[rabbitmq-discuss] Inconsistent queue synchronization behavior

Matt Pietrek mpietrek at skytap.com
Wed Mar 14 16:48:57 GMT 2012

> the data returned by rabbitmqctl reports is correct. Are you sure?

I'm working on reproing that exact scenario. However, I did uncover another
related inconsistency just now.

Starting from the original scenario (3 mirrored brokers, stop 1 of them),
the Node column in the "Queues" tab shows "+1" (as expected) for four of my
five queues.

However, one queue continues to show "+2". When hovering over it, the two
nodes listed include the broker node that I shut down. The "Overview" tab
confirms that this broker is indeed shut down.

As for the rabbitmqctl results, this time they *do* seem correct, i.e.

*rabbitmqctl list_queues name slave_pids synchronised_slave_pids*

indicates that all queues have exactly one mirror.

On Wed, Mar 14, 2012 at 6:29 AM, Simon MacMullen <simon at rabbitmq.com> wrote:

> On 13/03/12 21:42, Matt Pietrek wrote:
>> some
>> subset of the queues (varies from run to run) go back to "+2", while the
>> remaining queues switch to "+1 +1", meaning 1 synched, and 1 unsynched.
>> Given that there are no messages in the queue, I don't understand why
>> some queues don't synch back up properly.
> OK, given a bit of fiddling I have managed to recreate this.
> Congratulations, you just delayed the 2.8.0 release ;-)
>  Also, I've looked at the sync status of the queues in rabbitmqctl, and
>> the results seem consistent with what the HTTP API reports.
> Hmm. In my recreation I don't see that - the data returned by rabbitmqctl
> reports is correct. Are you sure?
> Cheers, Simon
> --
> Simon MacMullen
> RabbitMQ, VMware
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20120314/5eb60251/attachment.htm>

More information about the rabbitmq-discuss mailing list