Currently when an Amphora type load balancer resource is under a change (create, update, delete, failover) and the controller service handling that provisioning flow shuts down abruptly, i.e. no graceful shutdown, the Taskflow flow is interrupted and the resource left stuck in PENDING_* provisioning state forever.

A workaround to unlock from this pending state is to manually change the provisioning status to ERROR and then either recreating (delete + create) or failover the load balancer or amphora.

This work allows sub-flow resumption of change actions by detecting a flow is not progressing and automatically re-assigning the flow to an alternate controller.

Note: This is an Ussuri priority upstream.

Taskflow provides a persistence module which allows flow resumption.
https://docs.openstack.org/taskflow/latest/user/persistence.html
https://docs.openstack.org/taskflow/latest/user/resumption.html

https://storyboard.openstack.org/#!/story/2005072