Sqoop for database imports in Hadoop

Introduction

Sqoop helps move data from databases into Hadoop easily. It saves time and avoids manual copying.

You want to analyze data stored in a MySQL or Oracle database using Hadoop tools.

You need to import large tables from a database into HDFS for big data processing.

You want to automate regular data transfers from a database to Hadoop.

You want to combine data from databases with other big data sources in Hadoop.

You want to export processed data back from Hadoop to a database.

Syntax

Hadoop

sqoop import \
  --connect jdbc:mysql://hostname:port/dbname \
  --username user \
  --password pass \
  --table tablename \
  --target-dir /hdfs/path \
  [other options]

Use --connect to specify the database connection string.

--table tells Sqoop which table to import.

Examples

Imports the customers table from MySQL database salesdb into HDFS directory /user/hadoop/customers.

Hadoop

sqoop import --connect jdbc:mysql://localhost:3306/salesdb --username admin --password secret --table customers --target-dir /user/hadoop/customers

Imports transactions table from PostgreSQL, splitting import by transaction_id for parallelism.

Hadoop

sqoop import --connect jdbc:postgresql://dbserver:5432/finance --username finance_user --password pass123 --table transactions --split-by transaction_id --target-dir /data/transactions

Imports only sales department employees using a custom query. $CONDITIONS is required for splitting.

Hadoop

sqoop import --connect jdbc:mysql://localhost:3306/hrdb --username hr --password hrpass --query 'SELECT * FROM employees WHERE department = "Sales" AND $CONDITIONS' --split-by employee_id --target-dir /user/hadoop/sales_employees

Sample Program

This command imports the employees table from the MySQL database testdb into the HDFS directory /user/hadoop/employees. It uses 2 mappers to speed up the import.

Hadoop

sqoop import \
  --connect jdbc:mysql://localhost:3306/testdb \
  --username testuser \
  --password testpass \
  --table employees \
  --target-dir /user/hadoop/employees \
  --num-mappers 2

Important Notes

Always use $CONDITIONS in custom queries to allow Sqoop to split the data for parallel import.

Use --num-mappers to control parallelism; more mappers can speed up import but use more resources.

Passwords can be passed securely using --password-file instead of plain text.

Summary

Sqoop imports data from databases into Hadoop automatically.

Use connection string, table name, and target directory to specify import.

Parallel imports speed up data transfer using split columns and mappers.

Practice

1. What is the main purpose of Sqoop in Hadoop?

easy

A. To process data inside Hadoop using MapReduce

B. To export data from Hadoop to relational databases

C. To import data from relational databases into Hadoop

D. To visualize data stored in Hadoop

Solution

1. Step 1: Understand Sqoop's role

   Sqoop is designed to transfer data between relational databases and Hadoop.

2. Step 2: Identify import function

   Its main use is importing data from databases into Hadoop's storage.

3. Final Answer:

   To import data from relational databases into Hadoop -> Option C

4. Quick Check:

   Sqoop imports data = A [OK]

Hint: Remember Sqoop imports data into Hadoop from databases [OK]

Common Mistakes:

• Confusing import with export
• Thinking Sqoop processes data inside Hadoop
• Assuming Sqoop is for visualization

2. Which of the following is the correct basic Sqoop import command syntax?

easy

A. sqoop export --connect jdbc:mysql://localhost/db --table mytable --target-dir /user/hadoop/mytable

B. sqoop import --connect jdbc:mysql://localhost/db --table mytable --target-dir /user/hadoop/mytable

C. sqoop import --table mytable --connect jdbc:mysql://localhost/db --output-dir /user/hadoop/mytable

D. sqoop import --connect jdbc:mysql://localhost/db --table mytable --output-dir /user/hadoop/mytable

Solution

1. Step 1: Identify correct command type

   Import commands use 'sqoop import' with connection, table, and target directory.

2. Step 2: Check option correctness

   sqoop import --connect jdbc:mysql://localhost/db --table mytable --target-dir /user/hadoop/mytable uses correct flags: --connect, --table, --target-dir.

