Recording activity of software threads in a concurrent software environment

摘要

A technique for failure monitoring and recovery of a first application executing on a first virtual machine includes storing machine state information during execution of the first virtual machine at predetermined checkpoints. An error message that includes an application error state at a failure point of the first application is received, by a hypervisor, from the first application. The first virtual machine is stopped in response to the error message. The hypervisor creates a second virtual machine and a second application from the stored machine state information that are copies of the first virtual machine and the first application. The second virtual machine and the second application are configured to execute from a checkpoint preceding the failure point. In response to receipt of a failure interrupt by the second application, one or more recovery processes are initiated in an attempt to avert the failure point.