We understand the logistics of how to do it, but are unsure of the best practice. Right now we have a 5 node Cassandra DSE production cluster set up (replication factor of 3). We've recently experimented with Solr nodes on single node Cassandra clusters, and are really pleased with the resultant search capabilities as opposed to using secondary indexes and additional column families in Cassandra.
What is the best way to bring up DSE Search/Solr nodes in our production environment? From the responses to this question:
Can we have cassandra only nodes and solr enabled nodes in same datacenter?
it seems like we should just change the config to bring up search on each of the nodes. The ideal for us (we think), however, would be to add search only nodes to the same cluster+DC, although also from that question, it doesn't appear to be possible. What's our best way forward here? Is it to bring up two new nodes in a different DC (which appear to need both cassandra and solr on them) or to change the config on each of the current nodes in the cluster?
Thanks in advance!
Running search and cassandra in the same DC works if your SLAs aren't too tight and your workload is not too demanding.
If this is not the case, you want to data centers, one for search and one for cassandra for workload isolation.
Either way, search nodes have cassandra on them, that is how the DSE Search implementation works (both run in the same JVM). Cassandra takes care of replication, repair, partitioning etc. and solr takes care of indexing, etc.