Coordinated upgrades in distributed systems

• US 9,753,713 B2
• Filed: 10/22/2010
• Issued: 09/05/2017
• Est. Priority Date: 10/22/2010
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Mani, Ajay, Dournov, Pavel, Irun-Briz, Luis, Nagesharao, Pavithra, Xu, Haizhi, Hassan, Akram, Shah, Ashish, Pfleiger, Todd
• Primary Examiner(s): Zhen, Li B
• Application Number:
  US12/910,712
• Publication Number:
  US 20120102481A1
• Time in Patent Office: 2,510 Days
• Field of Search: 717172
• CPC Class Codes: G06F 11/0757 by exceeding a time limit, i.e. time-out, e.g. watchdogs
• CPC Class Codes: G06F 8/65 Updates security arrangements therefor G06F21/57
• CPC Class Codes: G06F 9/45504 Abstract machines for programme code execution, e.g. Java virtual machine [JVM], interpreters, emulators

Abstract

A distributed application may be updated by using information provided by the application to determine grouping, sequencing, and whether or not to advance an update sequence. The application may provide input to an update sequence, and may cause the update sequence to repair one or more replicas prior to updating. The update mechanism may be used to change the distributed application topology as well as make changes to the application executable code, application configuration, quest operating system, virtual machine, and host systems, among others. In some cases, the distributed application may use a quorum to determine which version is current, and the quorum may change during the update process.

Claims

1. A method performed on a plurality of computer processors, said method comprising:

executing a distributed application on said plurality of computer processors, said distributed application being executed by a plurality of replicas of application code;

said distributed application being executed on virtual machines having guest operating systems, said virtual machines being operated on host devices;

grouping said plurality of replicas into update domains, said update domains being groups of said replicas;

identifying an update to perform on said distributed application;

querying the distributed application, by a fabric controller, to identify a sequence of said update domains to update;

for each of said update domains, processing, by the fabric controller according to the sequence, a current update domain to perform said update, said processing comprising;

querying said application executing in a current update domain, before transmission of said update for reception by said current update domain, to determine whether said application considers said current update domain ready for updating;

after receiving a query response indicative of said current update domain being ready for updating, performing said update on said update domain;

querying said application executing in said current update domain to determine if said update results in said application operating properly within said current update domain;

when said update results in said application operating properly, proceeding to a next update domain; and

when said update results in said application operating improperly, suspending said update.

2. The method of claim 1, said update comprising updating a version of said application.

3. The method of claim 2, said updating said version of said application comprising updating application executable code.

4. The method of claim 2, said updating said version of said application comprising updating configuration information for said application.

5. The method of claim 1, said update comprising updating said host.

6. The method of claim 1, said update comprising updating said guest operating system.

7. The method of claim 1, said grouping being performed by querying said distributed application for said grouping and receiving input on said grouping from said distributed application.

8. The method of claim 7, said querying being directed to a primary replica of said distributed application.

9. The method of claim 1 further comprising:

determining a quorum from a plurality of said update domains, said quorum determining a current version of said distributed application.

10. The method of claim 9, said quorum being determined after each of said update domains are processed as said current domain.

11. The method of claim 10, said quorum defining a primary replica of said distributed application.

12. The method of claim 1, said update being performed in multiple update steps, each of said multiple update steps comprising processing each of said update domains in sequence.

13. The method of claim 12, said update comprising:

performing a first update step comprising;

querying said application executing in a current update domain to determine whether said current update domain is ready for updating; and

if said application executing in said current update domain is not ready for updating, performing a repair on said application;

performing a second update step comprising;

performing an update to said application.

14. The method of claim 1, each of said update domains being updated atomically.

15. A cloud computing platform comprising:

a plurality of host devices, each of said host devices comprising a processor;

for each of said host devices, at least one virtual machine having a guest operating system;

for each of said virtual machine, at least one replica of a distributed application, said distributed application comprising a plurality of said replicas;

a fabric controller that;

determines a grouping for said replicas, said grouping defining a plurality of update domains, each of said update domains comprising a plurality of said replicas;

queries the distribution application to determine a sequence of said update domains to update;

processes each of said update domains according to said sequence by processing a current update domain in an atomic fashion, said current update domain being processed by a method comprising;

querying said application executing in a current update domain, before transmission of said update for reception by said current update domain, to determine whether said application considers said current update domain ready for updating;

after receiving a query response indicative of said current update domain being ready for updating, performing said update on said update domain;

querying said application executing in said current update domain to determine if said update results in said application operating properly within said current update domain;

when said update results in said application operating properly, proceeding to a next update domain; and

when said update results in said application operating improperly, suspending said update.

16. The cloud computing platform of claim 15, said update comprising at least one of a group composed of:

updating said host;

updating said guest operating system;

updating said virtual machine;

updating configuration for said distributed application; and

updating executable code for said distributed application.

17. The cloud computing platform of claim 16, said grouping of said replicas being determined by querying said distributed application to allow the distributed application to provide input to determine how to group said replicas.

18. The cloud computing platform of claim 17, wherein the distributed application is queried to allow the distributed application to provide input to determine how to sequence said update domains.

19. A method performed on a plurality of computer processors, said method comprising:

executing a distributed application on said plurality of computer processors, said distributed application being executed by a plurality of replicas of application code;

said distributed application being executed on virtual machines having guest operating systems, said virtual machines being operated on host devices;

identifying an update to perform on said distributed application;

transmitting information about said update to said distributed application to be considered by said distributed application in connection with responding to queries regarding said update;

querying said distributed application to determine a grouping of said plurality of replicas into update domains for said update, said update domains being groups of said replicas;

querying said distributed application, by a fabric controller, to determine a sequence of said update domains to update;

for each of said update domains, processing, by the fabric controller according to the sequence, a current update domain to perform a first step of said update, said processing being performed as a first atomic operation and comprising;

querying said application executing in a current update domain, before transmission of said update for reception by said current update domain, to determine whether said application considers said current update domain ready for updating; and

if said application is not ready for updating, performing a repair on said application within said current update domain;

for each of said update domains, processing a current update domain to perform a second step of said update, said processing being performed as a second atomic operation and comprising;

after receiving a query response indicative of said current update domain being ready for updating, performing said update on said update domain;

after performing said update, querying said application executing in a current update domain to determine whether said current update domain is operating properly;

if said application is operating properly, completing said second step on said current update domain.

20. The method of claim 19, said update comprising implementing a topology change to said distributed application.