Providing high availability for a thin-provisioned container cluster

1. A system for executing containers, the system comprising:

a memory;

one or more processors in communication with the memory;

a scheduler executing on the one or more processors; and

a spot instance market monitor,

wherein the spot instance market monitor:

receives market information about a first plurality of spot instances in a cloud system at a first time, wherein the first plurality of spot instances are available to a client at the first time; and

determines, based on the market information, a respective reliability value of each of the first plurality of spot instances at the first time, wherein each respective reliability value corresponds to a respective likelihood of a respective spot instance being terminated by a second time after the first time;

wherein the scheduler:

selects a first spot instance in the first plurality of spot instances based on a first reliability value of the first spot instance;

responsive to the selection of the first spot instance, schedules a container on the first spot instance; and

executes the container on the first spot instance.

2. The system of claim 1 , wherein the first spot instance has the greatest reliability value among the first plurality of spot instances at the first time.

3. The system of claim 1 , wherein the spot instance market monitor determines the respective reliability value of each of the first plurality of spot instances by using an age of each of the first plurality of spot instances.

4. The system of claim 1 , wherein the spot instance market monitor determines the respective reliability value of each of the first plurality of spot instances periodically.

5. The system of claim 1 , wherein the spot instance market monitor determines a respective reliability value of each of a second plurality of spot instances in the cloud system at a third time after the first time, wherein the second plurality of spot instances are available to the client at the third time.

6. The system of claim 5 , wherein the scheduler:

selects a second spot instance in the second plurality of spot instances based on a second reliability value of the second spot instance; and

migrates the container from the first spot instance to the second spot instance.

7. The system of claim 6 , wherein the second spot instance has the greatest reliability value among the second plurality of spot instances at the third time.

8. The system of claim 6 , wherein the scheduler migrates the container from the first spot instance to the second spot instance by:

starting execution of the container on the second spot instance while the container is executing on the first spot instance; and

terminating the first spot instance once the second spot instance is fully executing on the second spot instance.

9. The system of claim 5 , wherein the scheduler sends a request to the container to take a snapshot of the container.

10. The system of claim 9 , wherein the container takes the snapshot more frequently as the first reliability value of the first spot instance decreases.

11. The system of claim 9 , wherein the scheduler:

detects that the first spot instance is terminated;

responsive to detecting that the first spot instance is terminated, selects a second spot instance in the second plurality of spot instances based on a second reliability value of the second spot instance; and

migrates the container to the second spot instance by:

pulling a latest version of the snapshot of the container; and

starting execution of the container on the second spot instance using the latest version of the snapshot of the container.

12. A method for executing containers, the method comprising:

receiving, by a spot instance market monitor, market information about a first plurality of spot instances in a cloud system at a first time, wherein the first plurality of spot instances are available to a client at the first time;

determining, by the spot instance market monitor, based on the market information, a respective reliability value of each of the first plurality of spot instances at the first time, wherein each respective reliability value corresponds to a likelihood of a respective spot instance being terminated by a second time after the first time;

selecting, by a scheduler executing on one or more processors in communication with a memory, a first spot instance in the first plurality of spot instances based on a first reliability value of the first spot instance;

responsive to the selection of the first spot instance, scheduling, by the scheduler, a container on the first spot instance; and

executing, by the scheduler, the container on the first spot instance.

13. The method of claim 12 , wherein the first spot instance has the greatest reliability value among the first plurality of spot instances at the first time.

14. The method of claim 12 , wherein the spot instance market monitor determines the respective reliability value of each of the first plurality of spot instances by using an age of each of the first plurality of spot instances.

15. The method of claim 12 , further comprising determining, by the spot instance market monitor, a respective reliability value of each of a second plurality of spot instances in the cloud system at a third time after the first time, wherein the second plurality of spot instances are available to the client at the third time.

16. The method of claim 15 , further comprising:

selecting, by the scheduler, a second spot instance in the second plurality of spot instances based on a second reliability value of the second spot instance;

migrating, by the scheduler, the container from the first spot instance to the second spot instance.

17. The method of claim 16 , wherein the scheduler migrates the container from the first spot instance to the second spot instance by:

starting execution of the container on the second spot instance while the container is executing on the first spot instance; and

terminating the first spot instance once the second spot instance is fully executing on the second spot instance.

18. The method of claim 15 , further comprising sending, by the scheduler, a request to the container to take a snapshot of the container.

19. The method of claim 18 , further comprising:

detecting, by the scheduler, that the first spot instance is terminated;

responsive to detecting that the first spot instance is terminated, selecting, by the scheduler, a second spot instance in the second plurality of spot instances based on a second reliability value of the second spot instance; and

migrating, by the scheduler, the container to the second spot instance by:

pulling a latest version of the snapshot of the container; and

starting execution of the container on the second spot instance using the latest version of the snapshot of the container.

20. A non-transitory machine readable medium storing instructions, which when executed by one or more processors, cause the one or more processors to:

receive, by a spot instance market monitor, market information about a first plurality of spot instances in a cloud system at a first time, wherein the first plurality of spot instances are available to a client at the first time;

determine, by the spot instance market monitor, based on the market information, a respective reliability value of each of the first plurality of spot instances at the first time, wherein each respective reliability value corresponds to a likelihood of a respective spot instance being terminated by a second time after the first time;

select, by a scheduler executing on the one or more processors, a first spot instance in the first plurality of spot instances based on a first reliability value of the first spot instance;

responsive to the selection of the first spot instance, schedule, by the scheduler, a container on the first spot instance; and

execute, by the scheduler, the container on the first spot instance.