[rabbitmq-discuss] 0-8 AMQP Clients and RabbitMQ 3.0.2?

Fri Feb 8 15:18:56 GMT 2013

Hey All,

While migrating our workers to a new cluster last night, some legacy
workers that use Celery 2.5.5 and the 0-8 AMQP python client got tons
of IO Broken Pipe errors. They successfully declared their queues and
exchanges but their normal message rates were about an 1/10 of what
their lower range normally is. We migrated the older workers back and
I'm spending my morning investigating the issue but I'm wondering if
there is any knowledge of this kind of issue.

Our setup is nothing fancy... just a four node cluster behind haproxy
with all queues designated as mirrored. Our 0-9-1 AMQP clients do not
seem to have the same issue. :)

Thanks,
James