Because of hardware faults, the situation that the processor cannot perform properly is occurred frequently in large scale software-intensive systems. Most of traditional fault-tolerant methods do not distinguish the type of hardware failure. In view of this, we propose self-repairing software architecture for predictable hardware faults. By introducing computational reflection, the software architecture designs and realizes a proactive processor fault tolerant manager to complete component active migration according to the result of the migration decision. So, users do not need to modify the code to complete the self-repairing process. Experimental results show the architecture help enable system to selfrepairing and reduce predictable hardware failure impact on the running software.