On 1/15/2016 4:14 AM, Mugeesh Husain wrote: > Actually i have a question , if i will use single zookeeper, > > suppose I have a 3 cluster and each of cluster used zookeeper instance(only > one zk). > > how we will manage zk in a way all of cluster will not communicate each > other?
This is not the proper way to set things up. Each of those three clusters will have a single point of failure - the zookeeper server. If zookeeper goes down, you will not be able to change your index, and if the clients are using (CloudSolrClient), new clients will not be able to connect to the cluster while zookeeper is down. You haven't described each cluster, so I do not know if there are more single points of failure. The first thing you need when setting up SolrCloud is a fully redundant three-node zookeeper ensemble. Then you configure your SolrCloud machines for you first cluster with a chroot on the zkHost parameter. You need a minimum of two Solr machines for redundancy. The second cluster gets configured with a different chroot on its zkHost, and the third cluster gets another chroot. Depending on how busy the servers are, you might be able to run both Solr and Zookeeper on some of the machines. If you have the extra hardware, it's recommended to separate them. Thanks, Shawn
