WO2018166474A1

WO2018166474A1 - Member group change method and apparatus, medium, and computer device

Info

Publication number: WO2018166474A1
Application number: PCT/CN2018/078989
Authority: WO
Inventors: 吴义谱; 张炎泼
Original assignee: 贵州白山云科技有限公司
Priority date: 2017-03-17
Filing date: 2018-03-14
Publication date: 2018-09-20
Also published as: CN107707595A; CN107707595B

Abstract

Disclosed are a member group change method and apparatus. The method comprises: receive a change request for an initial cluster member group, determine a target cluster member group according to the initial cluster member group and the change request, and set a transition cluster member group according to the initial cluster member group and the target cluster member group, the transition cluster member group comprising a first group and a second group, the first group comprising all elements in the initial cluster member group, and the second group comprising all elements in the target cluster member group; the initial cluster member group executes a consistency algorithm so as to update the initial cluster member group to the transition cluster member group; the transition cluster member group executes the consistency algorithm so as to update the transition cluster member group to the target cluster member group. According to embodiments of the present invention, the replacement of a plurality of elements in a member group can be finished once without suspending the cluster service, and the intelligence and the effectiveness of the cluster service can be effectively improved.

Description

Method, device, medium and computer equipment for changing member groups

This application claims the priority of the Chinese Patent Application, filed on March 17, 2017, to the Chinese Patent Office, Application No. 201710161757.2, entitled "A Member Group Change Method and Apparatus", the entire contents of which are incorporated herein by reference. In the application.

Technical field

The present invention relates to the field of Internet technologies, and in particular, to a method, device, medium, and computer device for changing a member group.

Background technique

The Paxos algorithm is a message-based consistency algorithm proposed by Leslie Lamport for solving distributed system consistency problems. Processes in distributed systems based on messaging communication models may stop due to errors and restart due to anomalies, messages in network transmissions may be delayed, lost or duplicated, regardless of possible tampering, ie Byzantine error Next, the Paxos algorithm solves the problem of how to agree on a value in a distributed system where the above anomalies may occur.

Participants in Paxos can act as at least one of the three roles of a proposer, an acceptor, and a learner. The proposer can propose a proposal, a proposal information including the proposal number n and the value of the proposal; the recipient is the processor of the proposal, and it has the right to decide whether to accept or reject the proposal, if the proposal receives the majority of the recipient Acceptance, the proposal is selected (chosen); learners can only learn the proposal that has been selected. The Paxos algorithm guarantees the correctness of the algorithm by satisfying security constraints and activity constraints. The active constraint means that a proposal is finally selected, and the security constraints include the following constraints:

1. Only proposals submitted by the proponent can be selected by the recipient;

2. Only one value can be selected (chosen) in a Paxos instance algorithm.

3. If a participant believes that a proposal has been selected, then the proposal must be the one that was actually selected.

Lamport obtained the Paxos algorithm by continuously strengthening the above three constraints, especially constraint 2. Lamport used a lot of space in the "Paxos make simple" paper to strengthen the constraint 2, thus obtaining the following stage submission process of the Paxos algorithm:

Stage 1a. The proposer determines the number n of the proposal and then sends a prepare request to the recipient, the request request carrying the number n of the proposal.

Stage 1b. After receiving the request, the recipient determines that the number n of the proposal is not less than the number of the proposal that it has responded to before, and the recipient responds to the request through the preparation and returns the value of the proposal with the highest number before (if present) At the same time, it is guaranteed that no further proposals with a number less than n will be passed.

Stage 2a. If the proposer receives a response from the majority recipient to the preparation request number n, then it sends an acceptance request including the proposal with the value v number n to the recipient, where v Is the value of the largest numbered proposal in the Phase 1b response, or the value that the proposer needs to submit in this paxos algorithm.

Phase 2b. The recipient receives the acceptance request for the proposal including the number n, and as long as it has not responded to the preparation request with the number greater than n, it can pass the proposal.

It is through the above two stages that Paxos completes a Paxos and determines that Paxos submits a value. Lamport also proved the correctness of the algorithm in his paper, but in practice there are many details of the Paxos algorithm that need to be further resolved, such as the Paxos membership group change problem.

So far, we have assumed that the cluster member group configuration is fixed. However, in practice, because the machine is down, the members are offline, and the replacement of new and old members is an inevitable problem. Then, after the member group configuration information is changed, how to maintain the consistency of the member group configuration information, Lamport is in the paper. No specific solution is given, but keeping the member group configuration information consistent in the cluster is a prerequisite for proper operation of Paxos. The current methods include the following two:

Method one, stop-the-word method

The stop-the-word method will suspend the cluster first, modify the configuration information of all members, and then restart the entire cluster. This method can solve the problem of consistency of configuration information of member groups, and will not cause the configuration information of different members to be inconsistent, but The problem revealed at the same time is also obvious, that is, the service cannot be provided during the suspension of the cluster. For the reality, service unavailability is unbearable in most cases.

