Suppose you need to import 5 tables from an external RDBMS (like MySQL) into Hadoop HDFS. Write the Sqoop command

Option A (import-all-tables with filter):

sqoop import-all-tables --connect jdbc:mysql://host:3306/mydb \
  --username user -P --warehouse-dir /user/hadoop/import \
  --exclude-tables audit_log,temp --num-mappers 4

Option B (table-by-table, recommended for prod):

for t in table1 table2 table3 table4 table5; do
  sqoop import --connect jdbc:mysql://host/mydb --table $t \
    --target-dir /user/hadoop/import/$t --num-mappers 4 \
    --split-by id
  # Use primary key or indexed column for split-by
done

Why Option B: Each table can have its own --split-by (indexed column), --num-mappers, and incremental logic. Failure in one doesn't block others.

Scalability Trade-offs: Parallelize 5 imports in Airflow; each gets 4 mappers. Source connection limit = 5 * 4 = 20; ensure MySQL max_connections > 20.

Cost Implications: Sequential vs. parallel affects wall-clock time and cluster cost. Compress with --compress to reduce HDFS storage.