I was part of Azure's open source database migration team, where we shepherded customer databases to the cloud one careful byte at a time. Migrating databases is like moving a sleeping toddler and a vase at the same time — one wrong move and someone cries or break(trust/revenue). Data cannot be corrupted or lost during migration. What I loved about the job was getting deep into the internals of Postgres and MySQL. Most of our solutions leaned on native DB technologies like WAL log replication, dump/restore, and native replication — basically the database equivalent of duct tape and a prayer. Thanks for reading this far; I just wanted to set the scene before I hit the point. One more pre-context note: I was exploring CRDT systems for replication when I ran into DRBD, and suddenly things got even more delightfully nerdy.