Method 2, one-stage member change method

The first-stage member change method was originally a method of changing the member group proposed by the raft paper. The method is also applicable to Paxos. The member configuration information is changed by running the Paxos instance once, and the member configuration information after the change is required as the Paxos algorithm proposal. The value of the proposal allows the proposal to pass that value among the majority recipients. The one-stage member group change algorithm has a very extreme limitation, that is, only one member can be changed at a time, which means that only one member can be added or deleted at a time. For example, replacing one member requires two Paxos, first member Remove from the configuration and add new members to the configuration. The algorithm can also ensure the consistency of the member group configuration information, because changing one member at a time can ensure that the transition of the member configuration before the change and the member configuration after the change is not empty, thus ensuring that only One of the majority of the two member configurations is met so that only one value can be selected.

Disadvantages of the prior art include:

1. The stop-the-word method modifies the configuration by suspending the service provided by the cluster, and then provides the service after the configuration of the change completer. The obvious disadvantage of this method is that the service cannot be provided during the change member configuration process, and the availability is relatively high. This is unbearable for the service.

2, the one-stage member change method can change the member configuration information while providing the service, but only one member can be added or deleted at a time, meaning that if it is impossible to replace multiple members at the same time, it can only be replaced one by one. Services that are more frequently changed in member configurations greatly reduce their performance.

Summary of the invention

The following is an overview of the topics detailed in this document. This Summary is not intended to limit the scope of the claims.

In order to solve the above technical problem, the present invention provides a method, device, medium and computer device for changing a member group.

The member group changing method provided in the embodiment of the present invention includes:

Receiving a change request for the initial cluster member group, determining a target cluster member group according to the initial cluster member group and the change request, and setting a transition cluster member group according to the initial cluster member group and the target cluster member group, The transition cluster member group includes a first group and a second group, the first group includes all members in the initial cluster member group, and the second group includes all members in the target cluster member group;

Performing a consistency algorithm on the initial cluster member group, and updating the initial cluster member group to the transition cluster member group;

The transition cluster member group executes the consistency algorithm, and updates the transition cluster member group to the target cluster member group.

The above member group change method also has the following characteristics:

The initial cluster member group performs a consistency algorithm, and updating the initial cluster member group to the transition cluster member group includes: setting a first request proposal by a member that receives the change request in the initial cluster member group, to preset The incrementing rule sets the number of the first request proposal, sets the value of the first request proposal to the transition cluster member group, sets the version number of the execution of the consistency algorithm, and accepts all the values in the initial cluster member group. The member sends at least one request, the at least one request carrying the version number, the number of the first request proposal, and the value of the first request proposal; the initial cluster member group executes the consistency algorithm Until the majority of the initial cluster member group passes the first request proposal.

The above member group change method also has the following characteristics:

The transition cluster member group executes the consistency algorithm, and updating the transition cluster member group to the target cluster member group includes: setting a second request proposal by a member of the transition cluster member group, by a preset increment The rule sets the number of the second request proposal, sets the value of the second request proposal to the target cluster member group, sets the version number of the execution of the consistency algorithm, and assigns all members in the transition cluster member group as recipients Sending at least one request, the at least one request carrying the version number, the number of the second request proposal, and the value of the second request proposal; the transition cluster member group executing the consistency algorithm until the The majority of the transition cluster member group passes the second request proposal.

The above member group change method also has the following characteristics:

The step 2 further includes: the initial cluster member group executing the consistency algorithm until the majority of the initial cluster member group passes the first request proposal, and the member receiving the change request goes to the transition cluster All members in the member group send a notification request carrying a version number. After the member of the transition cluster member group determines that the version number in the notification request is greater than the current version number saved by the member, the first request proposal is learned. Updating the saved current version number to the version number in the notification request and updating the value of the saved proposal is the value of the first request proposal, and thus the transition cluster member group is successfully selected by the initial cluster member group.

The above member group change method also has the following characteristics:

The step 3 further includes: the transition cluster member group executing the consistency algorithm until the majority of the transition cluster member group passes the second request proposal, and setting a member of the second request proposal to the transition All members in the cluster member group send a notification request carrying a version number, and the member in the transition cluster member group determines that the version number in the notification request is greater than the current version number saved by the member in the transition cluster member group, and learns the second request proposal. After updating the saved current version number to the version number in the notification request and updating the value of the saved proposal to the value of the second request proposal, the target cluster member group is successfully selected by the transition cluster member group. set.

