Data Fusion for ETL in GCP - Step-by-Step Execution

Process Flow - Data Fusion for ETL

Start Data Fusion Pipeline

↓

Extract Data from Source

↓

Transform Data

↓

Load Data to Destination

↓

Pipeline Completes Successfully

This flow shows how Data Fusion extracts data, transforms it, and loads it to the target system step-by-step.

Execution Sample

GCP

1. Create pipeline
2. Add source plugin
3. Add transform plugin
4. Add sink plugin
5. Run pipeline

This pipeline extracts data from a source, applies transformations, and loads it to a destination.

Process Table

Step	Action	Component	Data State	Result
1	Create pipeline	Pipeline	No data	Pipeline created and ready
2	Add source plugin	Source	No data	Source configured to read data
3	Add transform plugin	Transform	Raw data	Data transformation logic applied
4	Add sink plugin	Sink	Transformed data	Sink configured to write data
5	Run pipeline	Pipeline	Data flows	Data extracted, transformed, and loaded
6	Pipeline completes	Pipeline	Data loaded	Pipeline run successful

💡 Pipeline run completes after data is loaded to the destination

Status Tracker

Variable	Start	After Step 2	After Step 3	After Step 4	After Step 5	Final
Data	None	Raw data extracted	Data transformed	Transformed data ready	Data loaded to sink	Pipeline complete

Key Moments - 2 Insights

Why does the data state change from 'Raw data' to 'Data transformed' at Step 3?

At Step 3, the transform plugin applies changes to the raw data, modifying it as per the pipeline logic, as shown in the execution_table row 3.

What happens if the sink plugin is not configured before running the pipeline?

Without the sink plugin configured (Step 4), the pipeline cannot load data to the destination, so the run at Step 5 would fail or not complete properly.

Visual Quiz - 3 Questions

Look at the execution_table, what is the data state after Step 3?

A Data transformed

B Raw data

C Data loaded

D No data

💡 Hint

Check the 'Data State' column for Step 3 in the execution_table.

At which step does the pipeline start running and data flows through components?

A Step 2

B Step 3

C Step 5

D Step 6

💡 Hint

Look for the 'Run pipeline' action in the execution_table.

If the transform plugin is removed, how would the data state after Step 4 change?

A Data would be loaded transformed

B Data would remain raw

C Data would be transformed

D No data would flow

💡 Hint

Refer to variable_tracker and execution_table rows for transform plugin effects.

Concept Snapshot

Data Fusion ETL pipeline:
1. Extract data from source
2. Transform data as needed
3. Load data to destination
Configure source, transform, sink plugins
Run pipeline to process data end-to-end

Full Transcript

Data Fusion for ETL involves creating a pipeline that extracts data from a source, applies transformations, and loads it to a destination. The pipeline is built by adding source, transform, and sink plugins. When the pipeline runs, data flows through these components step-by-step, changing state from raw to transformed to loaded. Proper configuration of each plugin is essential for successful execution.

Practice

1. What is the main purpose of Google Cloud Data Fusion in ETL processes?

easy

A. To store large amounts of raw data permanently

B. To write complex SQL queries for data analysis

C. To visually build and automate data pipelines without coding

D. To create virtual machines for data processing

Solution

1. Step 1: Understand Data Fusion's role

   Data Fusion is designed to help users build data pipelines visually, making ETL easier without needing to write code.

2. Step 2: Compare options

   Options A, B, and D describe other tools or services, not Data Fusion's main purpose.

3. Final Answer:

   To visually build and automate data pipelines without coding -> Option C

4. Quick Check:

   Data Fusion = Visual ETL pipelines [OK]

Hint: Data Fusion is about visual pipelines, not coding or storage [OK]

Common Mistakes:

• Confusing Data Fusion with BigQuery for SQL queries
• Thinking Data Fusion stores data permanently
• Assuming Data Fusion manages virtual machines

2. Which of the following is the correct way to add a source plugin in a Data Fusion pipeline?

easy

A. Drag the source plugin from the palette and drop it onto the canvas

B. Write a SQL query to define the source

C. Create a virtual machine and install the source plugin manually

D. Use the command line to start the source plugin

Solution

1. Step 1: Recall Data Fusion UI usage

   Data Fusion uses a visual interface where you drag and drop plugins to build pipelines.

