Compound computer system and method for sharing PCI devices thereof

A management method for a compound computer system comprising:

• a plurality of computers comprising:
• a processor for carrying out arithmetic processing;
• a memory for storing information;
• and
• a chipset connected to the processor and having a PCI interface;
• a plurality of I/O devices comprising a multi-root I/O device sharable among at least two of the plurality of computers;
• a PCI switch for mutually connecting the plurality of computers and the plurality of I/O devices with each other;
• a resource management module for managing assignment of the plurality of I/O devices to the plurality of computers;
• and
• a multi-root I/O manager for controlling the PCI switch to logically connects the plurality of computers to the plurality of I/O devices with each other,
• the compound computer system carrying out, at an occurrence of a failure in one of the plurality of I/O devices, recovery processing for the failure,
• the management method comprising:
• setting, by the resource management module, I/O device management information containing, as management information on the multi-root I/O device, at least failure handling content information indicating, for each of the plurality of computers sharing the multi-root I/O device, a content of a failure handling at an occurrence of a failure in the multi-root I/O device, and failure handling availability status information indicating, for the each of the plurality of computers sharing the multi-root I/O device, whether a hardware reset of the multi-root I/O device is possible;
• monitoring, by the multi-root I/O manager, the PCI switch, and notifying, at the occurrence of the failure in the multi-root I/O device, the resource management module of the failure of the multi-root I/O device;
• monitoring, by the multi-root I/O manager, the PCI switch, and restraining, at the occurrence of the failure in the multi-root I/O device, the hardware reset of the multi-root I/O device in which the failure has occurred;
• obtaining, upon reception of a notification of the occurrence of the failure in the multi-root I/O device from the multi-root I/O manager, by the resource management module, progress information on the failure handling from the each of the plurality of computers sharing the multi-root I/O device;
• updating, by the resource management module, the failure handling availability status information based on the obtained progress information on the failure handling;
• instructing, by the resource management module, based on the updated failure handling availability status information, the multi-root I/O manager to one of restrain and cancel the hardware reset of the multi-root I/O device;
• carrying out, by the each of the plurality of computers, at the occurrence of the failure in the multi-root I/O device, the failure handling specified by the failure handling content information, and requesting, when the failure handling is the hardware reset of the multi-root I/O device, the multi-root I/O manager to carry out the hardware reset of the multi-root I/O device;
• and
• receiving, by the multi-root I/O manager, a request for the hardware reset of the multi-root I/O device in which the failure has occurred from the each of the plurality of computers, and restraining, upon receiving an instruction to restrain the hardware reset of the multi-root I/O device from the resource management module, the hardware reset of the multi-root I/O device.

A management server for controlling a compound computer system comprising:

• a plurality of computers comprising:
• a processor for carrying out arithmetic processing;
• a memory for storing information;
• and
• a chipset connected to the processor and having a PCI interface;
• an I/O device sharable among the plurality of computers;
• a PCI switch for mutually connecting the I/O device and the plurality of computers with each other;
• and
• an I/O manager for controlling the PCI switch,
• the management server comprising a resource management module for managing association between at least two of the plurality of computers and the I/O device,
• wherein the resource management module comprises:
• I/O device management information containing failure handling content information indicating, for each of the plurality of computers associated with the I/O device, a content of a failure handling at an occurrence of a failure in the I/O device, and failure handling availability status information indicating, for the each of the plurality of computers sharing the I/O device, whether a hardware reset of the I/O device is possible;
• and
• a failure handling arbitration module for obtaining, upon reception of a notification of the occurrence of the failure in the I/O device from the I/O manager, progress information on the failure handling from the each of the plurality of computers associated with the I/O device, updating the failure handling availability status information, and instructing, based on the failure handling availability status information, the I/O manager to one of restrain and cancel the hardware reset of the I/O device.