[rabbitmq-discuss] rabbit servers block in cluster environment

[Liwen Fan]

this situation also occurs when 2 servers are configured both disk.
and, if the client only connects to 1 server (2 servers are clustered), no
blocking is detected.
is that the recommended way?


2010/11/19 Liwen Fan <f at scalethink.com>

> Hi
>   we face some blocking problems in rabbitmq cluster environment.
>   we have some java test clients and 2 rabbit servers clustered, one ram
> and the other disk, all of them are 2.1.1.
>   the client creates two connections when it is initialized, and holds
> these two connections. For each action as basicGet/basicPublish, it randomly
> chooses one connection.
>   the queues and exchanges are created on one server (random choose) when
> the client starts up.
>   then the client creates many threads. Each thread randomly chooses one
> connection and randomly chooses several exchanges to publish and several
> queues to get.
>
>   but after several message published/got, both servers face blocking
> problems. The test client blocks and the server command like "rabbitmqctl
> list_queues" and "rabbitmqctl stop" blocks too.
>   but "rabbitmqctl status" does not block and returns ok.
>
>   when we move to 1 single server(disk), everything turns fine. Rabbit
> server behaves quite well as a single node.
>
>   by the way, we are running on erlang R14B. Both queues and messages are
> persistent and durable.
>   Any suggestions?
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20101119/3c420dce/attachment.htm>