3. Final Answer:

   sqoop import --connect jdbc:mysql://localhost/db --table mytable --target-dir /user/hadoop/mytable -> Option B

4. Quick Check:

   Correct import syntax = A [OK]

Hint: Use --target-dir for import destination, not --output-dir [OK]

Common Mistakes:

• Using 'export' instead of 'import'
• Mixing --output-dir with import command
• Incorrect order of parameters

3. What will be the effect of running this command?

sqoop import --connect jdbc:mysql://localhost/db --table employees --split-by emp_id --num-mappers 4 --target-dir /user/hadoop/employees

medium

A. Imports employees table using 1 mapper ignoring split-by

B. Fails because --split-by cannot be used with --num-mappers

C. Exports employees table to MySQL using 4 mappers

D. Imports employees table using 4 parallel mappers splitting by emp_id

Solution

1. Step 1: Analyze command options

   The command imports data from 'employees' table using 4 mappers and splits data by 'emp_id'.

2. Step 2: Understand parallel import

   Using --split-by with --num-mappers enables parallel import for faster data transfer.

3. Final Answer:

   Imports employees table using 4 parallel mappers splitting by emp_id -> Option D

4. Quick Check:

   --split-by + --num-mappers = parallel import [OK]

Hint: Use --split-by column to enable parallel import with multiple mappers [OK]

Common Mistakes:

• Thinking --split-by disables parallelism
• Confusing import with export
• Assuming default 1 mapper always used

4. You run this Sqoop import command but get an error:

sqoop import --connect jdbc:mysql://localhost/db --table sales --num-mappers 4 --target-dir /user/hadoop/sales

What is the most likely cause?

medium

A. Missing --split-by column for parallel import

B. Incorrect --target-dir path format

C. Using --num-mappers with import is invalid

D. Table 'sales' does not exist in database

Solution

1. Step 1: Check command parameters

   The command uses --num-mappers 4 but does not specify --split-by column.

2. Step 2: Understand parallel import requirements

   Sqoop requires a --split-by column to divide data for multiple mappers; missing it causes error.

3. Final Answer:

   Missing --split-by column for parallel import -> Option A

4. Quick Check:

   Parallel import needs --split-by column [OK]

Hint: Always specify --split-by when using multiple mappers [OK]

Common Mistakes:

• Assuming --num-mappers works without --split-by
• Thinking --target-dir format causes error
• Ignoring table existence check

5. You want to import a large 'orders' table from MySQL into Hadoop quickly. The table has a numeric primary key 'order_id'. Which Sqoop command will best speed up the import?

hard

A. sqoop import --connect jdbc:mysql://localhost/db --table orders --split-by order_id --num-mappers 8 --target-dir /user/hadoop/orders

B. sqoop import --connect jdbc:mysql://localhost/db --table orders --split-by order_date --num-mappers 8 --target-dir /user/hadoop/orders

C. sqoop import --connect jdbc:mysql://localhost/db --table orders --num-mappers 8 --target-dir /user/hadoop/orders

D. sqoop import --connect jdbc:mysql://localhost/db --table orders --target-dir /user/hadoop/orders

Solution

1. Step 1: Identify best split column

   Use the numeric primary key 'order_id' for --split-by to evenly divide data.

2. Step 2: Use multiple mappers for parallel import

   Setting --num-mappers 8 enables 8 parallel tasks to speed up import.

3. Step 3: Evaluate other options

   sqoop import --connect jdbc:mysql://localhost/db --table orders --target-dir /user/hadoop/orders uses single mapper (slow). sqoop import --connect jdbc:mysql://localhost/db --table orders --num-mappers 8 --target-dir /user/hadoop/orders misses --split-by (error). sqoop import --connect jdbc:mysql://localhost/db --table orders --split-by order_date --num-mappers 8 --target-dir /user/hadoop/orders splits by 'order_date' which may not be numeric or evenly distributed.

4. Final Answer:

   sqoop import --connect jdbc:mysql://localhost/db --table orders --split-by order_id --num-mappers 8 --target-dir /user/hadoop/orders -> Option A

5. Quick Check:

   Use numeric PK with --split-by and multiple mappers [OK]

Hint: Split by numeric primary key and increase mappers for speed [OK]

Common Mistakes:

• Not specifying --split-by with multiple mappers
• Choosing non-numeric or uneven split column
• Using single mapper for large tables