Redis is our primary data store for all data. We do also use a Lucene index for our search indexing, but aside from our search indexing ALL data is in Redis.
Are you glad you didn't use Redis for your indexing? We've just had some consultants show us search indexing in Redis, and it looks very exciting.
The issue is fuzzy searching. Our first prototype had all search indexing in Redis however things such as spelling corrections, weighting, grammar aware, as well as complex searches with boolean statements, and so on become exceptionally difficult. If you simply need an inverted-index search then Redis is pretty compelling, but if you need something more advanced than a basic inverted-index it becomes untenable.