ML Engine monitors free space in each container in all pods, checking disk capacity in the root directory every 30 seconds and canceling live queries when a disk is 85% full. ML Engine does not shut down when disk capacity is reached.
Query cancellation is automated and not configurable.