摘要:
A diagnosis system includes procedures to perform various operations, including diagnosis of problems, in a run-time environment recreated from dump information. The dump information is saved in response to a fault occurring in a node, which may be part of a parallel processing system. The diagnosis system may be separate from the parallel processing system or it may be one of the nodes of the parallel processing system. Setting up the run-time environment includes identifying memory sections associated with a task that caused the fault and filling the identified memory sections with data extracted from the dump information. Diagnosis procedures may then be run in the recreated run-time environment. The diagnosis may be the same diagnosis procedures used in a live run-time environment to save time and costs associated with writing such diagnosis procedures.