[rabbitmq-discuss] Best practices to manage failovers at the publishers side

[Michael Klishin]

On 17 Mar 2014, at 12:58, rails <stammailbox at gmail.com> wrote:

> What options does rabbitMQ (client side) have that will help me to manage this scenario?
> 1. Can he accumulate the messages ? 

It’s possible but since the outage can last for minutes or hours, your client risks
running out of RAM.

> 2. Can he manage retries? 

You can attempt to reconnect and it will help with short outages. You really should try to
reconnect to a different cluster node.

> 
> The second part of the question is regarding managing the failover itself. 
>  I know that in some databases I can write a connection string as follows 
> 
>        Data Source=myServerAddress;Failover Partner=myMirrorServerAddress;
> and if the action fails the client automatically switches to another one in the same action.

RabbitMQ Java and .NET clients can use a list of addresses when connecting. The first
reachable node will be used. Many other clients don’t have this feature but it’s pretty
easy to imitate.

MK

Software Engineer, Pivotal/RabbitMQ