[rabbitmq-discuss] getting started, broker runs; can't get status

Sun Feb 1 01:09:28 GMT 2009

> Dave, Christopher, to test the above theory, please re-enable Nagle by
> changing the following lines in the rabbitmq-server script from
> 
> [ "x" = "x$RABBITMQ_SERVER_ERL_ARGS" ] && RABBITMQ_SERVER_ERL_ARGS="+K
> true +A30 \
> -kernel inet_default_listen_options
> [{nodelay,true},{sndbuf,16384},{recbuf,4096}] \
> -kernel inet_default_connect_options [{nodelay,true}]"
> 
> to
> 
> [ "x" = "x$RABBITMQ_SERVER_ERL_ARGS" ] && RABBITMQ_SERVER_ERL_ARGS="+K
> true +A30"
> 
> (Alternatively you can override the value by setting
> RABBITMQ_SERVER_ERL_ARGS in your shell or SERVER_ERL_ARGS in
> /etc/default/rabbitmq)

The /etc/init.d script invokes rabbitmq-multi, not rabbitmq-server. 

When I add those args to rabbitmq-multi (via RABBITMQ_MULTI_ERL_ARGS), I see
they get passed to erl.   But if the cwd is "/" the startup still fails
(whereas if it is /var/lib/rabbitmq it works).

I have a clean ubuntu that I upgrade to jaunty, but I'm having trouble
reproducing the problem cleanly.  I can't make it fail yet.   I will
continue to try.   I've spent a lot of time with strace so far without
getting a clue.   I'm suspicious of the node name handling in the erlang
runtime/kernel, but even running all-but-identical /etc/hosts and identical
/etc/resolv.conf, /etc/nsswitch.conf, /etc/hosts.conf hasn't shed a clue.

-ch