Amazon.com, Inc.
Recovery protocols for system malfunctions in virtual computing environments

Abstract:

Generally described, the present disclosure relates to the use of a virtual machine monitor to instruct one or more virtual processing units belonging to a malfunctioning virtual machine instance to pause processing. While paused, the virtual machine monitor may store the register data of the virtual processing units to virtual memory of the virtual machine instance (e.g., perform a register dump on behalf of the virtual processing units of the malfunctioning virtual machine instance). The virtual machine monitor may then instruct virtual processing units to reinitialize and invoke protected software (e.g., a crash kernel) from virtual memory in an effort to recover from virtual memory the register data dumped there by the VM monitor.