Question

This might be a case of me over thinking the problem, but how should I document an HBase schema? In relational database land, it is common to use UML or similar diagramming techniques to document schemas. Those approaches don’t seem to fit HBase very well. To me the simplest approach is to use a spreadsheet (or any other table) to document the columns and column families. Is there a better way to do this?

Answer 1

There’s no common method at this point. You can do it in whatever way seems clearest to you. I’ve used entity diagrams (generally more trouble than they’re worth), XML/JSON, and pseudo CREATE TABLE statements. A couple things that might help:

• A presentation I did at HBaseCon 2012 about understanding HBase schema design, which proposes a few simple guidelines.
• An simple open-source tool called scoot that takes in XML files and outputs JRuby scripts for creating said schema

If you’re getting into advanced stuff like “entity nesting” (i.e. using variable columns at runtime) then you’re kind of on your own as far as modeling goes, for now.