Rolling changelog execution in clustered mode

[Jotschi]

It would be possible to execute database changelog entries online in a rolling update setup.

• The new mesh instance would execute the changelog and update the database structure in a way that keeps it compatible with older mesh versions which are also part of the cluster.
• The new mesh instance tracks operations that may be invoked by older mesh instances. Newly created elements may need to be migrated on-the-fly.
• Once the whole cluster has been switched to the new mesh version and no older mesh instance is online: The final changelog step will be executed which removes the compatibility layer from the db.

https://thoughts-on-java.org/update-database-schema-without-downtime/

[mephinet]

https://www.martinfowler.com/books/refactoringDatabases.html