Merge protocol for clustered computer system

• US 6,968,359 B1
• Filed: 08/14/2000
• Issued: 11/22/2005
• Est. Priority Date: 08/14/2000
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Miller, Robert, Thayib, Kiswanto, Morey, Vicki Lynn, Williams, Laurie Ann
• Primary Examiner(s): Thompson, Marc D.
• Assistant Examiner(s): Gerezgiher, Yemane M.
• Application Number:
  US09/638,328
• Time in Patent Office: 1,926 Days
• Field of Search: 709/230, 709/100, 709/102, 709/103, 709/104, 709/105, 709/106, 709/217, 709/219, 709/225, 709/201, 709/204, 709/205, 709/213, 709/215, 709/248, 714/15, 718/106
• US Class Current: 709/205
• CPC Class Codes: G06F 2209/505 Clust
• CPC Class Codes: G06F 9/5061 Partitioning or combining of resources

Abstract

An apparatus, program product and method utilize ordered messages in a clustered computer system to defer the execution of a merge protocol in a cluster group until all pending protocols in each partition of a group are handled, typically by ensuring either cancellation or completion of each pending protocol prior to execution of the merge protocol. From the perspective of each group member, the execution of the merge protocol is deferred by inhibiting processing of the merge request by such member until after processing of all earlier-received pending requests has been completed.

Claims

1. A method of processing a merge request received by a member of a group in a clustered computer system, the method comprising:

(a) receiving a merge request in the member, the merge request configured to initiate a recombination of a plurality of partitions of the group in the clustered computer system; and

(b) inhibiting processing of the merge request by the member until after processing of all earlier-received pending requests by the member is completed.

8. The method of claim 1, further comprising

(a) receiving the merge request in a plurality of additional members in the group;

and

(b) locally inhibiting processing of the merge request within each of the plurality of additional members until after processing of all earlier-received pending requests by each such member is completed.

2. A method of processing a merge request received by a member of a group in a clustered computer system, the method comprising:

(a) receiving a merge request in the member,

(b) inhibiting processing of the merge request by the member until after processing of all earlier-received pending requests by the member is completed;

(c) assigning a reception identifier to each request received by the member to indicate a relative order of reception of each received request by the member, including assigning a reception identifier to the merge request indicating reception of the merge request after reception of all earlier-received pending requests received by the member; and

(d) processing each received request in order according to the reception identifiers assigned thereto, wherein inhibiting processing of the merge request comprises processing the merge request after completing processing of each earlier-received pending request.

3. The method of claim 2, wherein processing each received request includes, for each received request:

(a) comparing the reception identifier assigned to such received request with a reception identifier for a last received merge request; and

(b) canceling such received request if the reception identifier assigned to such received request indicates earlier reception of such received request relative to the reception identifier for the last received merge request.

4. The method of claim 3, wherein comparing the reception identifier assigned to such received request with the reception identifier for the last received merge request includes comparing the reception identifier assigned to such received request with a stored last merge reception identifier, the method further comprising, in response to reception of the merge request, updating the stored last merge reception identifier to that of the merge request.

5. The method of claim 3, wherein canceling such received request includes returning a cancel completion message to an originator of such received request.

6. The method of claim 3, wherein processing each received request further includes, for each received request, executing a protocol associated with such received request if the reception identifier assigned to such received request does not indicate earlier reception of such received request relative to the reception identifier for the last received merge request.

7. A method of processing a merge request received by a member of a group in a clustered computer system, the method comprising:

(a) receiving a merge request in the member;

(b) inhibiting processing of the merge request by the member until after processing of all earlier-received pending requests by the member is completed;

(c) receiving a message other than a request;

(d) determining whether a request handle associated with the message matches that of a pending request for the member; and

(e) discarding the message if the request handle does not match that of a pending request for the member.

9. A method of processing a request received by a member of a group in a clustered computer system, the method comprising:

(a) assigning a reception identifier to a received request, the reception identifier indicating a relative order of reception of the received request;

(b) comparing the reception identifier assigned to the received request with a reception identifier for a last received merge request; and

(c) canceling the received request if the reception identifier assigned to the received request indicates earlier reception of the received request relative to the reception identifier for the last received merge request.

