gcp-pipeline-ref-orchestration 1.0.8

GCP Pipeline Reference: Generic Orchestration - Airflow DAGs for Cloud Composer (JOIN + MAP patterns)

Generic Orchestration

Unit 3 of Generic 3-Unit Deployment

Airflow DAGs for pipeline coordination and scheduling.

---

Flow Diagram

                         Generic ORCHESTRATION FLOW
                         ─────────────────────

  Pub/Sub                    Airflow DAGs                    External
  ───────                    ────────────                    ────────

  .ok file     ┌─────────────────────────────────────────────────────────┐
  arrives      │                                                         │
  │        │  ┌──────────────┐                                       │
  └───────►│  │ PubSub       │                                       │
               │  │ Sensor       │                                       │
               │  └──────┬───────┘                                       │
               │         │                                               │
               │         ▼                                               │
               │  ┌──────────────┐    ┌──────────────┐    ┌───────────┐ │
               │  │ Trigger      │───►│ Dataflow     │───►│ Wait for  │ │
               │  │ Ingestion    │    │ (generic-ingest)  │    │ Complete  │ │
               │  └──────────────┘    └──────────────┘    └─────┬─────┘ │
               │                                                │       │
               │  ┌─────────────────────────────────────────────┘       │
               │  │                                                     │
               │  ▼                                                     │
               │  ┌──────────────┐                                      │
               │  │ Entity       │  Waits for ALL 3 entities:           │
               │  │ Dependency   │  - customers ✓                       │
               │  │ Checker      │  - accounts  ✓                       │
               │  └──────┬───────┘  - decision  ✓                       │
               │         │                                              │
               │         ▼ (all ready)                                  │
               │  ┌──────────────┐    ┌──────────────┐                  │
               │  │ Trigger      │───►│ dbt run      │                  │
               │  │ dbt          │    │ (transform)  │                  │
               │  └──────────────┘    └──────────────┘                  │
               │                                                        │
               └────────────────────────────────────────────────────────┘

---

Pattern

MULTI-TARGET: Orchestrates ingestion of 3 entities → waits for all → triggers transformation to 2 FDP targets

Step	Description
1	Pub/Sub sensor detects .ok file
2	Triggers Dataflow ingestion job
3	EntityDependencyChecker waits for all 3 entities
4	When all ready, triggers dbt transformation to event_transaction_excess and portfolio_account_excess

---

Library-Driven Ease of Use

The Generic orchestration unit leverages the gcp-pipeline-orchestration library to coordinate a complex multi-entity arrival pattern:

1. Event-Driven Triggering: Uses BasePubSubPullSensor with built-in .ok file filtering and metadata extraction to XCom.
2. Cross-Entity Coordination: Uses EntityDependencyChecker to verify that all 3 entities (Customers, Accounts, Decision) are loaded before triggering the FDP transformation.
3. Local Development: All DAGs can be parsed and validated without a live Airflow environment thanks to the library's AIRFLOW_AVAILABLE stubbing mechanism.

---

How to Replicate this JOIN Orchestration

To create a new orchestration unit for a multi-entity system, follow the Creating New Deployment Guide and use the standardized DAG Templates.

Key steps for this JOIN pattern:

1. Pub/Sub Sensor: Configure BasePubSubPullSensor for your system's notification topic.
2. Job Control: Use JobControlRepository from core to track the state of each entity load.
3. Dependency Check: Initialize EntityDependencyChecker with your list of REQUIRED_ENTITIES.
4. DAG Triggering: Use TriggerDagRunOperator to chain the Ingestion and Transformation units.

---

Infrastructure & Configurations

Google Cloud Resources

This deployment requires the following GCP infrastructure, provisioned via Terraform:

• Orchestration: Cloud Composer (Managed Apache Airflow).
• Messaging: Pub/Sub Topic generic-file-notifications and Subscription generic-file-notifications-sub.
• Identity & Access: Service Account with roles for Dataflow, BigQuery, and GCS.

For detailed infrastructure definitions, see infrastructure/terraform/systems/generic/orchestration/.

Airflow Configuration

The DAGs use several Airflow variables and connections:

Type	Name	Description	Default / Source
Variable	gcp_project_id	Target GCP Project ID	GCP_PROJECT_ID env var
Variable	gcp_region	GCP Region for Dataflow	europe-west2
Variable	generic_pubsub_subscription	Pub/Sub subscription for file alerts	generic-file-notifications-sub
Variable	generic_landing_bucket	GCS bucket for landing files	<project>-generic-dev-landing
Variable	generic_error_bucket	GCS bucket for error files	<project>-generic-dev-error
Connection	google_cloud_default	Connection for GCP resources	-
Connection	bigquery_default	Connection for BigQuery	-

Technology Stack & Documentation

• Google Cloud Composer - Managed Apache Airflow
• Apache Airflow - Workflow orchestration
• Airflow Google Cloud Operators - GCP integration
• Cloud Pub/Sub - Messaging service for event triggers
• Airflow Cross-DAG Dependencies - Orchestrating complex flows

---

DAGs

DAG	Purpose
generic_pubsub_trigger_dag.py	Triggered by Pub/Sub on .ok file arrival
generic_odp_load_dag.py	Runs Dataflow for ODP load and checks entity dependencies
generic_fdp_transform_dag.py	Runs dbt for FDP transformation
generic_error_handling_dag.py	Error handling and DLQ

---

Dependencies

Library	Purpose
gcp-pipeline-core	Audit, logging, error handling
gcp-pipeline-orchestration	DAG factory, sensors, operators

NO Apache Beam dependency - ingestion is separate unit.

---

Execution & Testing

1. Local DAG Validation

You can validate DAG syntax locally:

# Setup venv
./scripts/setup_deployment_venv.sh data-pipeline-orchestrator
source deployments/data-pipeline-orchestrator/venv/bin/activate

# Validate syntax
python dags/data_ingestion_dag.py

2. Testing End-to-End Flow

Use the simulation script to trigger the full Generic flow:

./scripts/gcp/06_test_pipeline.sh generic

3. Manual Pub/Sub Trigger

Alternatively, you can manually publish a notification to trigger the DAG:

gcloud pubsub topics publish generic-file-notifications \
    --message='{"name": "generic/customers/customers_20260101.ok", "bucket": "my-landing-bucket"}'

4. Deployment to Composer

Deploy the DAGs to your Cloud Composer environment:

gsutil cp dags/*.py gs://<composer-bucket>/dags/