[rabbitmq-discuss] Count message without Ack on a queue

Thu Aug 7 16:29:06 BST 2014

 On 7 August 2014 at 19:23:10, bloodBaron (bloodbaronofvizkopa at gmail.com) wrote:
> > The problem about using an HTTP API is the HTTP call is like
> http://someIP:15672/api/vhosts.
> This will leads to a single point failure in a distributed system  
> if someIP
> goes down.

People run HTTP servers behind load balancers (or even multiple layers of)
at a massive scale at places like Google or Amazon. You could do
the same ;)

> Even if I use a domain instead of an IP, the system will rely on the  
> domain
> server, it is NOT a closed system anymore.

How do you define "closed"? As in security? Then it's up to how you configure
your network. You can use HTTPS with the management plugin, too.

Finally, rabbitmqctl uses distributed Erlang facilities directly, which
means if you can access it and it can connect to the cluster, you can
do *anything*. I'd rather use the HTTP API than try to use rabbitmqctl
directly and then try to make the resulting solution more secure.
--  
MK  

Staff Software Engineer, Pivotal/RabbitMQ