If you need to reindex without affecting queries, is it better to have two cores and swap, or create a new collection, index, create an alias, and delete the collection previously associated with the alias?
What are the trade offs?
Solr handles the index swap automatically so no manual steps are required.
If you index very large indexes you could index on one core and replicate the result to the slave cores for added throughput.
When we indexed our large index we swapped our core into the replication domain but that shouldn't be needed if you don't want to, as in our case, want to verify the data before release.