The above member group change method also has the following characteristics:

The step 2 further includes: after the initial cluster member group executes the consistency algorithm, the member that receives the change request in the initial cluster member group determines whether the majority of the initial cluster member group passes the first Requesting a proposal, if not, setting the number of the first request proposal by a preset incremental rule, and sending the first request proposal of the updated number to all members in the initial cluster, and executing in this manner until the initial cluster member The majority of the group passes the first request proposal.

The above member group change method also has the following characteristics:

The step 3 further includes: after the transition cluster member group executes the consistency algorithm, the member sending the request in the transition cluster member group determines whether the majority of the transition cluster member group passes the second request Proposal, if not, setting the number of the second request proposal by a preset increment rule, and sending the second request proposal of the update number to all members in the initial cluster, and executing in this manner until the initial cluster member group The majority sent the proposal through the second request.

The above member group change method also has the following characteristics:

The change request includes a member corresponding to the change type and the change type, and the change type includes adding and/or deleting.

The member group changing apparatus provided in the embodiment of the present invention is applied to a server, and includes:

a change request receiving module, configured to receive a change request for an initial cluster member group when the belonging server is a member of the initial cluster member group;

a determining module, configured to determine a target cluster member group according to the initial cluster member group and the change request, and set a transition cluster member group according to the initial cluster member group and the target cluster member group, where the transition cluster member group is configured The first group includes a first group and a second group, the first group includes all members in the initial cluster member group, and the second group includes all members in the target cluster member group;

An execution module, configured to execute the consistency algorithm until the initial cluster member group is updated to the transition cluster member group; and is further configured to execute the consistency algorithm again when the belonging server is a member of the transition cluster member group The transition cluster member group is updated to the target cluster member group.

The above member group changing device also has the following characteristics:

The execution module includes a first unit and a second unit;

The first unit is configured to: when the belonging server receives the member of the change request for the initial cluster member group, set a first request proposal, set a number of the first request proposal by using a preset incremental rule, and set the The value of the first request proposal is the transition cluster member group, setting a version number for executing the consistency algorithm, and sending at least one request to all members in the initial cluster member group as recipients, the at least one request Carrying the version number, the number of the first request proposal, and the value of the first request proposal; and further, after receiving the response of the other members in the initial cluster member group to the request, performing the Part of the role of the proposer in the consistency algorithm;

The second unit is configured to: when the belonging server is the other member of the initial cluster member group, in addition to receiving the member of the change request, send at least the member that is the proposer role in the initial cluster member group After a request, the part of the consistency algorithm that is the recipient role is executed.

The above member group changing device also has the following characteristics:

The execution module includes a first unit and a second unit;

The first unit is configured to: when the belonging server is a member of the sponsoring role in the transition cluster member group, set a second request proposal, set a number of the second request proposal by using a preset incremental rule, and set the The value of the second request proposal is the target cluster member group, setting a version number for executing the consistency algorithm, and sending at least one request to all members in the transition cluster member group as recipients, the at least one request Carrying the version number, the number of the second request proposal, and the value of the second request proposal; and further, after receiving the response of the other members in the transition cluster member group to the request, performing the Part of the role of the proposer in the consistency algorithm;

The second unit is configured to: when the belonging server is the transition cluster member group, in addition to the sending the other members of the at least one requested member, in the receiving the transition cluster member group as a proposer role After at least one request sent by the member, the part of the consistency algorithm that is the recipient role is executed.

The above member group changing device also has the following characteristics:

The first unit is further configured to: when the belonging server is a member of the sponsor role in the initial cluster member group, after the majority of the initial cluster member group passes the first request proposal, All members in the transition cluster member group send a notification request carrying the version number;

The second unit is further configured to: when the belonging server is a member of the learner in the transition cluster member group, determine that the version number in the notification request is greater than the saved current version number, and learn the first request. After the proposal is updated, the current version number saved is the version number in the notification request and the value of the updated proposal is updated to the value of the first request proposal.

The above member group changing device also has the following characteristics:

The first unit is further configured to: when the belonging server is a member of the sponsor role in the transition cluster member group, after the majority of the transition cluster member group passes the second request proposal, All members in the transition cluster member group send a notification request carrying the version number;

The second unit is further configured to: when the belonging server is a member of the learner in the transition cluster member group, determine that the version number in the notification request is greater than the saved current version number, and learn the second request. After the proposal is updated, the current version number saved is the version number in the notification request and the value of the updated proposal is updated to the value of the second request proposal.

