Scaling Solr: From Desktop to Cloud Scale

At the recent Community over Code Denver conference I got to speak on how we could scale Solr seamlessly from the Desktop install to the Cloud scale setup. The talk was born out of my long time frustration that Solr has had two modes of operation, a legacy “Solr User Managed” mode (often referred to as “Standalone”) and the Solr Cloud mode.

For years it was a blind tenet of faith that running a single node Solr with embedded ZooKeeper was not viable. And yet, myself, and others (Eric waves at Christopher Ball) have done just that very successfully in many setups with multiple organizations. Single node SolrCloud setup works great for places where you don’t need “all the trimmings”.

The other blind tenet was that leader/follower replication has no place in Solr deployments. And yet, it turns out, that for many users, having one leader and one follower deals with their needs. Oh, and by the way, it works great to have two single node SolrCloud’s talking to each other via leader/follower.

I shared my mental model for how to scale Solr seamlessly in this talk:

I’ve started a https://issues.apache.org/jira/browse/SOLR-17465 as a “Umbrella” ticket in JIRA to make this a reality in the Solr 10 line.

I’m looking for more folks to join me on making this mental model a reality. Please hit me up on slack, email, JIRA and let’s simplify the scaling store from small to massive!