HeroDB - Managing Data Like Source

We'll discuss HeroDB, a key/value database that Yieldbot developed that has been in use in production for about a year for all configuration data. HeroDB is built on top of git and keeps perfect history on all operations, can be set to any previous state, self contains a history of all operations, etc.

Since we can navigate, access, and modify a filesystem view of the data with standard git and filesystem tools, all of our developers are able to be database administrators.

Get a local copy of the database?

$ git clone user@databaseserver:/datapath/db.git

Get most recent changes to local view?

$ git pull

Get a log of all modifications?

$ git log

$ gitk

Start looking at data?

$ ls

Reset to last known good state?

$ git reset --hard

Undo someone's configuration change?

$ git revert

Put changes into production?

$ git push