The embodiment of the present invention further provides a computer readable storage medium, where the computer program stores a computer program, and when the program is executed by the processor, the steps of the foregoing method are implemented.

An embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and operable on the processor, where the processor implements the method when the program is executed step.

The embodiment of the present invention can complete the replacement of multiple members in one member group at one time without stopping the cluster service, thereby effectively improving the intelligence and effectiveness of the cluster service.

DRAWINGS

The accompanying drawings are intended to provide a further understanding of the embodiments of the embodiments of the invention Improper limitations. In the drawing:

1 is a flow chart of a method for changing a member group in an embodiment;

2 is a structural diagram of a member group changing device in the embodiment;

3 is a schematic diagram of a member group change in a specific embodiment.

detailed description

The embodiments of the present invention will be further described with reference to the drawings and specific embodiments.

1 is a flow chart of a method for changing a member group in an embodiment; the method includes:

Step 101: Receive a change request for the initial cluster member group, determine a target cluster member group according to the initial cluster member group and the change request, and set a transition cluster member group according to the initial cluster member group and the target cluster member group, where the transition cluster member group includes the first a group and a second group, the first group includes all members of the initial cluster member group, and the second group includes all members of the target cluster member group;

Step 102: The initial cluster member group executes a consistency algorithm, and the initial cluster member group is updated to a transition cluster member group.

Step 103: The transition cluster member group executes the consistency algorithm, and updates the transition cluster member group to the target cluster member group.

Step 102 includes: setting a first request proposal by a member that receives the change request in the initial cluster member group, setting a number of the first request proposal by using a preset incremental rule, and setting a value of the first request proposal as a transition cluster member group. Setting a version number for executing the consistency algorithm, and sending at least one request to all members in the initial cluster member group as recipients, at least one request carrying the version number, the number of the first request proposal, and the value of the first request proposal; The initial cluster member group executes the consistency algorithm until the majority of the initial cluster member group passes the first request proposal. The preset increment rule includes: the proposal number is a combination of a member identifier and a time stamp, and the time stamp increases with time.

In the same manner as the above, the step 103 includes: setting a second request proposal by a member of the transition cluster member group, setting a number of the second request proposal by using a preset incremental rule, and setting a value of the second request proposal as a target cluster member group. Setting a version number of the execution of the consistency algorithm; sending at least one request to all members of the transition cluster member group as recipients, at least one request carrying the version number, the number of the second request proposal, and the value of the second request proposal; The transition cluster member group executes the consistency algorithm until the majority of the transition cluster member group passes the second request proposal.

In the case where the proposal cannot be passed at one time, it is necessary to send the number of the update proposal multiple times and send the proposal multiple times. Specifically, the step 102 further includes: after the initial cluster member group executes the consistency algorithm, the member that receives the change request in the initial cluster member group determines whether the majority of the initial cluster member group passes the first request proposal, and if not, presets The incremental rule sets the number of the first request proposal, and sends the first request proposal of the update number to all members in the initial cluster, and executes in this manner until the majority of the initial cluster member group passes the first request proposal. Step 103 further includes: after the transition cluster member group executes the consistency algorithm, the first member of the transition cluster member group determines whether the majority of the transition cluster member group passes the second request proposal, and if not, sets the second with a preset incremental rule. Request the number of the proposal, send the second request proposal with the updated number to all members in the initial cluster, and execute in this way until the majority of the initial cluster member group passes the second request proposal.

The method also includes a learning process. Step 2 further includes: the initial cluster member group executing the consistency algorithm until the majority of the initial cluster member group passes the first request proposal, and the member receiving the change request sends a notification request carrying the version number to all members in the transition cluster member group. That is, the commit request, the member who is a learner in the transition cluster member group determines that the version number in the notification request is greater than the current version number saved by the user, and updates the current version number saved in the notification request after learning the first request proposal. The version number and the value of the updated proposal are the values of the first request proposal. At this point, the transition cluster member group is successfully selected by the initial cluster member group. The step 103 further includes: the transition cluster member group executes the consistency algorithm until the majority of the transition cluster member group passes the second request proposal, and then the member of the second request proposal sends a notification carrying the version number to all members of the transition cluster member group. The request is a commit request, and the member of the transition cluster member group as the learner determines that the version number in the notification request is greater than the current version number saved by the member, and updates the saved current version number to the notification request after learning the second request proposal. The version number and the value of the updated proposal are updated to the value of the second request proposal. At this point, the target cluster member group is successfully selected by the transition cluster member group.

In the method, the change request includes a member corresponding to the change type and the change type, and the change type includes adding and/or deleting.

