We have a client node, which was full (mem, cpu). But we wanted to deploy new jobs to this node with a contraint and a higher prio e.g. 51, so our intent was to let nomad migrate jobs which can run elsewhere from this node to another node.
To our suprise, this didn’t work. Jobs weren’t migrated.
You’ll need to enable preemption in order to achieve the goals you expect. This Preemption Learn guide holds some useful information regarding this feature.
Is this the expected behavior?
Yes, in your current configuration, the behaviour you are seeing is the expected one.