipped"],
       "amount": [156.0,89.0,120.0,147.0,89.0,199.0,25.0,88.0,80.0,224.0],
   }))

# order_items
mk("order_items",
   Schema(NestedField(1, "order_item_id", StringType()),
          NestedField(2, "order_id", StringType()),
          NestedField(3, "product_id", StringType()),
          NestedField(4, "quantity", LongType()),
          NestedField(5, "amount", DoubleType())),
   pa.table({
       "order_item_id": [f"OI{i}" for i in range(1, 16)],
       "order_id":  ["O1","O1","O2","O3","O3","O4","O5","O6","O6","O7","O8","O9","O10","O10","O2"],
       "product_id": ["P1","P2","P3","P1","P5","P4","P3","P6","P8","P7","P1","P2","P4","P5","P6"],
       "quantity":  [1,2,1,1,3,1,1,1,