The consistency algorithm in this method is typically the PAXOS algorithm.

2 is a structural diagram of a member group changing device in an embodiment, the device is applied to a server, and includes:

a determining module, configured to determine a target cluster member group according to the initial cluster member group and the change request, and set a transition cluster member group according to the initial cluster member group and the target cluster member group, where the transition cluster member group includes the first group and the second group, One group includes all members in the initial cluster member group, and the second group includes all members in the target cluster member group;

Execution module, which is used to execute the consistency algorithm until the initial cluster member group is updated to the transition cluster member group; and is also used to execute the consistency algorithm again until the transition cluster member group is updated to the target cluster when the belonging server is a member of the transition cluster member group Member group.

The execution module includes a first unit and a second unit.

When the server receives the member of the change request in the initial cluster member group, the first unit of the server is configured to set the first request proposal when the owner server receives the member of the change request in the initial cluster member group, to preset The incremental rule sets the number of the first request proposal, sets the value of the first request proposal to the transition cluster member group, sets the version number of the execution of the consistency algorithm, and sends at least one time to all members in the initial cluster member group as recipients. The request, at least once, carries the version number, the number of the first request proposal, and the value of the first request proposal; and is also used to execute the consistency algorithm as a proposal after receiving the response of the other members in the initial cluster member group to the request Part of the person's role;

When the server receives the member of the change request for the initial cluster member group, the second unit of the server is used to receive the initial cluster when the member belongs to the initial cluster member group except the other members of the member who received the change request. After at least one request sent by a member of the member group as a proposer role, the part of the consistency algorithm that acts as the recipient role is executed.

In the case where the proposal cannot be passed at one time, it is necessary to send the number of the update proposal multiple times and send the proposal multiple times. The first unit is configured to set a second request proposal when the belonging server is a member of the sponsor role in the transition cluster member group, set the number of the second request proposal by using a preset incremental rule, and set the value of the second request proposal. The target cluster member group sets the version number of the execution of the consistency algorithm, and sends at least one request to all members in the transition cluster member group as recipients, at least one request carrying the version number, the number of the second request proposal, and the second The value of the request proposal; also used to perform the part of the consensus algorithm as the proposer role after receiving responses from other members in the transition cluster member group. The second unit is configured to perform consistency after receiving at least one request sent by a member of the transition cluster member group as a member of the proposer role in the transitional cluster member group of the belonging server, in addition to sending the other members of the member at least once. The part of the algorithm that acts as the recipient's role.

Corresponds to the process of learning in this method. When the own server is a member of the initial cluster member group as the proposer role, the first unit is also used to pass the first majority of the initial cluster member group when the owning server is a member of the sponsor role in the initial cluster member group. After requesting the proposal, a notification request carrying the version number, that is, a commit request, is sent to all members of the transition cluster member group. The second unit is further configured to: when the belonging server is a member of the learner in the transition cluster member group, determine that the version number in the notification request is greater than the saved current version number, and update the current saved current after learning the first request proposal. The version number is the version number in the notification request and the value of the updated proposal is the value of the first request proposal.

Corresponds to the process of learning in this method. The owning server is a member of the transition cluster member group that is the sponsor role. The first unit is also used to send the version number to all members of the transition cluster member group after the majority of the transition cluster member group passes the second request proposal when the belonging server is a member of the sponsor role in the transition cluster member group. The notification request is the commit request. The second unit is further configured to: when the belonging server is a member of the learner in the transition cluster member group, determine that the version number in the notification request is greater than the saved current version number, and update the saved current after learning the second request proposal. The version number is the version number in the notification request and the value of the updated proposal is the value of the second request proposal.

Specific embodiment

3 is a schematic diagram of a member group change in a specific embodiment. As shown in FIG. 3, members a, b, and c are included in the group member group C_old. A list of members in this cluster is stored in members a, b, and c. Member a receives the change request, which includes two parts:

In the first part, the change type is increased, and the change target is d, e.

In the second part, the change type is delete, and the change target is a, b.

The target cluster after the cluster member group C_old is changed is C_new, and C_new includes members c, d, and e.

Set the transition cluster member group C_middle to {{a,b,c},{c,d,e}}.

The identifiers corresponding to the members a, b, c, d, and e are: 01, 02, 03, 04, 05.

The specific implementation part includes the following:

The last request proposal that member a responded to is numbered 1400000010_01, the last request proposal that member b responded to is numbered 1400000009_02, and the last request proposal that member c should pass is number 1400000008_03. The proposal number includes the number of seconds and the member ID in seconds from a fixed historical time to the current time. When a member compares numbers, when the timestamps are different, the size of the timestamps is compared to the number comparison result. When the timestamps are the same, the size of the member identifiers is compared as the number comparison result.

