How to Cut RocketMQ Consumer Latency from 10 s to 500 ms with Five Overlooked Settings
During a traffic peak the order service saw 30‑second consumer latency and a backlog of over 20 000 messages; by adjusting five RocketMQ consumer parameters—pullBatchSize, consumeThreadMin/Max, adjustThreadPoolNumsThreshold, pullTimeoutMillis, socketTimeoutMillis, and consumeMessageBatchMaxSize—the team reduced end‑to‑end latency to 500 ms, raised TPS to over 2500, and cleared the backlog.
