Method for controlling group membership in a distributed multinode data processing system to assure mutually symmetric liveness status indications

• US 7,043,550 B2
• Filed: 02/15/2002
• Issued: 05/09/2006
• Est. Priority Date: 02/15/2002
• Status: Active Grant

• Current Assignee: X Corp. (f/k/a Twitter, Inc.) (X Holdings Corp.)
• Original Assignee: International Business Machines Corporation
• Inventors: Knop, Felipe, Lin, Tseng-Hui
• Primary Examiner(s): Dharia, Rupal
• Assistant Examiner(s): Gillis, Brian J.
• Application Number:
  US10/077,051
• Publication Number:
  US 20030158936A1
• Time in Patent Office: 1,544 Days
• Field of Search: 709/224, 709/230, 709/225, 709/249, 709/227, 714/25, 714/47
• US Class Current: 709/224
• CPC Class Codes: G06F 2209/505 Clust
• CPC Class Codes: H04L 67/10 in which an application is distributed across nodes in the network software deployment G06F8/60; multiprogramming arrangements G06F9/46
• CPC Class Codes: H04L 69/329 in the application layer [OSI layer 7]
• CPC Class Codes: H04L 69/40 for recovering from a failure of a protocol instance or entity, e.g. service redundancy protocols, protocol state redundancy or protocol service redirection management of faults, events, alarms or notifications in data switching networks H04L41/06
• CPC Class Codes: H04L 9/40 Network security protocols

Abstract

In distributed multinode data processing systems, mechanisms are employed to insure that the nodes are continually informed about the liveness of the other nodes within node groups in the network and across networks. The method of the present invention employs the concept of node stability which it uses to provide indications of membership in a previous stable group as part of messages transmitted in a protocol for forming, maintaining and dissolving node groupings in such a way as to assure that all of the nodes in the group have a consistent indication of liveness status for all of the nodes within any given group of nodes.

Claims

1. In a distributed, multinode data processing system in which nodes communicate via adapters coupled to each node, a method for detecting node reachability inconsistencies in the presence of temporary communication failure or temporary daemon blockage, said method comprising:

upon detecting at a first node conditions suitable for instituting change in adapter group membership, sending to a recipient node, a message indicative of previous membership of said first node in a stable adapter membership group upon the condition that said recipient node is also indicated as being part of said stable group; and

initiating a group join protocol in which said message is rejected by a recipient node under a condition selected from the group of conditions consisting of (1) said message includes indicia showing that the sending node and the recipient node are in the same adapter membership class, said indicia being in conflict with corresponding indicia stored within said recipient node; and

(2) said message includes indicia showing that the sending node and the recipient node are in different adapter membership classes, said indicia being in conflict with corresponding indicia stored within said recipient node.

2. The method of claim 1 in which said conditions include failure to detect heartbeat messages at said first node.

3. The method of claim 2 further including a step in which said first node, after its first message is rejected, transmits a second message which is indicative of nonmembership in a previously stable group, whereby the goal of having the sender and receiver agree on adapter group membership is advanced.

4. The method of claim 3 further including the step of processing said second message so as to provide mutually consistent views of the liveness of nodes in the membership group.

5. A multinode data processing system comprising:

a plurality of data processing nodes connected in a network capable of transmitting messages between nodes;

storage mails within said nodes containing program code for transmitting from a sending node to a recipient node, upon detecting, at said sending node, conditions suitable for instituting change in adapter group membership, a message indicative of previous membership of said sending node in a stable adapter membership group upon the condition that said recipient node is also indicated as being part of said stable group and for initiating a group join protocol in which said message is rejected by a recipient node under a condition selected from the group of conditions consisting of (1) said message includes indicia showing that the sending node and the recipient node are in the same adapter membership class, said indicia being in conflict with corresponding indicia stored within said recipient node; and

(2) said message includes indicia showing that the sending node and the recipient node are in different adapter membership classes, said indicia being in conflict with corresponding indicia stored within said recipient node.

6. The data processing system of claim 5 in which said program code further includes means for transmitting from said first node, after its first message is rejected, a second message which is indicative of nonmembership in a previously stable group, whereby the goal of having the sender and receiver agree on adapter group membership is advanced.

7. The data processing system of claim 6 in which said program code further includes means for processing said second message so as to provide mutually consistent views of the liveness of nodes in the membership group.

8. A program product comprising machine readable medium containing program code for use in a multinode data processing system for transmitting to a recipient node, upon detecting at a first node, conditions suitable for instituting change in adapter group membership, a message indicative of previous membership of said first node in a stable adapter membership group upon the condition that said recipient node is also indicated as being part of said stable group and for initiating a group join protocol in which said message is rejected by a recipient node under a condition selected from the group of conditions consisting of (1) said message includes indicia showing that the sending node and the recipient node are in the same adapter membership class, said indicia being in conflict with corresponding indicia stored within said recipient node;

and (2) said message includes indicia showing that the sending node and the recipient node are in different adapter membership classes, said indicia being in conflict with corresponding indicia stored within said recipient node.

9. The program product of claim 8 further including means for transmitting from said first node, after its first message is rejected, a second message which is indicative of nonmembership in a previously stable group, whereby the goal of having the sender and receiver agree on adapter group membership is advanced.

10. The program product of claim 9 further including means for processing said second message so as to provide mutually consistent views of the liveness of nodes in the membership group.