| CPC G06F 11/2236 (2013.01) [G06F 9/4881 (2013.01)] | 20 Claims |
|
1. A computer-implemented method comprising:
receiving, at a managed compute service of a cloud provider network, a request to terminate a compute instance hosted for a user by the managed compute service, wherein the compute instance is executed by a computing device and is attached to one or more graphics processing units (GPUs) of the computing device;
executing a termination workflow, comprising:
updating a status of the compute instance to be in a shutting down state, wherein while in the shutting down state, capacity associated with the compute instance is not available to the user,
terminating the compute instance of the user,
launching an auxiliary compute instance at the computing device, wherein the auxiliary compute instance is also attached to the one or more GPUs,
executing a cleanup workflow by the auxiliary compute instance, comprising executing an application, by the auxiliary compute instance, to test a functionality of the one or more GPUs and to reset one or more user-modifiable configuration values of the one or more GPUs,
determining that the cleanup workflow has been completed, and
terminating the auxiliary compute instance; and
after the execution of the termination workflow, updating the status of the compute instance of the user to be in a terminated state, wherein while in the terminated state, the capacity associated with the compute instance is available to the user.
|