启动后不久，RabbitMQ服务器关闭

我在保持RabbitMQ方面遇到了一些麻烦。

我通过提供的启动它/etc/init.d/rabbitmq-server start，它启动正常。status表示没事。

但是过了一会儿，服务器死了。status印刷

Error: unable to connect to node 'rabbit@myserver': nodedown

检查日志文件，看来我已经达到内存阈值。以下是日志：

    # start
=INFO REPORT==== 26-Mar-2014::03:24:13 ===
Limiting to approx 924 file handles (829 sockets)

=INFO REPORT==== 26-Mar-2014::03:24:13 ===
Memory limit set to 723MB of 1807MB total.

=INFO REPORT==== 26-Mar-2014::03:24:13 ===
Disk free limit set to 953MB

=INFO REPORT==== 26-Mar-2014::03:24:13 ===
Management plugin upgraded statistics to fine.

=INFO REPORT==== 26-Mar-2014::03:24:13 ===
msg_store_transient: using rabbit_msg_store_ets_index to provide index

=INFO REPORT==== 26-Mar-2014::03:24:13 ===
msg_store_persistent: using rabbit_msg_store_ets_index to provide index

=WARNING REPORT==== 26-Mar-2014::03:24:13 ===
msg_store_persistent: rebuilding indices from scratch

=INFO REPORT==== 26-Mar-2014::03:24:27 ===
started TCP Listener on [::]:5672

=INFO REPORT==== 26-Mar-2014::03:24:27 ===
Management agent started.

=INFO REPORT==== 26-Mar-2014::03:24:27 ===
Management plugin started. Port: 55672, path: /

=INFO REPORT==== 26-Mar-2014::03:24:39 ===
accepting AMQP connection <0.1999.0> (127.0.0.1:34788 -> 127.0.0.1:5672)

=WARNING REPORT==== 26-Mar-2014::03:24:40 ===
closing AMQP connection <0.1999.0> (127.0.0.1:34788 -> 127.0.0.1:5672):
connection_closed_abruptly

=INFO REPORT==== 26-Mar-2014::03:24:42 ===
accepting AMQP connection <0.2035.0> (127.0.0.1:34791 -> 127.0.0.1:5672)

=INFO REPORT==== 26-Mar-2014::03:24:46 ===
accepting AMQP connection <0.2072.0> (127.0.0.1:34792 -> 127.0.0.1:5672)

=INFO REPORT==== 26-Mar-2014::03:25:19 ===
vm_memory_high_watermark set. Memory used:768651448 allowed:758279372

=INFO REPORT==== 26-Mar-2014::03:25:19 ===
    alarm_handler: {set,{{resource_limit,memory,'rabbit@myserver'},
                         []}}

=INFO REPORT==== 26-Mar-2014::03:25:48 ===
Statistics database started.
    # server dies here

我似乎已经达到内存阈值，但是阅读文档，它不应该关闭服务器吗？只是在释放一些内存之前阻止发布？

是的，我知道我的芹菜工人是内存使用量的原因，我只是以为RabbitMQ可以正确处理它，而这似乎是文档暗示的。所以我做错了什么？

编辑：重构了我的任务，因此它的消息只是一个字符串（最多15个字符）。似乎没有什么不同。

我尝试启动RabbitMQ和celery worker --purge，没有任何事件触发任务，但是RabbitMQ的内存使用率似乎仍稳步上升到40％。然后不久后崩溃。它崩溃了，我的所有任务都没有运行的机会。