rabbitmq集群脑裂

[rabbitmq@hs-10-20-30-100 log]$ cat crash.log.0
2020-11-25 15:09:56 =ERROR REPORT====
** Node 'rabbitmq@hs-10-20-30-99' not responding **
** Removing (timedout) connection **
2020-11-25 15:09:56 =ERROR REPORT====
Mnesia('rabbitmq@hs-10-20-30-100'): ** ERROR ** mnesia_event got {inconsistent_database, running_partitioned_network, 'rabbitmq@hs-10-20-30-99'}
2020-11-25 15:09:57 =ERROR REPORT====
Mnesia('rabbitmq@hs-10-20-30-100'): ** ERROR ** mnesia_event got {inconsistent_database, starting_partitioned_network, 'rabbitmq@hs-10-20-30-99'}
2020-11-25 19:45:12 =ERROR REPORT====
** Node 'rabbitmq@hs-10-20-30-99' not responding **
** Removing (timedout) connection **
2020-11-25 19:45:12 =ERROR REPORT====
Mnesia('rabbitmq@hs-10-20-30-100'): ** ERROR ** mnesia_event got {inconsistent_database, running_partitioned_network, 'rabbitmq@hs-10-20-30-99'}
2020-11-25 19:45:14 =ERROR REPORT====
Mnesia('rabbitmq@hs-10-20-30-100'): ** ERROR ** mnesia_event got {inconsistent_database, starting_partitioned_network, 'rabbitmq@hs-10-20-30-99'}

出现上述日志说明系统出现脑裂了。

https://www.cnblogs.com/fswhq/p/13859578.html 

https://www.cnblogs.com/liyongsan/p/9640361.html 脑裂问题

rabbitmq部署，请参考最佳实践：https://www.cloudamqp.com/blog/rabbitmq-checklist-for-production-environments-a-complete-guide.html

https://www.rabbitmq.com/partitions.html

https://jack-vanlightly.com/blog/2018/9/10/how-to-lose-messages-on-a-rabbitmq-cluster

https://www.kubernetesquestions.com/questions/52390055

https://github.com/rabbitmq/rabbitmq-server/discussions/5787

https://www.tavisca.com/blog/handling-rabbitmq-network-partition-in-aws/

https://blog.csdn.net/u013256816/article/details/71097186 rabbitmq镜像模式和quorum模式。理论上ALL镜像模式性能低于quorum模式，一个是ALL，一个是n，只会小于，但实际好像并非如此。