[rabbitmq-discuss] "Zombie" RabbitMQ Processes

[Emile Joubert]

Hi Chris,


On 10/06/13 19:49, Chris wrote:
> (1) Is there any way we can troubleshoot this further than just looking
> at the logs in /var/log/rabbitmq?


Check whether "rabbitmqctl report" works. Inspect the whole logfile, and
check the sasl logfile too. It is exceedingly rare for the broker to
crash or become unresponsive in a way that does not leave some evidence
behind. Check your OS logfile also.


> (2) Should we be concerned about the connection_closed_abruptly
> warnings?  Currenlty we connect to RabbitMQ using node-amqp and the Java
> client.  I /think/ we usually should be closing the connection on exit.


No, that is normally innocuous. You can turn the logging for that off
with the log_levels configuration option.


> (3)  Can anyone suggest how we might be getting into this state?  Might
> it be possible if we started RabbitMQ in a script and somehow killed it
> or restarted it before it was finished initializing (just guessing
> here)?  So far I cannot reproduce it, but we've had several reports of it.


It is hard to venture guesses without knowing what's wrong. A broker
that has run out of disk or memory can appear to be stuck. A very busy
broker can appear unresponsive. Some network problems can cause a broker
to appear stuck also.




-Emile