| CPC G06F 11/3476 (2013.01) [G06F 9/542 (2013.01); G06F 11/3006 (2013.01); G06F 11/3037 (2013.01)] | 8 Claims |
|
1. An integrated circuit to control operations in a partitionable multi-processor system that includes a first central processing unit connected to a second central processing unit by a processor interconnect, the operations comprising:
selectively disabling the processor interconnect that connects the first central processing unit to the second central processing unit;
initiating operation of a first partitioned node including a first subset of hardware devices of the multi-processor system that includes the first central processing unit in response to disabling the processor interconnect;
initiating operation of a second partitioned node including a second subset of hardware devices of the multi-processor system that includes the second central processing unit in response to disabling the processor interconnect;
initiating operation of the multi-processor system as a single unified node in response to the processor interconnect not being disabled;
receiving a predictive fault alert identifying a hardware device of the multi-processor system;
determining whether the multi-processor system is operating in a unified state as the single unified node or in a partitioned state including the first partitioned node and the second partitioned node;
determining, in response to determining that the multi-processor system is operating in the partitioned state, whether the identified hardware device is included in the first partitioned node or the second partitioned node;
forwarding, in response to the multi-processor system operating in the partitioned state and the identified hardware device being included in the first partitioned node, the predictive fault alert to a first designated reporting entity included in the first partitioned node; and
forwarding, in response to the multi-processor system operating in the partitioned state and the identified hardware device being included in the second partitioned node, the predictive fault alert to a second designated reporting entity included in the second partitioned node.
|