[rabbitmq-discuss] Bringing down Nodes in a cluster

[Patrick Long]

Michael Klishin <mklishin at ...> writes:

> 
> On 28 Feb 2014, at 12:27, Patrick Long <pat at ...> wrote:
> 
> > So I can stop the node by using stop_app or I can stop the RabbitMQ 
service on that node. Is one better that the other?
> 
> Stop the service. stop_app does not stop the VM, which may be necessary 
with some upgrades
> (e.g. you also want to upgrade the Erlang VM).
> 
> Use stop_app when resetting nodes, as described in “Breaking up a cluster” 
[1].
> 
> > When Ops want to update a machine that is a Node in a 3 node RabbitMQ 
cluster what is the recommended approach?
> > 
> > Is it to stop the service on one node, run Windows Update etc etc then 
bring that node back. Once that node is
> confirmed move to next node always ensuring there is at least one 
synchronised slave? 
> 
> See “Breaking up a cluster” in [1].
> 
> Removing nodes from the cluster and bringing them back one by one is a 
good
> strategy.
> 
> HTH.
> 
> 1. http://www.rabbitmq.com/clustering.html
> 
> MK
> 
> Software Engineer, Pivotal/RabbitMQ
> 
> 
What is the benefit to breaking up the cluster rather than simply stop_app 
then stop the service, patching and rebooting? 

Is it to stop the cluster trying to find the downed node?

Thanks

pat