I have setup a cluster of rabbit brokers and within the management portal plugin I get the following message
"Statistics database could not be contacted. Message rates and queue lengths will not be shown"
I have searched for this erro开发者_如何学Gor but google is not being kind. Can anyone shed any light on this?
I had the same problem recently on an old installation of RabbitMQ (2.8.7) and found that there was no solution in this question. I found that to restart the statistics database, you can execute:
rabbitmqctl eval 'application:stop(rabbitmq_management), application:start(rabbitmq_management).'
(Source: http://en.it-usenet.org/thread/15496/19206/#post19199)
Update 2012-01-11: We believe we found the source of this and fixed it in 2.7.1. However it's very timing-dependent so we were never able to replicate it. If anyone sees this in 2.7.1 or later, can you mail support@rabbitmq.com please?
Old answer follows:
That's not really a message you should even be able to see in 2.6.1. I would be interested to know how you got there.
The management plugin maintains a single in-memory statistics database on one node in the cluster, even if it's installed on several / all nodes. Prior to 2.6.0 you would see that error if the node hosting the database failed or was removed. But in 2.6.0 we ensured (or I thought we did) that the statistics database would migrate to another node and reconstruct itself if necessary.
So this might not be a very good Stack Overflow question as it's looking more like a bug report. Do you see this after a node has failed / been removed or right from the start? Is mgmt installed on all nodes or just one? Can you send us the logs from the various nodes? Probably better to reply to support@rabbitmq.com in fact.
精彩评论