The topic of rolling upgrades in database management is discussed, with the goal of upgrading a cluster without disrupting its operation. The development of rolling upgrades was prompted by the need to address breaking changes in protocol used between instances, particularly when upgrading from 1.5.3 to 1.7 series. A rolling upgrade allows each version to gracefully coordinate and continue serving requests while other versions are being upgraded or brought online. Each slave is detected as it's brought up, and falls back into a compatibility mode that prevents it from becoming the master but still executes transactions. The order of upgrades matters, with slaves typically started first and then the master. This approach has been implemented in Neo4j 1.8.M06, which offers immediate download, provisioning on Heroku, discussion on the Google Group, and questions on Stack Overflow.