Using Fleet I'm trying to implement high-availability solution for a certain daemon which does master-slave replication. There are two instances - "master" and "shadow master". When "master" is down, "shadow master" can be promoted to "master" which is trivial to implement with Fleet. However the problem that is difficult to solve is re-starting "shadow master" somewhere else after "shadow master" to "master" fail-over.

A new fleet-specific option "Replaces" could be helpful in this scenario. Unlike stronger "Conflicts", I suppose "Replaces" could force re-scheduling of another unit to another machine. For example, "master.service" could contain

[X-Fleet]
Replaces=shadow.service

so when "master.service" is started on machine where "shadow.service" is running the latter is stopped and automatically re-scheduled to run on another machine.

I hope that such option can be considered useful.