Lucene/Solr Revolution 2017

For the majority of the cases, current SolrCloud distributed indexing works great. There is a subset of use cases for which the legacy Master/Slave replication may be a better fit, like cases where NRT is not required, and where read availability is more important than consistency. For such cases we are adding to Solr a way to choose different types of replicas, that treat updates in different ways. With a combination of replica types, one can create a SolrCloud cluster that behaves like a Master/Slave architecture of Solr < 4.0 and provides separation of responsibilities (search vs index) while still getting most of the SolrCloud benefits, like high availability of writes, replica discovery, collections API, etc.

This talk will be a deep dive into the Replica Type feature, reasons for implementing it, differences between the existing types and how/why one would choose to use them, and implementation details of the feature.