I am running a Wildfly application in standalone mode inside docker containers that can be scaled up or down according to current load.
I configured Transaction Recovery in transactions subsystem to use JDBC Store since I have no persistent disk to use. Problem is, since my containers can be killed or created at any time they don't have anything to uniquely identify them. Hostnames are randomly generated and IPs are randomly assigned.
So if for example I try to set the node-identifier of the transactions subsystem to the hostname and this container is killed, it might leave orphaned uncompleted transactions.
Am I right about this? Do you have any suggestions regarding this issue?