Member a runs the Paxos algorithm as the proposer, determines the first request proposal, determines the number of the request proposal as 1400000011_01, and determines the value of the request proposal as C_middle. Member a determines that the version number at which the Paxos algorithm is executed is, for example, T. Member a sends a preparation request to all members of C_old, which carries the version number T and the number of the first request proposal.

All members of C_old as acceptors receive this preparation request, for example, all members are recipients. After receiving the preparation request, member a determines that the number carried in the preparation request is 1400000011_01, which is greater than the number 1400000010_01 of the last request proposal that the member has responded to, and determines that the version number in the preparation request is greater than the version number saved by the member. With this proposal, the response to the preparation request is returned, and the number of the last requested request that was responded is updated to 1400000011_01, and the saved version number is updated to T. After receiving the preparation request, the member b determines that the number 1400000011_01 carried in the preparation request is greater than the number 1400000009_02 of the last request proposal that the member has responded to, and determines that the version number in the preparation request is greater than the version number saved by the member, and then passes the Proposal, returning a response to the preparation request, and updating the number of the last requested request that was responded to 1400000011_01, and updating the saved version number to T. After receiving the preparation request, the member c determines that the number carried in the preparation request is 1400000011_01 is greater than the number 140000008_03 of the last request proposal that the member has responded to, and determines that the version number in the preparation request is greater than the version number saved by the member, and then passes This proposal returns a response to the prepare request and updates the number of the last requested request that was responded to 1400000011_01.

After receiving the response from the majority of the C_old to the preparation request, the member a sends an acceptance request to the member as the acceptor, where the acceptance request carries the version number T, the number of the first request proposal 1400000011_01, and the value of the first request proposal C_middle . After receiving the acceptance request as a member of the acceptor in C_old, it judges that the version number of the currently accepted request is greater than its saved version number, and has not responded to the preparation request with the number greater than 1400000011_01, responds to the acceptance request, and updates The version number it holds is the version number in the acceptance request.

After member a receives the response from the majority of C_old to accept the request, it determines that the majority in C_old passes the first request proposal.

Member a sends a notification request to all members of C_middle. After learning the first request proposal as a member of the learner in C_middle, C_middle is successfully selected by C_old.

The update process of C_middle to C_new is performed in the same manner as above.

After the first Paxos algorithm is successfully run, the member a is still selected as the first member to run the Paxos algorithm. Member a, as the proposer, runs the Paxos algorithm for the second time, determines the second request proposal, determines the number of the request proposal as 1400000012_01, and determines that the value of the request proposal is C_new. Member a determines that the version number of the implementation of this Paxos algorithm is T+1. The member a sends a preparation request to all members of the C_middle, and the preparation request carries the version number T+1 and the number of the second request proposal.

All members of C_old who are recipients receive this preparation request, for example, all members are recipients. After receiving the preparation request, member a determines that the number carried in the preparation request is 1400000012_01, which is greater than the number 1400000011_01 of the last request proposal that the member has responded to, and determines that the version number in the preparation request is greater than the version number saved by the member. With this proposal, the response to the preparation request is returned, and the number of the last requested request that was responded is updated to 1400000012_01, and the saved version number is updated to T+1. After receiving the preparation request, the member b determines that the number 1400000012_01 carried in the preparation request is greater than the number 1400000011_01 of the last request proposal that the member has responded to, and determines that the version number in the preparation request is greater than the version number saved by the member, and then passes the The proposal returns a response to the preparation request and updates the number of the last requested proposal that was responded to 1400000012_01, updating the saved version number to T+1. After receiving the preparation request, the member c determines that the number carried in the preparation request is 1400000012_01 is greater than the number 1400000011_01 of the last request proposal that the member has responded to, and determines that the version number in the preparation request is greater than the version number saved by the member. This proposal returns a response to the prepare request and updates the number of the last requested request that was responded to 1400000012_01.

Member a receives the response to the preparation request from the majority in C_middle (the majority of C_middle is C_old majority and C_new), and then sends an acceptance request to the member who is the recipient, and carries the version number T in the acceptance request. +1, the number of the second request proposal 1400000012_01 and the value of the second request proposal C_new. After receiving the acceptance request, the member who is the recipient of C_middle determines that the version number in the current acceptance request is greater than the version number it has saved, and has not responded to the preparation request with the number greater than 1400000012_01, and responds to the request. And update its saved version number to accept the version number in the request.

After member a receives the response from the majority in C_middle to accept the request, it determines that the majority in C_middle passes the second request proposal.

