On 9/1/2016 6:48 AM, Shawn Heisey wrote: > You can use the load balancer URL with HttpSolrClient instead of using > the zookeeper information with CloudSolrClient. At that point, you > will be dependent on haproxy to balance the load. It probably will not > react as fast to servers going down as SolrCloud's built-in load > balancing, but it should work well.
I need to tell you about one possible hiccup with this approach -- SolrCloud will *still* load balance the queries across the cloud, even without CloudSolrClient, unless you tell it not to. There is a parameter "preferLocalShards". You can find documentation on this parameter here: https://cwiki.apache.org/confluence/display/solr/Distributed+Requests The documentation covers when it is a good or bad idea to use the parameter. Pairing an external load balancer with SolrCloud is the right time to use preferLocalShards. Thanks, Shawn
