[rabbitmq-discuss] RabbitMQ node stopped while broker still up

Wed Oct 17 10:40:22 BST 2012

On 17/10/12 03:04, Wong Kam Hoong wrote:
> Hi Simon,
>
> I checked the log and noticed every time when start this particular
> RabbitMQ, the log showed some error messages as follow:
>
> =ERROR REPORT==== 4-Oct-2012::05:46:57 ===
> Mnesia(nplay at app2): ** ERROR ** mnesia_event got {inconsistent_database,
> starting_partitioned_network, nplay at web1}
>
> =ERROR REPORT==== 4-Oct-2012::05:46:57 ===
> Mnesia(nplay at app2): ** ERROR ** mnesia_event got {inconsistent_database,
> starting_partitioned_network, nplay at web2}
>
> Why the errors logged? Is it related to my cluster setting?

That log message ("starting_partitioned_network") means that Mnesia 
detected it was partitioned last time it was up, and is recovering from 
it now (where the recovery is fairly brutal - the node logging that 
message has thrown away any local changes, and has resynchronised with 
whichever nodes it found running).

> This is production environment, I can guaranteed no one dare to execute
> the command "rabbitmqctl stop".

Well, I can promise that we only log that message "Halting Erlang VM" in 
response to "rabbitmqctl stop". If you don't believe me you can check - 
it's emitted in rabbit:stop_and_halt():

http://hg.rabbitmq.com/rabbitmq-server/file/rabbitmq_v2_8_x/src/rabbit.erl#l310

which is only invoked here:

http://hg.rabbitmq.com/rabbitmq-server/file/rabbitmq_v2_8_x/src/rabbit_control.erl#l161

- the handler for "rabbitmqctl stop".

Cheers, Simon

-- 
Simon MacMullen
RabbitMQ, VMware