Reviewed by Aditya Kumar · Last reviewed 2026-03-25
**Situation**: Multi-DB ingestion DAG runs sequentially; SLA at risk; DB teams report connection exhaustion. **Task**: Increase throughput without overloading source DBs or Airflow. **Action**: (1) **TaskGroup per DB**—isolate connection pools; `max_active_tasks_per_dag`...
This easy-level Spark/Big Data question appears frequently in data engineering interviews at companies like Dunnhumby. While less common, it tests deeper understanding that distinguishes strong candidates. Mastering the underlying concepts (airflow, sql) will help you answer variations of this question confidently.
Start by clearly defining the core concept being asked about. Interviewers want to see that you understand the fundamentals before diving into implementation details. Structure your answer with a definition, then explain the practical application with a concise example.
Situation: Multi-DB ingestion DAG runs sequentially; SLA at risk; DB teams report connection exhaustion.
Task: Increase throughput without overloading source DBs or Airflow.
Action: (1) TaskGroup per DB—isolate connection pools; max_active_tasks_per_dag limits concurrency per DB. (2) Dynamic task mapping (Airflow 2.3+): @task returns list; expand() creates N tasks. (3) Connection pooling—SQLAlchemy pool_size, max_overflow; one connection per worker, not per task. (4) CeleryExecutor—scale workers; separate queues for DB-heavy vs. light tasks. (5) Circuit breaker—pause DAG if DB error rate spikes.
Result: 3x parallelism, DB CPU within limits, SLA met.
Scalability Trade-offs: parallelism_factor = min(worker_count, DB_max_connections / tasks_per_DB). Over-parallelism causes connection timeouts.
Cost Implications: More workers = more cost; right-size from DB capacity and SLA.
Pro-Move: 'We use queue-based routing: db-heavy to high-mem workers, light to spot.' Red Flag: Cranking parallelism without DB connection limits—causes cascading failures.
Some links below are affiliate links. If you buy through them we may earn a small commission at no extra cost to you — it helps keep DataEngPrep free.
According to DataEngPrep.tech, this is one of the most frequently asked Spark/Big Data interview questions, reported at 1 company. DataEngPrep.tech maintains an editor-reviewed database of 1,863 data engineering interview questions across 7 categories.