In this presentation, I will talk about two problems. The first problem involves debugging long-running multi-threaded programs. When a fault occurs during an execution of such a program, we use a logging/replay system to replay a reduced execution that allows us to efficiently capture tracing information that is relevant to the fault. The other problem involves avoiding environmental faults during executions of server programs. Our approach is able to capture and recover from these environmental faults when they occur, and prevent them from occurring again.