10. The method of claim 9, wherein comparing the reception identifier assigned to the received request with the reception identifier for the last received merge request includes comparing the reception identifier assigned to the received request with a stored last merge reception identifier, the method further comprising:

(a) determining whether the received request is a merge request; and

(b) if so, updating the stored last merge reception identifier to that of the merge request.

11. The method of claim 9, wherein canceling the received request includes returning a cancel completion message to an originator of the received request.

12. The method of claim 9, further comprising executing a protocol associated with the received request if the reception identifier assigned to the received request does not indicate earlier reception of the received request relative to the reception identifier for the last received merge request.

13. The method of claim 9, further comprising:

(a) receiving a message other than a request;

(b) determining whether a request handle associated with the message matches that of a pending request for the member; and

(c) discarding the message if the request handle does not match that of a pending request for the member.

14. An apparatus, comprising:

(a) a memory; and

(b) a program resident in the memory, the program configured to process a merge request received by a member of a group in a clustered computer system by inhibiting processing of the merge request until after processing of all earlier-received pending requests is completed, wherein the merge request is configured to initiate a recombination of a plurality of partitions of the group in the clustered computer system.

21. The apparatus of claim 14, wherein the program is further configured to receive the merge request in a plurality of additional members in the group, and locally inhibit processing of the merge request within each of the plurality of additional members until after processing of all earlier-received pending requests by each such member is completed.

22. The apparatus of claim 14, wherein the apparatus is a computer in the clustered computer system.

15. An apparatus, comprising:

(a) a memory; and

(b) a program resident in the memory, the program configured to process a merge request received by a member of a group in a clustered computer system by inhibiting processing of the merge request until after processing of all earlier-received pending requests is completed, wherein the program is further configured to assign a reception identifier to each request received by the member to indicate a relative order of reception of each received request by the member, wherein a reception identifier is assigned to the merge request indicating reception of the merge request after reception of all earlier-received pending requests received by the member, and wherein the program is further configured to process each received request in order according to the reception identifiers assigned thereto such that the program inhibits processing of the merge request by processing the merge request after completing processing of each earlier-received pending request.

16. The apparatus of claim 15, wherein the program is configured to process each received request by comparing the reception identifier assigned to such received request with a reception identifier for a last received merge request, and canceling such received request if the reception identifier assigned to such received request indicates earlier reception of such received request relative to the reception identifier for the last received merge request.

17. The apparatus of claim 16, wherein the program is configured to compare the reception identifier assigned to such received request with the reception identifier for the last received merge request by comparing the reception identifier assigned to such received request with a stored last merge reception identifier, and wherein the program is further configured to update the stored last merge reception identifier to that of the merge request in response to reception of the merge request.

18. The apparatus of claim 16, wherein the program is configured to cancel such received request by returning a cancel completion message to an originator of such received request.

19. The apparatus of claim 16, wherein the program is configured to process each received request further by, for each received request, executing a protocol associated with such received request if the reception identifier assigned to such received request does not indicate earlier reception of such received request relative to the reception identifier for the last received merge request.

20. An apparatus, comprising:

(a) a memory; and

(b) a program resident in the memory, the program configured to process a merge request received by a member of a group in a clustered computer system by inhibiting processing of the merge request until after processing of all earlier-received pending requests is completed, wherein the program is further configured to receive a message other than a request, determine whether a request handle associated with the message matches that of a pending request for the member, and discard the message if the request handle does not match that of a pending request for the member.

23. A clustered computer system, comprising:

(a) a plurality of nodes coupled to one another over a network;

(b) a plurality of member jobs defining a group and configured to be executed by at least one of the plurality of nodes; and

(c) a program configured to be executed by at least one of the plurality of nodes to process a merge request received by a member job from the plurality of member jobs by inhibiting processing of the merge request until after processing of all earlier-received pending requests is completed, wherein the merge request is configured to initiate a recombination of a plurality of partitions of the group in the clustered computer system.

24. A program product, comprising:

(a) a program configured to process a merge request received by a member of a group in a clustered computer system by inhibiting processing of the merge request until after processing of all earlier-received pending requests is completed, wherein the merge request is configured to initiate a recombination of a plurality of partitions of the group in the clustered computer system; and

(b) a computer readable medium upon which the program is tangibly embodied.