2. Step 2: Eliminate incorrect methods

   Options A, C, and D involve coding, manual installs, or command line, which are not how Data Fusion works for adding plugins.

3. Final Answer:

   Drag the source plugin from the palette and drop it onto the canvas -> Option A

4. Quick Check:

   Data Fusion = Drag & drop plugins [OK]

Hint: Data Fusion uses drag-and-drop for plugins, not code [OK]

Common Mistakes:

• Trying to write SQL to add plugins
• Thinking manual VM setup is needed
• Using command line instead of UI

3. Given a Data Fusion pipeline with a source plugin reading from Cloud Storage and a sink plugin writing to BigQuery, what happens when the pipeline runs successfully?

medium

A. The pipeline deletes data in BigQuery

B. Data is moved from Cloud Storage to BigQuery after any transformations

C. The pipeline creates a new Cloud Storage bucket

D. Data is copied from BigQuery to Cloud Storage

Solution

1. Step 1: Understand source and sink roles

   The source plugin reads data from Cloud Storage, and the sink plugin writes data to BigQuery.

2. Step 2: Identify pipeline behavior

   When the pipeline runs, data flows from source to sink, optionally transformed in between.

3. Final Answer:

   Data is moved from Cloud Storage to BigQuery after any transformations -> Option B

4. Quick Check:

   Source to sink = Cloud Storage to BigQuery [OK]

Hint: Source reads, sink writes; data flows forward [OK]

Common Mistakes:

• Confusing source and sink directions
• Thinking pipeline creates buckets automatically
• Assuming pipeline deletes data

4. You created a Data Fusion pipeline but it fails to start. Which of the following is the most likely cause?

medium

A. The pipeline has too many source plugins

B. The Cloud Storage bucket is empty

C. You did not write any SQL queries in the pipeline

D. The sink plugin is missing or not configured properly

Solution

1. Step 1: Check pipeline requirements

   Every Data Fusion pipeline needs a properly configured sink plugin to write data.

2. Step 2: Analyze failure reasons

   If the sink is missing or misconfigured, the pipeline cannot complete and will fail to start.

3. Final Answer:

   The sink plugin is missing or not configured properly -> Option D

4. Quick Check:

   Missing sink = pipeline start failure [OK]

Hint: Always configure sink plugin to start pipeline [OK]

Common Mistakes:

• Assuming empty bucket causes start failure
• Thinking SQL queries are mandatory
• Believing multiple sources cause failure

5. You want to build a Data Fusion pipeline that reads sales data from multiple Cloud Storage folders, cleans the data by removing duplicates, and writes the result to BigQuery daily. Which approach best fits this requirement?

hard

A. Use multiple source plugins for each folder, a Wrangler plugin to clean duplicates, and a BigQuery sink; schedule pipeline daily

B. Write a custom script outside Data Fusion to clean data, then use Data Fusion to load into BigQuery

C. Use a single source plugin pointing to one folder and manually merge data later

D. Create a pipeline with only a BigQuery source and sink without cleaning

Solution

1. Step 1: Identify data sources and cleaning needs

   Multiple folders require multiple source plugins or a pattern that reads all folders; cleaning duplicates needs a Wrangler or transform plugin.

2. Step 2: Design pipeline and scheduling

   Connect sources to cleaning plugin, then to BigQuery sink; schedule pipeline to run daily for automation.

3. Final Answer:

   Use multiple source plugins for each folder, a Wrangler plugin to clean duplicates, and a BigQuery sink; schedule pipeline daily -> Option A

4. Quick Check:

   Multiple sources + cleaning + sink + schedule = Use multiple source plugins for each folder, a Wrangler plugin to clean duplicates, and a BigQuery sink; schedule pipeline daily [OK]

Hint: Combine sources, cleaning plugin, sink, and schedule for ETL [OK]

Common Mistakes:

• Skipping cleaning step inside Data Fusion
• Using external scripts unnecessarily
• Ignoring scheduling for daily runs