Member a sends a notification request to all members of C_middle. After C_new learns the second request proposal as a member of the learner, C_new is successfully selected by C_middle.

Any step in the whole process of change can be interrupted without affecting the cluster to provide normal services. Since C-old and C-new are not required to intersect, all members can be changed at the same time, and the consistency of the algorithm is determined by Paxos. It is guaranteed by itself.

The invention can complete the replacement of multiple members in one member group at one time without stopping the cluster service, and can effectively improve the intelligence and effectiveness of the cluster service.

It should be understood by those skilled in the art that the present invention may be modified or equivalently substituted without departing from the spirit and scope of the invention.

Those of ordinary skill in the art will appreciate that all or some of the steps, systems, and functional blocks/units of the methods disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof. In a hardware implementation, the division between functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be composed of several physical The components work together. Some or all of the components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer readable medium, which may include computer storage media (or non-transitory media) and communication media (or transitory media). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and nonvolatile, implemented in any method or technology for storing information, such as computer readable instructions, data structures, program modules or other data. Sex, removable and non-removable media. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridge, magnetic tape, magnetic disk storage or other magnetic storage device, or may Any other medium used to store the desired information and that can be accessed by the computer. Moreover, it is well known to those skilled in the art that communication media typically includes computer readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave or other transport mechanism, and can include any information delivery media. .

Industrial applicability

In this paper, the replacement of multiple members in a member group can be completed at one time without stopping the cluster service, which can effectively improve the intelligence and effectiveness of the cluster service.

Claims

A method for changing a member group, comprising:

Receiving a change request for the initial cluster member group, determining a target cluster member group according to the initial cluster member group and the change request, and setting a transition cluster member group according to the initial cluster member group and the target cluster member group, The transition cluster member group includes a first group and a second group, the first group includes all members in the initial cluster member group, and the second group includes all members in the target cluster member group;

Performing a consistency algorithm on the initial cluster member group, and updating the initial cluster member group to the transition cluster member group;

The transition cluster member group executes the consistency algorithm, and updates the transition cluster member group to the target cluster member group.
A member group changing method according to claim 1, wherein

The initial cluster member group performs a consistency algorithm, and updating the initial cluster member group to the transition cluster member group includes: setting a first request proposal by a member that receives the change request in the initial cluster member group, to preset The incrementing rule sets the number of the first request proposal, sets the value of the first request proposal to the transition cluster member group, sets the version number of the execution of the consistency algorithm, and accepts all the values in the initial cluster member group. The member sends at least one request, the at least one request carrying the version number, the number of the first request proposal, and the value of the first request proposal; the initial cluster member group executes the consistency algorithm Until the majority of the initial cluster member group passes the first request proposal.
A member group changing method according to claim 1, wherein

The transition cluster member group executes the consistency algorithm, and updating the transition cluster member group to the target cluster member group includes: setting a second request proposal by a member of the transition cluster member group, by a preset increment The rule sets the number of the second request proposal, sets the value of the second request proposal to the target cluster member group, sets the version number of the execution of the consistency algorithm, and assigns all members in the transition cluster member group as recipients Sending at least one request, the at least one request carrying the version number, the number of the second request proposal, and the value of the second request proposal; the transition cluster member group executing the consistency algorithm until the The majority of the transition cluster member group passes the second request proposal.
The member group changing method according to claim 2, wherein

The step 2 further includes: the initial cluster member group executing the consistency algorithm until the majority of the initial cluster member group passes the first request proposal, and the member receiving the change request goes to the transition cluster All members in the member group send a notification request carrying a version number. After the member of the transition cluster member group determines that the version number in the notification request is greater than the current version number saved by the member, the first request proposal is learned. Updating the saved current version number to the version number in the notification request and updating the value of the saved proposal is the value of the first request proposal, and thus the transition cluster member group is successfully selected by the initial cluster member group.
A member group changing method according to claim 3, wherein

The step 3 further includes: the transition cluster member group executing the consistency algorithm until the majority of the transition cluster member group passes the second request proposal, and setting a member of the second request proposal to the transition All members in the cluster member group send a notification request carrying a version number, and the member in the transition cluster member group determines that the version number in the notification request is greater than the current version number saved by the member in the transition cluster member group, and learns the second request proposal. After updating the saved current version number to the version number in the notification request and updating the value of the saved proposal to the value of the second request proposal, the target cluster member group is successfully selected by the transition cluster member group. set.
The member group changing method according to claim 2, wherein

