What contains the offset of the last message copied to all replicas?

• A. High watermark
• B. Low watermark
• C. End offset
• D. Log start offset

Answer: (A)

The key idea is understanding what marks the point where messages are considered safe and visible to consumers due to full replication. In Kafka, the high watermark is defined as the offset of the last message that has been replicated to all in-sync replicas. This boundary determines which messages are guaranteed to be stored everywhere that can serve reads, and thus what consumers can reliably read.

As the followers in the in-sync set catch up to the leader, the high watermark advances. It can lag behind the log’s end offset, which is simply the offset of the next message to be appended, and it is not a guarantee of replication to all replicas. The log start offset is just the first offset in the log, not related to replication completeness. The concept of a low watermark isn’t the standard term used here in Kafka’s replication model.

So, the offset of the last message copied to all replicas is exactly what the high watermark represents, making it the best answer. For example, if the log has offsets up to 9 but only up to offset 7 is replicated on all in-sync replicas, the high watermark is 7, and messages up to 7 are considered committed and readable by consumers.