Question 1054

Two Out of Three

A topic has replication factor 3 with min.insync.replicas=2, and producers use acks=all. During a rolling restart one broker is already down when a second broker holding replicas for the same partitions fails. Producers start getting errors. What is happening?

A

Consumers have read past the partition's high watermark, so the leader blocks further appends until the group commits those offsets

B

The ISR has fallen to one replica, below min.insync.replicas, so the leader refuses acks=all writes rather than accept them unsafely

C

acks=all requires an acknowledgement from all three assigned replicas, so writes actually stopped when the first broker went down and are only now being noticed

D

The controller promoted a lagging out-of-sync replica to leader, and that replica rejects producer writes until it has caught up to the old high watermark