Rabbitmq normal cluster mode, accessed through VIP, the node goes offline and goes online again, and the original consumers disappear #9528
-
Describe the bugRabbitmq normal cluster mode, accessed through VIP, the node goes offline and goes online again, and the original consumers disappear Reproduction steps1、Rabbitmq Normal Cluster Deployment Mode Expected behaviorThe consumer of node A exists and can continue to consume messages Additional contextNo response |
Beta Was this translation helpful? Give feedback.
Replies: 3 comments 4 replies
-
It's not clear what you mean here. Of course the connections are dropped when the node goes down. It's the client's responsibility to reconnect. Please clarify. Probably unrelated but 2-node clusters are a bad idea: https://www.rabbitmq.com/partitions.html#pause-minority, |
Beta Was this translation helpful? Give feedback.
-
The current environment is: two nodes are deployed in the same cluster, and RabbitMQ nodes after accessing the cluster through VIP (Keepalived+HAProxy). Take one of the nodes offline, and for the client, the connection should not be disconnected |
Beta Was this translation helpful? Give feedback.
-
In normal cluster mode, will brokers within the cluster synchronize with consumers subscribing to the queue? If not synchronized, in VIP access mode: does it mean that after the node is offline, the original consumer channel will not be aware of the broker change, resulting in no consumers in the broker's view |
Beta Was this translation helpful? Give feedback.
It's a responsibility of the client to reconnect. RabbitMQ does not connect to clients, clients connect (and reconnect as needed) to RabbitMQ. Some client libraries can do this automatically if configured and if you accept their recovery strategy as suitable.
Clients also share their part of responsibility for using publisher and consumer-side data safety features.