When a connection is made to the JanusGraph DB an identifier generated by the JanusGraph client is stored in the DB (as entry "?system-registration.ac170007125-sdc-BE1.startup-time" in "column1" in sdc-titan.system_properties.
During upgrade when the sdc-be starts it attempts to connect to the DB, the JanusGraph client reads the existing config (from the above DB table) and re-uses the identifier if one is found (UniqueInstanceIdRetriever::getOrGenerateUniqueInstanceId()). This then causes the error "A JanusGraph graph with the same instance id [ac170007125-sdc-BE1] is already open"

The existing identifer should however no longer be present as the previous client has exited. The continued presence of the identfier is highly likely due only to the caching in the DB and we have observed where the sdc-be is delayed in restarting that this issue does not arise. The caching has only recently been turned on in https://gerrit.onap.org/r/c/sdc/+/122516
The proposed solution is to config JanusGraph to replace the existing instance if still present