What is the ISR (In-Sync Replicas) set?

The leader of each partition keeps track of the set of replicas that are in sync with it; this set is the ISR, and the leader is always part of it. A follower stays in the ISR while it keeps an active session with the controller and is not too far behind: the leader removes a follower that has not sent fetch requests, or has not caught up to the leader's log end offset, for replica.lag.time.max.ms (default 30000 ms, i.e. 30 seconds). A message is considered committed once all replicas in the ISR have applied it, and only committed messages are given to consumers. Only members of the ISR are eligible for election as the new leader, which is how the new leader has all committed data; by default Kafka favors waiting for a consistent replica over an unclean election. On newer clusters with Eligible Leader Replicas the choice is not limited strictly to the ISR.

Expected key concepts: in sync, ISR, replica.lag.time.max.ms, 30000, leader, follower, committed, election, eligible, fetch requests