We're setting up an Index Cluster with a Master Node. From the documentation it looks like the Cluster will take care of replicating data and configuration between the Indexers. However, we're also wanting to have multiple search heads that work with the cluster.
What have people used to setup the search heads behind a load balancer. Do we need to use shared storage for the search heads or is there better configuration?
@gkanapathy did mention something similar to that in his Architecting for Scale talk at .conf 2012, maybe he can shed some light on viable alternatives to search head pooling.
Thanks for the answer. I'd like to avoid using shared storage as this adds complexity. I heard that some folks rsync knowledge bundles between search heads. We don't plan on running scheduled searches on the search heads, but rather have a separate server for that.
Have you seen this as an option?