On 8/13/2014 12:49 AM, waqas sarwar wrote: > Hi, I'm using Solr. I need a little bit assistance from you. I am bit > stuck with Solr replication, before discussing issue let me write a brief > description. Scenario:- I want to set up solr in distributed architecture, > suppose start with least no of nodes (suppose 3), how can i replicate data of > each node to 2 others and vice versa. My Solution:- I set up “REPEATER” on > all nodes, each node is master to other, and configured circular replication. > Issue i'm facing:- All nodes are working fine replicating data from other > node, but when node1 replicate data from node2, node1 loses its own data. I > think node1 don’t have to atleast lose its own data & have to merge new data. > I think now question is pretty simple and clear, I want to set up solr in > distributed architecture, each node is replica to other, how may i achieve > it. Is there be any other way except Repeater and circular replication using > repeater, to replicate data of each node to all others. Environme nt:- LA N, Solr (3.6 to 4.9), Redhat
With master-slave replication, there must be a clear master, from which slaves replicate. You can't set up fully circular replication, or the master will replicate from the empty slave and your data will be gone. This form of replication does not merge data -- it makes the slave index identical to the master by copying the actual files on disk for the index. I think you'll want to use SolrCloud. You have three machines, so you have the minimum number for a redundant zookeeper ensemble. SolrCloud relies on zookeeper to handle cluster functions. SolrCloud is a true cluster -- no replication, no master. https://cwiki.apache.org/confluence/display/solr/SolrCloud Thanks, Shawn
