Take a look at the following scenario:
- I have a CoreOS / fleet cluster with 4 nodes.
There are 4 apache units with the following line
Conflicts=apache@%i.service
One machine goes down.
Fleet would try to relaunch the unit on another host but there is no possible host available. What does fleet do in this situation? Are there any differents between the handeling of different options like MachineMetadata and Conflicts
A unit that can't be scheduled just stays down. When a suitable machine rejoins the cluster, the unit will be started again.
That shouldn't be a big problem in your example, since there's not a lot of point having two identical apache units running on the same machine.