[rabbitmq-discuss] HQ queue seems never to get fully created, is never usable

Sat Sep 24 00:09:56 BST 2011

I'm having some difficulty with creating a mirrored queue so that I
can test it. The management plugin only thinks the queue is half-
present, rabbitmqctl doesn't think it's there at all, and the AMQP API
seems to be blocking when it shouldn't.

1. I created a two-node cluster.  This much seems fine.
cluster_status gives the right report on both machines:
Cluster status of node rabbit at ps7 ...

[{nodes,[{disc,[rabbit at ps8,rabbit at ps7]}]},
 {running_nodes,[rabbit at ps8,rabbit at ps7]}]

2. I used the Perl API to create an HA queue.  This reported success
via the AMQP, and the management plugin sees the queue and its
parameters correctly.  When I hover over HA I see "x-ha-policy:
nodes", and when I hover over Args I see "x-ha-policy-params:
[rabbit at ps7,rabbit at ps8]".  A get to the HTTP API shows the same
data, in JSON of course.

3. When I issue a consume to the queue, my client hangs, rather than
continuing as it did in my non-HA test.  This may be a quirk of the
client, but it is a change from the non-HA working behavior.

4. The output of "rabbitmqctl list_queues" is completely empty.

5. The management plugin seems conflicted as to whether the queue
exists or not.  The queue is present and visible in the queue list,
but when I go to the queue's page the big items for Ready, Unacknowledged,
and Total all show "?" instead of a number, and a GET to the queue's
specific URL returns a 404 error.

How can I diagnose this situation and/or make this work right?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20110923/dda87898/attachment.htm>