[rabbitmq-discuss] Clustering On EC2

Michael Bridgen mikeb at rabbitmq.com
Mon Dec 12 12:58:30 GMT 2011


Hi Mick,

> But am I right in saying the only solution is to use short names, in
> effect using the hosts file on the actual cluster
> machines to identify each other?
>
> In my naive thinking I would have assumed supporting FQDN's to be
> straight forward, they all resolve to ip's right?

The main difficulty is simply that Erlang's clustering needs either all 
short names or all "long" names, and its starting arguments discriminate 
between these two situations.

I tried this, locally:

I changed '-sname' to '-name' in the scripts rabbitmq-server and 
rabbitmqctl, then ran two nodes with full names (e.g., 
rabbit at example.com and hare at example.com) using RABBITMQ_NODENAME. These 
happily clustered together and ran the management plugin, through which 
I created a mirrored queue and sent some messages.

This suggests to me that it will work fine with the minimal modification 
to scripts given in the attached patch. A release that supports this may 
be longer in coming.


-Michael
-------------- next part --------------
A non-text attachment was scrubbed...
Name: longnames.patch
Type: text/x-patch
Size: 775 bytes
Desc: not available
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20111212/cd1e1056/attachment.bin>


More information about the rabbitmq-discuss mailing list