The step 2 further includes: after the initial cluster member group executes the consistency algorithm, the member that receives the change request in the initial cluster member group determines whether the majority of the initial cluster member group passes the first Requesting a proposal, if not, setting the number of the first request proposal by a preset incremental rule, and sending the first request proposal of the updated number to all members in the initial cluster, and executing in this manner until the initial cluster member The majority of the group passes the first request proposal.
A member group changing method according to claim 3, wherein

The step 3 further includes: after the transition cluster member group executes the consistency algorithm, the member sending the request in the transition cluster member group determines whether the majority of the transition cluster member group passes the second request Proposal, if not, setting the number of the second request proposal by a preset increment rule, and sending the second request proposal of the update number to all members in the initial cluster, and executing in this manner until the initial cluster member group The majority sent the proposal through the second request.
A member group changing method according to claim 1, wherein

The change request includes a member corresponding to the change type and the change type, and the change type includes adding and/or deleting.
A member group changing device, which is applied to a server, and includes:

a change request receiving module, configured to receive a change request for an initial cluster member group when the belonging server is a member of the initial cluster member group;

a determining module, configured to determine a target cluster member group according to the initial cluster member group and the change request, and set a transition cluster member group according to the initial cluster member group and the target cluster member group, where the transition cluster member group is configured The first group includes a first group and a second group, the first group includes all members in the initial cluster member group, and the second group includes all members in the target cluster member group;

An execution module, configured to execute the consistency algorithm until the initial cluster member group is updated to the transition cluster member group; and is further configured to execute the consistency algorithm again when the belonging server is a member of the transition cluster member group The transition cluster member group is updated to the target cluster member group.
A member group changing device according to claim 9, wherein:

The execution module includes a first unit and a second unit;

The first unit is configured to: when the belonging server receives the member of the change request for the initial cluster member group, set a first request proposal, set a number of the first request proposal by using a preset incremental rule, and set the The value of the first request proposal is the transition cluster member group, setting a version number for executing the consistency algorithm, and sending at least one request to all members in the initial cluster member group as recipients, the at least one request Carrying the version number, the number of the first request proposal, and the value of the first request proposal; and further, after receiving the response of the other members in the initial cluster member group to the request, performing the Part of the role of the proposer in the consistency algorithm;

The second unit is configured to: when the belonging server is the other member of the initial cluster member group, in addition to receiving the member of the change request, send at least the member that is the proposer role in the initial cluster member group After a request, the part of the consistency algorithm that is the recipient role is executed.
A member group changing device according to claim 9, wherein:

The execution module includes a first unit and a second unit;

The first unit is configured to: when the belonging server is a member of the sponsoring role in the transition cluster member group, set a second request proposal, set a number of the second request proposal by using a preset incremental rule, and set the The value of the second request proposal is the target cluster member group, setting a version number for executing the consistency algorithm, and sending at least one request to all members in the transition cluster member group as recipients, the at least one request Carrying the version number, the number of the second request proposal, and the value of the second request proposal; and further, after receiving the response of the other members in the transition cluster member group to the request, performing the Part of the role of the proposer in the consistency algorithm;

The second unit is configured to: when the belonging server is the transition cluster member group, in addition to the sending the other members of the at least one requested member, in the receiving the transition cluster member group as a proposer role After at least one request sent by the member, the part of the consistency algorithm that is the recipient role is executed.
A member group changing device according to claim 10, wherein:

The first unit is further configured to: when the belonging server is a member of the sponsor role in the initial cluster member group, after the majority of the initial cluster member group passes the first request proposal, All members in the transition cluster member group send a notification request carrying the version number;

The second unit is further configured to: when the belonging server is a member of the learner in the transition cluster member group, determine that the version number in the notification request is greater than the saved current version number, and learn the first request. After the proposal is updated, the current version number saved is the version number in the notification request and the value of the updated proposal is updated to the value of the first request proposal.
A member group changing device according to claim 11, wherein:

The first unit is further configured to: when the belonging server is a member of the sponsor role in the transition cluster member group, after the majority of the transition cluster member group passes the second request proposal, All members in the transition cluster member group send a notification request carrying the version number;

The second unit is further configured to: when the belonging server is a member of the learner in the transition cluster member group, determine that the version number in the notification request is greater than the saved current version number, and learn the second request. After the proposal is updated, the current version number saved is the version number in the notification request and the value of the updated proposal is updated to the value of the second request proposal.
A computer readable storage medium having stored thereon a computer program, the program being executed by a processor to perform the steps of the method of any one of claims 1-8.
A computer device, comprising: a memory, a processor, and a computer program stored on the memory and operable on the processor, the processor executing the program to implement the claims 1 to 8 The steps of any of the methods described.