[rabbitmq-discuss] RabbitMQ Clustering with SSL

[Francesco Mazzoli]

Hi Jignesh,

At Tue, 11 Sep 2012 14:55:41 +0000,
Purohit, Jignesh wrote:
> We are facing problem in RabbitMQ clustering on two existing node enable for
> SSL. First we have set up SSL on both the nodes on SSL default port 5671 and
> then we have applied cluster set up with two disk nodes. After successful
> cluster set up we can able to see the cluster environment on management
> console which shows two disk nodes clustered.

I'm not sure what you mean here with "set up SSL".  I guess that you're opening
SSL connections to clustered nodes - if your intention is to encrypt the
inter-node communications, that is a bit tricky (see
<http://www.erlang.org/doc/apps/ssl/ssl_distribution.html>).

> 1.  But after few hours While verifying it, we observed the management console
>     doesn’t appear so we checked the services and its stops on both the
>     machines.

If I understand correctly you have a durable, HA queue on a two nodes cluster,
where both nodes are on disc.

Then after some activity the queue disappears.  I need more details on the
activity on the queue and on the broker, and the logs on both nodes, otherwise
it is impossible for us to know what's going on.

Also, what version of RabbitMQ are you using?
 
> So we have started service of first node and then we can see the management
> console and it shows the second server disk node gets down. The time duration
> between set up and testing was hardly 1 hr. 30 mins. Node didn’t went up by
> following command. We followed attached installation document for RabbitMQ
> installation. We also checked the cookies files and its correct.

I don't understand this.  You say that you can't bring a node up, but why was it
down in the first place?  Do you start your cluster with one node running?  Did
it crash? I need more details.

> 2.  To manage the cluster set up, we have created another disk node on the
>     machine with diff name and same port (5671). The node is created
>     successfully and shown the active status in management console. But the
>     node working is not correct and is as follows.

Again, it's unclear what this is referring to - is this what you do when trying
to set up a cluster at the beginning?  In any case, trying to run two cluster
node on the same port on the same machine won't work.

> 1.  The messages are not transport properly on the queues generated after
>     clustering. The queues are mirror queues. The queue status also shows “?”
>     and when any message send to the queue, the management console is not able
>     to open the queue detail window and the page gets hanged. It seems it goes
>     into some loop. When we restart the RabbitMQ service, the page display
>     properly but the queue status is same.

Is this referring to the queue crash described above, or is it another issue?

> 2.  When we shut down the first node and shift the focus on 2^nd node then the
>     messages get deleted from all the queues and not able to retrieve the
>     messages. The queues are persistence and durable queues. When we send the
>     message on 2^nd node then the behaviours of queue is same as mentioned in
>     #1.

I am really struggling to understand what's going on here, please try to explain
more clearly.

> 3.  If any node of RabbitMQ or its service get shut down abruptly then is
>     there any provision to send notifications to administrator? Currently this
>     functionality is exists for other Windows services

If the node gets shut down abruptly, then surely it has no way to notify the
administrator.  You can surely set up some monitoring for port on which RabbitMQ
is listening to, but my Windows experience is limited and I'm not sure what the
best way to do that is.

--
Francesco * Often in error, never in doubt