[rabbitmq-discuss] RabbitMQ Cluster, option "-setcookie cookie" + EC2 experiences

[Alex Clemesha]

Quick update on this...

Currently, I have "fixed" my setcookie issue by bundling a Rabbit AMI, which,
in turn, bundles the .erlang.cookie with it.  Now every instance of the AMI will
have the same cookie - so these instances can cluster with each other
... Not the
cleanest solution, but it works for now.


For completeness, here are a couple issues I've hit during this
(RabbitMQ cluster on EC2) process:

=> There are issue with bundling "/var/lib/rabbitmq/mnesia/rabbit/*"
because the Mnesia tables have a dependence on the current hostname.
I just did "rm -rf /var/lib/rabbitmq/mnesia/rabbit/*" before bundling,
which was good enough for now.
This issue is discussed in detail here:
http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/2009-January/003171.html


=> Sometime (but not always) when I instantiate a couple Rabbit ec2
instances at once, I see one of the instances fail to start up,
with the following message from  "rabbitmqctl status":

Status of node 'rabbit at ip-10-250-19-15' ...\n\n=ERROR REPORT====
10-Mar-2009::21:10:49 ===\nError in process <0.35.0> on node
'rabbitmqctl1437 at ip-10-250-19-15' with exit value:
{badarg,[{erlang,list_to_existing_atom,[\"rabbit@(none)\"]},{dist_util,recv_challenge,1},{dist_util,handshake_we_started,1}]}\n\n{badrpc,nodedown}\n...done.

This interesting part is the "rabbit@(none)" part I think.  A hostname
issue of some sort?  Here's another ec2 hostname rabbitmq-discuss
thread: http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/2008-March/000972.html
, but I'm fairly sure that I'm not having the same issue,
as I'm starting all my Cluster instance at the "availability_zone"
"us-east-1c" - which all have internal hostnames of
"ip-XXX-XXX-XXX-XXX"


Anyways, back to work!
thanks,
Alex




-- 
Alex Clemesha
clemesha.org