[rabbitmq-discuss] rabbitmq server slows down for cluster set up - (multiple stop_app scenario)

[Francesco Mazzoli]

Hi Sanjiv,
At Mon, 9 Jul 2012 10:38:07 +0000,
sanjiv kumar jha wrote:
> Hi,
> 
> We have a RabbitMQ cluster of 2 nodes (say NODE-A and NODE-B) having mirrored
> queues running on 2 separate Linux servers.
> 
> We’ve multiple clients connecting to RabbitMQ cluster.
> 
> If NODE-A is brought down, the clients are able to connect to NODE-B
> successfully; no issues observed till this point.
> 
> Now if NODE-A is brought up and NODE-B is brought down, the rabbitmq server
> tends to slow down and does not even respond to simple command like
> rabbitmqctl list_queues – keeps waiting endlessly.
> 
> Please let us know, if there any clues or ideas to rectify this issue.

I'm not sure I understand what you're doing. My understanding is that

  * You have node A and B clustered and with HA queues on them
  * You bring down A
  * You bring up A
  * You bring down B *after you brought up A*

Is that correct? If that the case, that should be fine. My fear is that you are
bringing down B before A has fully started, in which case A will wait endlessly
for B since it needs its tables to get the most up-to-date data.

If you are making sure that A is fully started before stopping B, then what you
are observing should not happen and more information would help to investigate -
logs and `rabbitctl status'.

--
Francesco * Often in error, never in doubt