Answer: Checkpointing and idempotent sinks

Structured Streaming achieves end-to-end fault tolerance through **Checkpointing and Idempotent Sinks**. - **Checkpointing** saves the current state of the streaming query, including input stream offsets, allowing the query to restart from the last checkpoint after a failure, ensuring no data loss. - **Idempotent Sinks** can process the same data multiple times without duplicating results, ensuring consistency even if data is reprocessed due to failures or retries. Other options are incorrect because: - **Watermarking** manages late data and triggers micro-batch processing but doesn't directly aid in fault tolerance. - **Write-ahead logging**, while useful in some systems, isn't a primary fault tolerance mechanism in Structured Streaming. - **Failover to available nodes** is part of fault tolerance but doesn't alone ensure data consistency or prevent loss without checkpointing and idempotent sinks.

Author: LeetQuiz Editorial Team