Estou tentando simular a queda de primeiro broker(id=0), com o objetivo do segundo(id=2) assumir.
Quando executo o stop do primeiro o segundo desliga sozinho logo em seguida. A pasta do log.dirs foi criada separadamente para os 2 brokers, alguma outra coisa pode estar influenciando nessa dependência?
Estou usando o Kafka v2.13-3.5.1
Segue log apresentado no segundo broker antes do desligamento:
[2023-08-22 11:01:58,426] INFO [KafkaServer id=2] Retrying controlled shutdown (2 retries remaining) (kafka.server.KafkaServer)
[2023-08-22 11:01:58,427] INFO [KafkaServer id=2] Client requested connection close from node 0 (org.apache.kafka.clients.NetworkClient)
[2023-08-22 11:01:58,482] INFO [KafkaServer id=2] Controlled shutdown request returned after 53ms with 12 partitions remaining to move (kafka.server.KafkaServer)
[2023-08-22 11:02:03,482] INFO [KafkaServer id=2] Retrying controlled shutdown (1 retries remaining) (kafka.server.KafkaServer)
[2023-08-22 11:02:03,493] INFO [KafkaServer id=2] Controlled shutdown request returned after 11ms with 12 partitions remaining to move (kafka.server.KafkaServer)
[2023-08-22 11:02:03,495] WARN [KafkaServer id=2] Proceeding to do an unclean shutdown as all the controlled shutdown attempts failed (kafka.server.KafkaServer)
[2023-08-22 11:02:03,496] INFO [/config/changes-event-process-thread]: Shutting down (kafka.common.ZkNodeChangeNotificationListener$ChangeEventProcessThread)
[2023-08-22 11:02:03,496] INFO [/config/changes-event-process-thread]: Stopped (kafka.common.ZkNodeChangeNotificationListener$ChangeEventProcessThread)
[2023-08-22 11:02:03,497] INFO [/config/changes-event-process-thread]: Shutdown completed (kafka.common.ZkNodeChangeNotificationListener$ChangeEventProcessThread)
...