How does the high watermark bound a consumer's reads?

How does the high watermark bound a consumer's reads?

• A. It bounds reads to the end offset of the log
• B. It bounds reads to the high watermark offset
• C. It bounds reads to the oldest offset
• D. It bounds reads to the next offset to produce

Answer: (B)

The high watermark defines the boundary of data that is safely readable. In a partition, it represents the highest offset that all in-sync replicas have acknowledged, so data up to this point is considered committed and durable. When a consumer fetches messages, it can only read up to the high watermark; messages beyond it may exist but are not yet replicated to all in-sync replicas and are not exposed to readers. This guarantees that consumers see only data that is replicated and committed, avoiding uncommitted or potentially lost data. The oldest offset or the absolute end of the log are not the bounds used for reads, and the next offset to produce is a producer concept, not the consumer read boundary.

The high watermark defines the boundary of data that is safely readable. In a partition, it represents the highest offset that all in-sync replicas have acknowledged, so data up to this point is considered committed and durable. When a consumer fetches messages, it can only read up to the high watermark; messages beyond it may exist but are not yet replicated to all in-sync replicas and are not exposed to readers. This guarantees that consumers see only data that is replicated and committed, avoiding uncommitted or potentially lost data. The oldest offset or the absolute end of the log are not the bounds used for reads, and the next offset to produce is a producer concept, not the consumer read boundary.