Virtual machine instance migration using a hypervisor

Patent metadata

Field	Value
Publication number	US-10404579-B1
Application number	US-201514961765-A
Country	US
Kind code	B1
Filing date	Dec 7, 2015
Priority date	Dec 7, 2015
Publication date	Sep 3, 2019
Grant date	Sep 3, 2019

Abstract

Techniques for managing a critical phase of a virtual machine migration are described herein. During the critical phase, which must be kept as short as possible, a virtual machine instance is paused, network packets are rerouted to a temporary packet queue, and messages associated with managing synchronization of block storage devices are dispatched to an entity configured to manage those connections. After the block storage devices are synchronized, the network packets are released from the temporary packet queue to the new location of the virtual machine instance.

First claim

What is claimed is:

1. A computer-implemented method, comprising:
receiving a request to begin a critical phase of a virtual machine migration for a first virtual machine instance;
pausing, using a first hypervisor, execution of the first virtual machine instance, the first virtual machine instance having access to resources provided by the first hypervisor, the first virtual machine instance managed by the first hypervisor, the first hypervisor running on a first computer system of one or more computer systems;
after pausing the execution, receiving network packets that are directed to the first virtual machine instance to an incoming packet queue provided by a privileged instance running on the first computer system, the privileged instance having access to resources of the first computer system, the privileged instance managed by the first hypervisor;
validating a standby lease to a block storage device for a second virtual machine instance on a second computer system of the one or more computer systems, the standby lease acquired using a second hypervisor executing on the second computer system;
completing a copy of memory from the first virtual machine instance to the second virtual machine instance;
after copying the memory:
forwarding the received network packets in the incoming packet queue to the second virtual machine instance;
receiving new network packets directed to the first virtual machine instance; and
routing the new network packets to the second virtual machine instance;

converting an active lease to the block storage device for the first virtual machine instance to an inactive lease; and
converting the standby lease to an active lease.

2. The computer-implemented method of claim 1 , wherein receiving the network packets directed to the first virtual machine instance to the incoming packet queue is performed by the first hypervisor.

3. The computer-implemented method of claim 1 , wherein receiving the network packets directed to the first virtual machine instance to the incoming packet queue is performed by a network translation service.

4. The computer-implemented method of claim 1 , wherein memory and state of the first virtual machine instance is copied to the second virtual machine instance by the first hypervisor in parallel with pausing execution of the first virtual machine instance and in parallel with routing the network packets directed to the first virtual machine instance to the incoming packet queue.

5. A system, comprising at least one computing device configured to implement one or more services, wherein a first service of the one or more services is configured to:
in a pause stage, pause a first virtual machine instance;
in a route stage, route, while the first virtual machine instance is in the pause stage, network packets directed to the first virtual machine instance to an incoming packet queue;
send a message indicating completion of the route stage;
receive a message indicating completion of a standby lease stage associated with validating a standby lease for a block storage device;
in a packet forward stage, transfer the network packets from the incoming packet queue to a second virtual machine instance and route, by at least bypassing the incoming packet queue, new incoming network packets directed to the first virtual machine instance to the second virtual machine instance, the second virtual machine instance configured based at least in part on the first virtual machine instance, the second virtual machine instance located at a target location associated with a migration of the first virtual machine instance; and
send a message indicating completion of the packet forward stage.

6. The system of claim 5 , wherein the first service is a hypervisor.

7. The system of claim 5 , wherein a second service of the one or more services is configured to:
receive the message from the first service indicating the completion of the route stage;
validate the standby lease by associating the second virtual machine instance with the block storage device, the standby lease specifying a first policy of access to the block storage device by the second virtual machine instance; and
send the message indicating the completion of the standby lease stage.

8. The system of claim 7 , wherein the second service is configured to:
receive the message indicating completion of the packet forward stage;
convert an active lease associating the first virtual machine instance with the block storage device to an inactive lease; and
convert the standby lease associated with the second virtual machine instance to an active lease, the active lease associated with the second virtual machine instance specifying a second policy of access to the block storage device by the second virtual machine instance.

9. The system of claim 7 , wherein the second service is a hypervisor.

10. The system of claim 5 , wherein the first service is configured to route the new incoming network packets addressed to the first virtual machine instance to the second virtual machine instance after memory is copied from the first virtual machine instance to the second virtual machine instance.

11. The system of claim 5 , wherein the incoming packet queue is provided by a privileged instance associated with the first service.

12. The system of claim 5 , wherein the first service is further configured to, before pausing the first virtual machine instance, at least:
perform a first iteration of a memory and state copy, the first iteration copying all the memory and state of the first virtual machine instance; and
perform one or more additional iterations of the memory and state copy, the one or more additional iterations copying the memory and state that have changed after previous iterations, until it is determined that a number of changes still remaining is below a threshold value.

13. One or more non-transitory computer-readable storage media having stored thereon executable instructions that, as a result of being executed by one or more processors of a computer system, cause the computer system to at least:
specify a set of flip stages, the set of flip stages including a pause stage associated with pausing a first virtual machine instance, a route stage associated with routing packets to an incoming packet queue, a standby lease stage associated with validating a standby lease for a block storage device, a packet forward stage associated with forwarding packets from the incoming packet queue to a second virtual machine instance and routing new packets, by at least bypassing the incoming packet queue, directed to the first virtual machine instance to the second virtual machine instance, and an active lease stage associated with activating the second virtual machine instance such that:
the pause stage is executed after a critical phase of a migration of the first virtual machine instance has started;
the route stage is executed after the pause stage completes;
the standby lease stage is executed after the route stage completes;
the packet forward stage is executed after the standby lease stage completes; and
the active lease stage is executed after the packet forward stage completes; and

instantiate a first hypervisor configured to cause the computer system to at least:
validate the standby lease by associating the second virtual machine instance with the block storage device;
convert an active lease by associating the first virtual machine instance with the block storage device to an inactive lease; and
convert the standby lease to an active lease.

14. The one or more non-transitory computer-readable storage media of

Assignees

• Amazon Tech Inc

Inventors

• Biemueller Sebastian Martin
• Elissaios Georgios
• Gupta Diwakar
• Ma Hui
• Brouwer Pieter Kristian
• Husain Mohammad Farhan
• Friebel Thomas

Classifications

• G06F2009/45595

  Network integration; Enabling network access in virtual machine instances · CPC title

• G06F9/45558

  Hypervisor-specific management and integration aspects · CPC title

• G06F2009/4557

  Distribution of virtual machine instances; Migration and load balancing · CPC title

• H04L47/50

  Queue scheduling · CPC title

• H04L61/25

  of the same type · CPC title