There's nothing in your case that makes it much different from a "single domain" setup.
Solr works as REST service, so in order to index content, all you have to do is provide access from different domains. For bulk import (e.g. from a db or csv) we use DataImportHandler.
If you want to facet on domain-bases (narrow down search results by one or more domains using check boxes, after results are already displayed) and you have e.g. an article entity that you want to index, on top of its regular attributes (Solr fields like Body, Author, Date, ...) you should add additional one that would represent its domain.
To setup security (who can see a document) I'd use ManifoldCF, which knows how to work with Solr out-of-the-box.
If you're just starting with Solr I suggest you first go through this tutorial, which is a great starting point to grasp Solr (and general search engine) concepts.