<html><head><meta http-equiv="content-type" content="text/html; charset=utf-8"></head><body dir="auto"><div>Unless you mostly do request-reply workloads, just use disk nodes.<br><br>MK</div><div><br>On 01/05/2014, at 15:26, srikanth tns <<a href="mailto:srikanthtns@gmail.com">srikanthtns@gmail.com</a>> wrote:<br><br></div><blockquote type="cite"><div><div dir="ltr">Also , we are trying to use 2DISC 1RAM in the cluster . Is this good enough ? Also which node should be set as RAM , is there any preference to set only a particular node the cluster to be RAM ?<div><br></div>
<div>Thanks</div><div>Srikanth</div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Wed, Apr 30, 2014 at 1:48 PM, srikanth tns <span dir="ltr"><<a href="mailto:srikanthtns@gmail.com" target="_blank">srikanthtns@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Thanks Micheal , here is the settings we are doing on rabbitmq cluster<div><br></div><div>{rabbit, [{vm_memory_high_watermark_paging_ratio, 0.8},{vm_memory_high_watermark, 0.9},{disk_free_limit, 1000000000}]}<br>
</div><div><br></div><div>Apart from this ,is there any QoS value that we need to set ?</div></div><div class="HOEnZb"><div class="h5"><div class="gmail_extra"><br><br><div class="gmail_quote">On Tue, Apr 29, 2014 at 1:13 AM, Michael Klishin <span dir="ltr"><<a href="mailto:mklishin@gopivotal.com" target="_blank">mklishin@gopivotal.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div><br>
<br>
On 29 April 2014 at 07:22:03, srikanth tns (<a href="mailto:srikanthtns@gmail.com" target="_blank">srikanthtns@gmail.com</a>) wrote:<br>
> > Can you let us know what would be optimal settings on rabbitmq<br>
> cluster to prevent it from crashing and having reliable messaging?<br>
<br>
</div>It's really hard to make general recommendations without knowing what<br>
kind of workload you use.<br>
<br>
> 1) sysctl configurations<br>
<br>
nlimit > 20,000 or so (definitely can't be less than ~ 6K)<br>
<div><br>
> 2) Disk limitations to setup on cluster . the default is 50MB<br>
<br>
</div>I'd use 500 MB or so if RabbitMQ is the primary thing that runs on the machine.<br>
<div><br>
> 3) Memory to used , right now its default 40%<br>
<br>
</div>Can be up to 90%, again, depending on what else may run on the same machine.<br>
<div><br>
<br>
> 6) any disk space limitation on the hosts<br>
<br>
</div>Host is this different from 2) above?<br>
<div><br>
> For HA , we are using ha-mode all to duplicate the queues across<br>
> all nodes in the cluster. Do you think its reliable during the<br>
> failover scenario ?During the failover it would like 100k queues<br>
> on nthe new master? Or should we use ha-mode = exactly/nodes ?<br>
<br>
</div>With 3 nodes ha-mode=all is a reasonable amount of duplication. If you are confident<br>
you will never lose more than 1 node at a time, use ha-mode=exactly with 2 nodes.<br>
Less duplication means better throughput.<br>
<br>
As to how well new master election will work with 100K queues, try it. <br>
--<br>
MK<br>
<br>
Software Engineer, Pivotal/RabbitMQ<br>
</blockquote></div><br></div>
</div></div></blockquote></div><br></div>
</div></blockquote></body></html>