Combined with the information on the Internet, copied a few pictures, recorded.
1.solrcloud-collection/shard/replica
1.Replica, leader is the role of core, in the index, search the process of different roles.
2. The different shard are distributed as far as possible on different SOLR nodes. The core on SOLR is as evenly distributed as possible, ensuring an average load.
2.solrcloud-index
1. Send the index request to any core.
2. If core is replica, forward to leader on the same shard.
3. Determine which shard index to (routing rule), if not on the current Shard, forward to the corresponding Shard leader (step 3 '), if on the current Shard, forward to the corresponding replica.
3.solrcloud-search
1. Send the search request to any core.
2. The core, based on the number of Shard, forwards the request to any replica of different Shard, and searches each.
3. The replica that participate in the search returns the search results to "initiator", which completes the merge of the result set and returns (step 4).
Solrcloud index Search (figure)