broker.id重复、log.dirs路径不存在)、Zookeeper未启动、端口冲突、权限不足。server.properties中broker.id唯一性、log.dirs路径有效性(需提前创建目录并赋予权限)。systemctl status zookeeper,未启动则执行systemctl start zookeeper。netstat -tuln | grep 9092,修改listeners配置避免冲突。Connection refused或TimeoutException。advertised.listeners配置错误、网络分区。sudo ufw allow 9092(或对应Kafka端口)。advertised.listeners是否为集群节点实际IP或域名(如PLAINTEXT://node1:9092)。ping或telnet测试节点间网络连通性。num.partitions(主题分区数)提升并行度。compression.type=gzip)并增大batch.size/linger.ms参数。OutOfMemoryError或磁盘写满。buffer.memory配置不合理、日志保留策略不当。kafka-server-start.sh中KAFKA_HEAP_OPTS,建议设置为可用内存的50%-70%(如-Xms4G -Xmx4G)。log.retention.hours(如设置为168小时)和log.retention.bytes控制日志保留量。log.dirs所在磁盘有足够空间。ZooKeeper connection lost或Session expired。zookeeper.connect配置错误、会话超时。systemctl status zookeeper,确保正常运行。zookeeper.connect参数为正确的Zookeeper地址(如localhost:2181)。zookeeper.session.timeout.ms参数(默认10秒,可根据网络延迟调整)。Permission denied或NoSuchFileException。log.dirs、dataDir等路径存在且为绝对路径,使用正斜杠(/)或双反斜杠(\\)。chmod -R 777 /path/to/kafka/logs。inter.broker.protocol.version参数。tail -f /path/to/kafka/logs/server.log定位错误详情。Kafka Manager或Prometheus+Grafana监控集群状态、内存和磁盘使用情况。参考来源:[1,2,3,4,5,6,7,8,9,10,11]