69 real Flink & Streaming questions from the Big Data bank, as asked in Indian campus drives and tech interviews. Every question has a verified answer and an AI-tutor explanation on placd — free to start.
1. What is Watermark?
Junior
A.a condition where a slow downstream operator throttles upstream operators by slowing record flow through the pipeline
B.a marker in a stream that asserts event time has progressed past a timestamp, used to trigger event-time windows and handle lateness
C.a Flink state pattern that replicates a low-volume control stream (such as rules) to all parallel instances so it can join against a high-volume keyed stream
D.a Flink sink that ties external commits to checkpoints via prepare and commit phases to achieve end-to-end exactly-once delivery
2. Which term means: "a marker in a stream that asserts event time has progressed past a timestamp, used to trigger event-time windows and handle lateness"?
A.Watermark — a marker in a stream that asserts event time has progressed past a timestamp, used to trigger event-time windows and handle lateness
B.Watermark — the Flink model that processes each event individually for low latency, contrasted with Spark Structured Streaming's small batches
C.Watermark — the Complex Event Processing library that detects ordered patterns of events in a stream using a pattern API with conditions, quantifiers, and time constraints
D.Watermark — the Flink state backend storing keyed state in an embedded RocksDB instance on local disk, enabling state larger than memory with incremental checkpoints
A.the Flink feature decoupling checkpointing from state materialization by continuously logging state changes, reducing checkpoint duration and tail latency
B.a condition where a slow downstream operator throttles upstream operators by slowing record flow through the pipeline
C.the Flink 2.0 architecture that stores primary operator state on remote storage like S3 with local disk as cache for cloud-native elasticity
D.a periodic, consistent snapshot of all operator state in Flink that enables exactly-once recovery after a failure
A.the unified Flink connector (KafkaSource) supporting bounded and unbounded reads with offset initialization, watermark generation, and exactly-once via committed offsets
B.Flink state partitioned by key so each parallel instance maintains independent state for its assigned keys
C.a manually triggered, durable snapshot of Flink state used for upgrades, rescaling, and version migration
D.the distinction between when an event actually occurred and when it is processed, the former giving deterministic results despite delays
A.Savepoint — a Flink state pattern that replicates a low-volume control stream (such as rules) to all parallel instances so it can join against a high-volume keyed stream
B.Savepoint — the special marker Flink injects into streams so all operators snapshot their state at a consistent point for fault-tolerant recovery
C.Savepoint — a manually triggered, durable snapshot of Flink state used for upgrades, rescaling, and version migration
D.Savepoint — the unified Flink connector (KafkaSource) supporting bounded and unbounded reads with offset initialization, watermark generation, and exactly-once via committed offsets
A.a Flink mechanism using an OutputTag to emit selected records (such as late events or errors) to a secondary stream separate from the main output
B.the Flink feature decoupling checkpointing from state materialization by continuously logging state changes, reducing checkpoint duration and tail latency
C.a Flink state pattern that replicates a low-volume control stream (such as rules) to all parallel instances so it can join against a high-volume keyed stream
D.the distinction between when an event actually occurred and when it is processed, the former giving deterministic results despite delays
11. Which term means: "the distinction between when an event actually occurred and when it is processed, the former giving deterministic results despite delays"?
A.Event time vs processing time — a Flink mechanism using an OutputTag to emit selected records (such as late events or errors) to a secondary stream separate from the main output
B.Event time vs processing time — computation that retains information across events, such as running aggregates, kept in fault-tolerant operator state
C.Event time vs processing time — Flink state partitioned by key so each parallel instance maintains independent state for its assigned keys
D.Event time vs processing time — the distinction between when an event actually occurred and when it is processed, the former giving deterministic results despite delays
A.Keyed state — the choice between aligned barriers that deduplicate state versus faster unaligned or relaxed modes that may reprocess
B.Keyed state — a Flink marker flowing with the stream that asserts no events earlier than its timestamp should still arrive, driving event-time window firing
C.Keyed state — Flink state partitioned by key so each parallel instance maintains independent state for its assigned keys
D.Keyed state — the Flink 2.0 disaggregated state store built for asynchronous remote access and lightweight checkpoint and recovery
A.a manually triggered, durable snapshot of Flink state used for upgrades, rescaling, and version migration
B.a Flink sink that ties external commits to checkpoints via prepare and commit phases to achieve end-to-end exactly-once delivery
C.the Flink feature decoupling checkpointing from state materialization by continuously logging state changes, reducing checkpoint duration and tail latency
D.the Flink state backend storing keyed state in an embedded RocksDB instance on local disk, enabling state larger than memory with incremental checkpoints
17. Which term means: "a Flink sink that ties external commits to checkpoints via prepare and commit phases to achieve end-to-end exactly-once delivery"?
