Hi, I have requirements to make large amounts of data (> 5 million) documents search-able. The problem is that more than half have highly volatile field values. I will also have a data store specifically for Meta Data. Committing frequently isn't a solution. What I'm basically trying to achieve is NRT. I've read so many postings and articles everywhere and even considered sharing a single index amongst one WriteOnly Solr instance with >1-n Solr instances. Apparently this will not work since, calling commit on a searcher is the only way new documents will become search-able. I've also considered using one WriteOnly Master Instance with > 1-n ReadOnly Solr Slaves but that would mean there will be lag between snapshots of the master. Another solution that I was thinking about is having a smaller R/W Dynamic Master Solr instance that would only store deltas while I will still have a WriteOnly Master with a set of ReadOnly slaves. That would mean I would have to add some logic to combine and intersect the results from the dynamic Solr instance and R/O slaves. In this scenario, I wonder what would happen if I were to search for the top 25 documents that contains "x"? What would happen to scoring and other factors? Would sharding be better in this situation?
One more question is that I have not seen a lot of people discuss Solr-RA NRT? Is anyone familiar with it? There's not much mention of it except here http://solr-ra.tgels.com. Thanks, Zarni