ODL has a somewhat complex upgrades story. We should create some specific docs dedicated to known/recommended upgrade strategies and their limitations. trade-offs and mitigations.

For example, topics to cover:

• Data store upgrades being difficult at the moment, how to do yang schema migrations
• Possible to work-around in some deployments by doing replay of events (say from OpenStack Neutron) to ODL. Can do OVS table updates in atomic batch, no dataplane downtime.
• Additional suggestion is to deploy ODL in a container so can do update by swapping for newer-version container.
• Configuration management tooling like ansible-opendaylight and puppet-opendaylight handle upgrades because they assume no yang model schema upgrades. This is normally okay, and almost certainly okay between SR releases.
• RPM upgrade tests are WIP, link to patch.