https://4f0a9a3b.wangpeng.pages.dev/en/tags/exactly-once/Recent content in Exactly-Once on Zhuifeng NotesHugoenTue, 25 Aug 2026 00:00:00 +0000https://4f0a9a3b.wangpeng.pages.dev/en/tech/kafka/reliability/Tue, 25 Aug 2026 00:00:00 +0000https://4f0a9a3b.wangpeng.pages.dev/en/tech/kafka/reliability/<p>Kafka&rsquo;s reliability requires coordinated configuration across the producer, broker and consumer. We cover it through three goals: no loss, no duplication, ordered.</p> <h2 id="producer-side-reliability">Producer-Side Reliability </h2> <p>Control write durability level via <code>acks</code>:</p> <ul> <li><code>acks=0</code>: considered successful as soon as sent; may lose messages; highest throughput.</li> <li><code>acks=1</code>: success once the leader writes; may lose if the leader crashes.</li> <li><code>acks=all</code>: success only after all ISR replicas sync; safest, combined with <code>min.insync.replicas</code> to avoid single points.</li> </ul> <p>Enabling retries (<code>retries</code>) handles transient failures but introduces <strong>duplicates</strong>: network jitter causes the producer to retry without receiving the ack, writing the same message twice.</p>