A technique that allows existing reservations to be partially preempted by higher-priority reservations. An intermediate node receives a request to reserve resources for the higher-priority reservation. The higher priority entitles the reservation to acquire resources allocated to existing lower priority reservations, if necessary. If sufficient unallocated resources are not available on the preempting system to accommodate the higher-priority reservation, an error message is generated and forwarded by the intermediate node towards a receiver associated with the reservation that is to be partially preempted. The error message contains an error code and an amount of resources available on the intermediate node. The receiver processes the error message including selecting a reservation to be torn down, tearing down the selected reservation and re-reserving resources for remaining reservations, as necessary.