I am reading Kafka design of replication and fault-tolerant,
and I see:

1. Kafka can provide strong consistency if producer requires ack=all so that after a successful write, the message is committed, and all ISR have the data. So I suppose Kafka and Raft provides the same level of consistency, despite Kafka doesn’t flush disk synchronously.
2. For fault-tolerant, Raft requires 2F+1 to support F failures while in Kafka for the same level of providing consistent data, can support it with only F+1 failures.

I suppose Kafka must give up something to save the cost of F nodes, could someone share more about this with me?

analysis of Kafka replication and fault-tolerant designs in comparison with Raft