In one of our environments, we have an issue where one shard has two replicas with smaller document counts than the third one. This is on Solr 4.10.3 (Cloudera's build). We've found that shutting down the smaller replicas, deleting their data folders and restarting one by one will do the trick of forcing them to get the bigger and fresher index from the third one.
We aren't doing anything different with the document router configuration or anything like that. It's a really simple and straight forward installation of Solr that is largely based on defaults for everything. Any suggestions on what might be getting us into this situation? Also, is there a SolrCloud API for forcing those two replicas to sync with the third or do we have to continue using that manual process? Thanks, Mike
