Container cloning and branching

1. A method, comprising:

invoking a library to clone a container running on a host, wherein the invoking comprises acquiring from the library a predetermined execution point that identifies in code where, during execution of the running container, a cloning operation should occur, and a location of a second host, associated with the execution point, to which a clone of the running container is to be deployed;

making, by the library, a synchronous call to a controller requesting that the running container be cloned;

in response to the call, the controller performing:

pausing the container at the execution point specified in the library;

generating a checkpoint image of the running container, wherein the checkpoint image includes a memory dump of the running container and is stored in a checkpoint image storage;

storing the checkpoint image as a child of the running container in a list maintained by the controller, wherein the list identifies a tree of containers that designate a container as a parent container and one or more other containers as children of the parent and leaf containers, wherein the tree of containers allows restoration of a leaf to be performed using checkpoint images in a corresponding branch of the tree of containers and wherein the running container and the clone container are included in the tree of containers;

restoring the checkpoint image into a clone container on the second host that is specified in the library, wherein restoring the checkpoint image includes restoring the memory dump of the running container into the clone container; and

issuing a response to the call to the clone container indicating that the clone container may proceed to new instructions or requests;

in response to receiving the response, resuming operation of the clone container at the execution point; and

resuming operation of the running container at the execution point.

2. The method of claim 1 , wherein the library identifies an environment for the clone container, the environment including the second host and settings of the clone container.

3. The method of claim 1 , further comprising cloning the clone container by generating a second checkpoint image of the clone container and restoring the second checkpoint image into a second clone container, wherein the second clone container is a child of the clone container.

4. The method of claim 3 , wherein the second checkpoint image is an incremental checkpoint image, wherein the incremental checkpoint image has a smaller size compared to a full checkpoint image, wherein restoring the second checkpoint image includes restoring at least a portion from the checkpoint image.

5. The method of claim 1 , wherein the second host comprises specialized hardware.

6. The method of claim 1 , further comprising terminating, live-migrating or backing up all containers associated with a portion of the tree using the list.

7. The method of claim 1 , further comprising selectively killing the running container.

8. The method of claim 1 , further comprising establishing a communication channel for the running container and the clone container, wherein the communication channel includes one of a message bus or a default hostname and port.

9. The method of claim 1 , wherein the containers in the tree of containers are configured to operate in a distributed style such that different children containers execute different aspects of a repeated workflow for a parent container.

10. The method of claim 9 , further comprising restoring the children containers selectively such that each of the children containers is configured to perform a different portion of the repeated workflow.

11. A non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to perform operations, the non-transitory storage medium is not a propagation signal, the operations comprising:

invoking a library to clone a container running on a host, wherein the invoking comprises acquiring from the library a predetermined execution point that identifies in code where, during execution of the running container, a cloning operation should occur, and a location of a second host, associated with the execution point, to which a clone of the running container is to be deployed;

making, by the library, a synchronous call to a controller requesting that the running container be cloned;

in response to the call, the controller performing:

pausing the container at the execution point specified in the library;

generating a checkpoint image of the running container, wherein the checkpoint image includes a memory dump of the running container and is stored in a checkpoint image storage;

storing the checkpoint image as a child of the running container in a list maintained by the controller, wherein the list identifies a tree of containers that designate a container as a parent container and one or more other containers as children of the parent and leaf containers, wherein the tree of containers allows restoration of a leaf to be performed using checkpoint images in a corresponding branch of the tree of containers and wherein the running container and the clone container are included in the tree of containers;

restoring the checkpoint image into a clone container on the second host that is specified in the library, wherein restoring the checkpoint image includes restoring the memory dump of the running container into the clone container; and

issuing a response to the call to the clone container indicating that the clone container may proceed to new instructions or requests;

in response to receiving the response, resuming operation of the clone container at the execution point; and

resuming operation of the running container at the execution point.

12. The non-transitory storage medium of claim 11 wherein the library identifies an environment for the clone container, the environment including the second host and setting of the clone container.

13. The non-transitory storage medium of claim 11 , the operations further comprising cloning the clone container by generating a second checkpoint image of the clone container and restoring the second checkpoint image into a second clone container, wherein the second clone container is a child of the clone container.

14. The non-transitory storage medium of claim 13 , wherein the second checkpoint image is an incremental checkpoint image, wherein the incremental checkpoint image has a smaller size compared to a full checkpoint image, wherein restoring the second checkpoint image includes restoring at least a portion from the checkpoint image.

15. The non-transitory storage medium of claim 11 , wherein the second host comprises specialized hardware.

16. The non-transitory storage medium of claim 11 , further comprising terminating, live-migrating or backing up all containers associated with a portion of the tree using the list.

17. The non-transitory storage medium of claim 11 , the operations further comprising selectively killing the running container.

18. The non-transitory storage medium of claim 11 , the operations further comprising establishing a communication channel for the running container and the clone container, wherein the communication channel includes one of a message bus or a default hostname and port.