Ping on this question - My guess it that it just got lost in the shuffle.<br><br>Also, just realized in the above question that I transposed something. The two lines that say:<br><br><br>federation: skytap -> rabbit@slave<br>
federation: skytap -> rabbit@slave-alternate<br><br>Should instead be:<br><br>federation: skytap -> rabbit@master<br>federation: skytap -> rabbit@master-alternate<br><br>Thanks,<br><br>Matt<br><div class="gmail_extra">
<br><br><div class="gmail_quote">On Mon, Dec 10, 2012 at 1:17 PM, Matt Pietrek <span dir="ltr"><<a href="mailto:mpietrek@skytap.com" target="_blank">mpietrek@skytap.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
I realize this is a somewhat esoteric question and apologize for the complexity. We're still on 2.8.7 for a while, although my guess is that federation in 3.0 won't drastically change the answer.<br><br>In our setup, we have bidirectionally federated nodes, i.e. 'slave' <-> 'master' with an exchange named 'skytap'. To this mix I add a set of alternate nodes to enable non-stopping upgrades: 'master-alternate' and 'slave-alternate'. The primary and alternate nodes are identically configured.<br>
<br>I'm using VIPs (keepalive based) for connections between nodes. There's a vip-master and a vip-slave. Most of the time vip-master points at the 'master" node, but occasionally points at the 'master-alternate" node. In the rabbitmq.config, I specify the VIP names everywhere (i.e vip-master, vip-alternate), and never mention the actual node names.<br>
<br>When I first start this up, I see a federation support queue like this:<br><br>federation: skytap -> rabbit@master<br><br>This is what I'd expect based on my understanding of federation.<br><br>However, after doing an upgrade process (i.e., vip-master goes from 'master' to 'master-alternate' and back to 'master'), I have two queues now:<br>
<br>federation: skytap -> rabbit@slave<br>federation: skytap -> rabbit@slave-alternate<br><br>This isn't so good. The vast majority of the time the slave-alternate node isn't around. Messages sent to the skytap exchange just pile up in the second queue. (The do of course get delivered to the non-alternate slave node.)<br>
<br><b>Question 1:</b><br>I'm trying to understand this behavior. It's almost as if the federation logic is burrowing underneath my VIP and discovering the actual node names. In my naive understanding, I'd expect that the use of VIPs would hide away the primary/alternate nodes.<br>
<br>That said, I do notice in the connections list that incoming connections come from the 'regular' node IP, not from the VIP that's assigned to the same node. Not sure if this is relevant or not.<br><br><b>Question 2:</b><br>
In a related vein, looking at the Exchanges I see exchanges like this:<br><br>federation: skytap -> rabbit@slave A<br>federation: skytap -> rabbit@slave-alternate B<br><br>What do the A/B signify?<br><br>Thanks,<br>
<br>Matt<br><br>
</blockquote></div><br></div>