[Pacemaker] Web farm question

[Andrew Beekhof]

On 25/04/2013, at 12:49 AM, Robert Parsons <rparsons at tappublishing.com> wrote:

> 
> We are building a new web farm to replace our 7 year old system. The old system used ipvs/ldirectord/heartbeat to implement redundant load balancers. All web server nodes were physical boxes.
> 
> The proposed new system will utilize approximately 24 virtual machines as web servers. Load balancing and redundancy will be done via ClusterIP rather than an explicit load balancer. We're looking at potentially 50 services running on 24 nodes. Pacemaker seemed like it would accommodate such a setup, however, from my research I am worried that perhaps it will not scale up to such a level.

You would need to significantly tune corosync for your network and be sure to run at least rc1 of pacemaker 1.1.10 (a lot of performance/scaling work and memory leak fixes went into 1.1.9+1.1.10 - see http://blog.clusterlabs.org/blog/2013/large-cluster-performance-redux/)
But it should be doable.

> 
> Does anyone have any experience or feedback on this proposed approach to a web farm? Should I consider another approach?
> 
> Thanks for any info.
> 
> - Rob Parsons
> 
> 
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org