A method and technique for data lifecycle management includes identifying a fault from a monitored system. A time period window is defined associated with the fault based on a time the fault occurred. One or more metrics that are related to the fault that fall within the time period window are identified from the monitored system and stored in a memory. A lifespan is assigned to the one or more metrics based on the fault, and the one or more metrics are removed from the memory when their associated lifespans are over.