Normally the suggested approach for high availability is a blue/green deployment to a separate and isolated 0.11 cluster before pull the LB switch over, you can run migrations on a existing copy of DB before hand and validate the installation of course, just make sure not to add/modify/delete Kong entities during this process and you should be fine.
As @Cooper pointed out is a must to review that document to address any breaking changes between versions.
I don’t have specific recommendations about Bucardo though.