A.Two-phase commit sink — the special marker Flink injects into streams so all operators snapshot their state at a consistent point for fault-tolerant recovery
B.Two-phase commit sink — a Flink sink that ties external commits to checkpoints via prepare and commit phases to achieve end-to-end exactly-once delivery
C.Two-phase commit sink — the Flink operator that issues concurrent requests to external systems and processes responses out of order, hiding lookup latency without blocking the pipeline
D.Two-phase commit sink — a Flink marker flowing with the stream that asserts no events earlier than its timestamp should still arrive, driving event-time window firing
A.Flink state partitioned by key so each parallel instance maintains independent state for its assigned keys
B.the Complex Event Processing library that detects ordered patterns of events in a stream using a pattern API with conditions, quantifiers, and time constraints
C.the Flink 2.0 architecture that stores primary operator state on remote storage like S3 with local disk as cache for cloud-native elasticity
D.a Flink sink achieving end-to-end exactly-once delivery via a two-phase commit that aligns external commits with Flink checkpoints
20. Which term means: "the Flink 2.0 architecture that stores primary operator state on remote storage like S3 with local disk as cache for cloud-native elasticity"?
A.Disaggregated state — a manually triggered, durable snapshot of Flink state used for upgrades, rescaling, and version migration
B.Disaggregated state — a Flink marker flowing with the stream that asserts no events earlier than its timestamp should still arrive, driving event-time window firing
C.Disaggregated state — a Flink state pattern that replicates a low-volume control stream (such as rules) to all parallel instances so it can join against a high-volume keyed stream
D.Disaggregated state — the Flink 2.0 architecture that stores primary operator state on remote storage like S3 with local disk as cache for cloud-native elasticity
A.ForSt state backend — Flink state partitioned by key so each parallel instance maintains independent state for its assigned keys
B.ForSt state backend — the Flink feature decoupling checkpointing from state materialization by continuously logging state changes, reducing checkpoint duration and tail latency
C.ForSt state backend — a Flink mechanism using an OutputTag to emit selected records (such as late events or errors) to a secondary stream separate from the main output
D.ForSt state backend — the Flink 2.0 disaggregated state store built for asynchronous remote access and lightweight checkpoint and recovery
26. Which term means: "the Flink model that processes each event individually for low latency, contrasted with Spark Structured Streaming's small batches"?
A.True streaming vs micro-batch — computation that retains information across events, such as running aggregates, kept in fault-tolerant operator state
B.True streaming vs micro-batch — the Flink model that processes each event individually for low latency, contrasted with Spark Structured Streaming's small batches
C.True streaming vs micro-batch — the Complex Event Processing library that detects ordered patterns of events in a stream using a pattern API with conditions, quantifiers, and time constraints
D.True streaming vs micro-batch — a manually triggered, durable snapshot of Flink state used for upgrades, rescaling, and version migration
A.a condition where a slow downstream operator throttles upstream operators by slowing record flow through the pipeline
B.a Flink sink that ties external commits to checkpoints via prepare and commit phases to achieve end-to-end exactly-once delivery
C.the special marker Flink injects into streams so all operators snapshot their state at a consistent point for fault-tolerant recovery
D.the Flink state backend storing keyed state in an embedded RocksDB instance on local disk, enabling state larger than memory with incremental checkpoints
Showing 30 of 69 Flink & Streaming questions — the full set, with answers, explanations and an AI tutor on every question, is inside.
Free to start
Answers, AI explanations, and a scored voice mock interview
Sign up free to check your answers with explanations, ask the AI tutor anything on any question, and take one full AI mock interview — scored like a real panel.