On 8/9/2013 12:24 PM, kaustubh147 wrote:
We have Solr installed on Glassfish cluster which has 4 nodes and we have a single solr.data directory which is shared among all 4 nodes.
This doesn't work well at all. Solr expects exclusive access to the Lucene index, and if you have more than one server accessing a shared directory, you have to play a complicated game of commits and/or restarts to make it work right. I've never done it, but from what I've seen on this list when others do, the risks aren't worth the hassle. It's even more complicated and messy when Solr is version 4.x.
If high availability and load balancing with four nodes is what you're after, you should run SolrCloud and build your collection with numShards=1 and replicationFactor=4. Each node will have its own independent copy of the index. Load balancing of both updates and queries is automatic, no matter which one it gets sent to. I think you could even run it under glassfish cluster. Three of those nodes would also run standalone zookeeper in addition to Solr.
Thanks, Shawn
