Rolling version update deployment utilizing dynamic node allocation

• US 10,001,983 B2
• Filed: 07/27/2016
• Issued: 06/19/2018
• Est. Priority Date: 07/27/2016
• Status: Active Grant

• Current Assignee: Salesforce.com, Inc.
• Original Assignee: Salesforce.com, Inc.
• Inventors: Roy, Saptarshi, Silver, Daniel C., Schneider, Donovan, Pradhan, Medha, Lamore, Ryan, Rao, Naveen Purushothama, Geh, Nicholas, Koneru, Srirama
• Primary Examiner(s): Kang, Insun
• Application Number:
  US15/221,496
• Publication Number:
  US 20180032323A1
• Time in Patent Office: 692 Days
• Field of Search: None
• CPC Class Codes: G06F 8/65 Updates security arrangements therefor G06F21/57
• CPC Class Codes: G06F 9/50 Allocation of resources, e.g. of the central processing unit [CPU]
• CPC Class Codes: G06F 9/5083 Techniques for rebalancing the load in a distributed system
• CPC Class Codes: H04L 67/34 involving the movement of software or configuration parameters network booting or remote initial program loading [RIPL] G06F9/4416

Abstract

Methods are disclosed for dynamic node allocation for a server system that can automatically heal on failure—minimizing the need for static configuration—dynamically adjusting server resources to match load, and minimize end user wait times. The disclosed methods dynamically allocate nodes to increase capacity for a platform that accepts data queries. Additionally disclosed is a system for rolling version update deployment: workers maintain org lists of org-task-queues that they service; org-affinities between the workers and the org-task-queues require the workers to have access to local copies of immutable data sets to service org-tasks from the org-task-queues of the orgs that they service. A leader running on a worker implements a healing and balancing service that maintains worker redundancy, manages the workers'"'"' org-affinities to accumulate orgs on their respective org-lists. The leader implements messaging to the workers to update to a new software version and monitors completion of updates.

Claims

1. A hardware system with rolling version update deployment, including:

workers on a set of devices in the system, the workers maintain lists of org-task-queues to be serviced by the workers;

org-affinities between the workers and the org-task-queues to provide access to local copies of org-data-sets to service org-tasks from the org-task-queues of orgs serviced by the workers;

a configuration leader running on a device in the set of devices;

a healing and balancing service implemented by the configuration leader that maintains worker redundancy, that manages the workers'"'"' org-affinities, and that causes workers to accumulate orgs on their respective org-lists and to have heterogeneous org-affinities;

a messaging service, implemented by the configuration leader, messaging workers to update from a legacy software version to a new software version and to implement monitoring of completion of updates in a set of update cycles including;

the configuration leader selects workers to update in a cycle and ensures that a selected level of worker redundancy to service particular org-task-queues is not compromised by coordinating the selection of workers taken out of service during the update cycle based on the selected workers'"'"' org-affinities;

the configuration leader informs the selected workers in the cycle to proceed with updating; and

when the configuration leader learns that the selected workers have successfully completed updating, the configuration leader updates version accounting over the selected workers and moves on to another cycle of updating; and

the configuration leader repeating the cycle to update all update-eligible workers.

2. The system of claim 1, further including the devices on which the workers run being organized by racks and redundant workers running on devices in disjoint racks.

3. The system of claim 2, further including satisfying the selected level of worker redundancy by updating to the new software version on a rack-by-rack basis.

4. The system of claim 1, further including the configuration leader running on a worker or management device.

5. The system of claim 1, wherein the configuration leader suspends the healing and balancing service during updating.

6. The system of claim 1, wherein the worker redundancy maintained by the healing and balancing service has a flexible integer value of at least three redundant workers.

7. The system of claim 1, wherein the configuration leader increases redundancy of workers servicing the org-task-queues with which the selected workers have org-affinities, including provisioning the org-data sets to increased redundancy workers to establish org-affinities.

8. The system of claim 7, wherein the configuration leader for at least one cycle of updating reports results from the cycle of updating and waits for instructions from an administrator to proceed before repeating the cycle of updating.

9. The system of claim 1, wherein the workers after being informed to proceed with updating, stop taking new tasks, complete pending tasks, shut off services, update to the new software version, restart, and report available for duty.

10. A tangible non-transitory computer readable medium with instructions that are combined with a processor and memory coupled to the processor to construct the system of claim 1.

11. A method of rolling version update deployment, with workers on devices, org-affinities between the workers and org-task-queues, and a configuration leader running on a worker or management device, including:

wherein org-affinities between the workers and the org-task-queues provide access to local copies of org-data-sets to service org-tasks from the org-task-queues serviced by the workers;

implementing, by the configuration leader running on a worker or management device, a healing and balancing service that maintains worker redundancy, that manages the workers'"'"' org-affinities, and that causes workers to accumulate orgs on their respective org-lists and to have heterogeneous org-affinities;

messaging workers from a messaging service, implemented by the configuration leader, to update to a new software version and monitoring completion of updates, in cycles including;

selecting workers to update in a cycle, and ensuring that a selected level of worker redundancy to service particular org-task-queues is not compromised, by coordinating the selection of workers taken out of service during the update cycle based on the selected workers'"'"' org-affinities;

informing the selected workers in the cycle to proceed with updating; and

updating version accounting over the selected workers by the configuration leader, upon learning that the selected workers have successfully completed updating, and moving on to another cycle of updating; and

the configuration leader repeating the cycle to update all update-eligible workers.

12. The method of claim 11, further including the devices on which the workers run being organized by racks and redundant workers running on devices in disjoint racks.

13. The method of claim 11, further including the configuration leader satisfying the selected level of worker redundancy in worker availability by updating to the new software version on a rack-by-rack basis.

14. The method of claim 11, wherein the org-data-sets are cached locally, at least in the same rack, to the workers that service the org-task-queues of the orgs.

15. The method of claim 11, wherein the configuration leader suspends the healing and balancing service during updating.

16. The method of claim 11, wherein the worker redundancy maintained by the healing and balancing service has a flexible integer value of at least three redundant workers.

17. The method of claim 11, wherein the configuration leader increases redundancy of workers servicing the org-task-queues with which the selected workers have org-affinities, including provisioning the org-data sets to increased redundancy workers to establish org-affinities.

18. The method of claim 11, wherein the workers after being informed to proceed with updating, stop taking new tasks, complete pending tasks, shut off services, update to the new software version, restart, and report available for duty.

19. The method of claim 11, wherein the configuration leader for at least one cycle of updating reports results from the cycle of updating and waits for instructions from an administrator to proceed before repeating the cycle of updating.

20. A tangible non-transitory computer readable medium with instructions that implement the method of claim 11.

21. A hardware system including devices organized in racks, each device including a processor and memory coupled to the processor, the memory loaded with instructions that implement the method of claim 11.