<div style="font-family: Helvetica; font-size: 13px; "><span style="color: rgb(160, 160, 168); ">On Saturday, 29. December 2012 at 15:37, Francesco Mazzoli wrote:</span></div>
<blockquote type="cite" style="border-left-style:solid;border-width:1px;margin-left:0px;padding-left:10px;">
<span><div><div><div>OK, I think I understand your issues better now: you’d like better tools to</div><div>exist to setup clustering in an automated environment like Chef. Right now we</div><div>don’t provide much in that space.</div><div><br></div><div>What we had in mind is something along the lines of “take a declarative</div><div>description of a cluster, e.g. ‘I want a rabbit cluster consisting of four</div><div>nodes, A,B,C,D. A & B shall be disk nodes, C & D ram nodes’, and try to convert</div><div>that into a set of commands to execute on the four machines, with the commands</div><div>being idempotent and execution order between machines being irrelevant.”. This</div><div>sounds reasonable and not too hard to do. I can’t promise anything though :).</div><div><br></div></div></div></span></blockquote><div>That sounds like a great idea, and I'm all for it. In general it'd be great to have some more operational and automation knowledge collected in the documentation too. </div><blockquote type="cite" style="border-left-style:solid;border-width:1px;margin-left:0px;padding-left:10px;"><span><div><div><div></div><div>For what concerns now, things like</div><div><br></div><blockquote type="cite"><div><div>join_cluster does fit my needs, but in an automated environment, e.g. with</div><div>Chef, that command would be triggered every 30 minutes. Would you consider</div><div>that a good practice to do?</div></div></blockquote><div><br></div><div>Are perfectly fine, you can easily check periodically if a node is clustered as</div><div>you want and fix it if something is not of your liking. However be aware that</div><div>`join_cluster' will leave the current cluster the node is in and thus reset the</div><div>node, so act carefully.</div><div><br></div></div></div></span></blockquote><div>That sound reasonable. It's also why I like the configuration file approach though. It's a declarative approach and doesn't require any commands to be fired to e.g. check if the node is part of the cluster and make it join via join_cluster, when run via Chef, reducing resource overhead just a little bit :)</div><div><br></div><div>Thanks!</div><div><br></div><div>Cheers, Mathias</div><div><br>
</div>