WO2019119263A1

WO2019119263A1 - Distributed cluster split-brain processing method, apparatus, and device

Info

Publication number: WO2019119263A1
Application number: PCT/CN2017/117131
Authority: WO
Inventors: 关超
Original assignee: 海能达通信股份有限公司
Priority date: 2017-12-19
Filing date: 2017-12-19
Publication date: 2019-06-27

Abstract

Disclosed are a distributed cluster split-brain processing method, apparatus, and device. The method comprises: when running states of opposite parties cannot be sensed between two mutually connected service clusters, switching the two service clusters into a read-only mode; determining, according to the result of whether the two service clusters can be normally connected to a database, at most one service cluster having the right to initiate voting, the service cluster having the voting right initiating voting to a preset voter set, and the service cluster not having the right to initiate voting not initiating voting; and if the service cluster that initiates voting obtains a predetermined number of votes within a preset time period, switching the service cluster that initiates voting from the read-only mode to a normal service mode, notifying the service cluster that does not initiate voting in the two clusters of a voting result, and stopping the work of the service cluster that does not initiate voting. Therefore, by means of the method of the present invention, the occurrence of split-brain can be completely avoided, and the voting result is closer to an actual use situation.

Description

Method, device and device for processing distributed cluster brain split

Technical field

The present invention relates to the field of distributed clusters, and in particular, to a method, device and device for processing a distributed cluster brain split.

Background technique

In a highly available distributed cluster, there are two clusters that are mutually backed up. When two clusters are connected, one of the clusters serves the user, and the other cluster acts as a backup cluster and passes each other through the heartbeat link. Achieved the main and standby negotiations. However, when there is a heartbeat link failure in two clusters, only one of the two clusters can be in a viable state. That is, only one of the clusters can serve the user. If both are in a viable state, the user is provided with the user. Service, and the inability of the two to synchronize data, indicates that there is a situation of brain splitting, which will lead to confusion of user data. Specifically, there may be inconsistencies in user data stored in the two clusters, so It is impossible to know which user data stored in the cluster is valid data.

In the prior art, many methods for solving brain splitting are provided, but many methods fail to solve how to select a cluster of activities without disagreement to provide services for users, and when votes are not consistent between voters and users, voting results may be caused. Not available, and if a non-voting party fails, it will not be able to know the result of the vote, resulting in brain splitting.

Summary of the invention

In view of this, the embodiment of the present invention discloses a method, a device, and a device for processing a distributed cluster brain split, which solves the problem that the cluster that selects an activity cannot be provided for the user in the prior art, for the voter and the use. When the voting is inconsistent, the voting result may be unavailable, and if the non-voting party fails, the voting result may not be known, resulting in a problem of brain splitting.

An embodiment of the present invention provides a method for processing a distributed cluster brain split, and the method may include:

When the two service clusters connected to each other in the distributed system cannot detect the running status of the other party, the two service clusters are switched to the read-only mode.

Determining whether the connection between the two service clusters and the database is normal, and determining, according to a judgment result of whether the connection between the two service clusters and the database is normal, determining at most one of the two service clusters a business cluster that initiates voting rights;

The service cluster having the right to initiate voting initiates a vote to the set of voters; wherein the set of voters is determined from the use terminal and/or the arbitration terminal according to a preset selection rule; a terminal that performs network operation in any one of two service clusters, where the arbitration terminal is a server other than the use terminal in the network node;

Determining whether the business cluster that initiated the voting has obtained a predetermined number of votes within a preset time;

If the polling service cluster obtains a predetermined number of votes within a preset time, the polling service cluster is switched from the read-only mode to the normal service mode, and the voting result is notified that the two clusters are not initiated. Voting business cluster.

Optionally, determining, according to the judgment result that the connection between the two service clusters and the database is normal, determining, from the two service clusters, at most one service cluster having the right to initiate voting, includes:

If the connection between the two service clusters and the database is normal, each of the two service clusters detects whether another service cluster initiates voting;

When the first service cluster of the two service clusters does not detect that other service clusters initiate voting, the first service cluster has the right to initiate voting;

If only one of the two service clusters is properly connected to the database, the service cluster that is connected to the database has the right to initiate voting;

If the connection between the two service clusters and the database is abnormal, the two service clusters do not have the right to initiate voting.

Optionally, the service cluster with the right to initiate voting initiates a vote to the set of voters, including:

The business cluster that initiated the vote writes a voting record and a timestamp in the database.

Optionally, if the connection between the two service clusters and the database is normal, each of the two service clusters detects whether other service clusters initiate voting, including:

The first service cluster of the two service clusters queries other voting records and timestamps in the database;

When the voting records and time stamps of the other service clusters exist and the current time does not exceed the preset time starting from the time stamp, the other service clusters initiate voting;

When the voting records of the other service clusters do not exist, or the voting records and time stamps of the other service clusters exist and the current time exceeds a preset time starting from the timestamp, the other service clusters do not initiate voting.

Optionally, determining the set of voters from the using terminal and/or the arbitration terminal according to a preset selection rule, including:

Determining whether the number of used terminals satisfying the preset condition in the using terminal reaches a preset number of voting sets;

Determining a voter in the set of voters from the use terminal if the number of use terminals satisfying the preset condition exceeds the preset number of voter sets;

If the number of used terminals that satisfy the preset condition does not exceed the preset number of voters, the arbitration terminal is selected from the arbitration terminals that satisfy the preset condition to determine the voter together with the use terminal that meets the preset condition. set.

Optionally, the preset condition is:

The terminal is online for more than the preset time and the state is stable.

Optionally, the notifying the voting result to the service clusters in the two clusters that do not initiate voting, including:

Writing the voting result to the database;

The service cluster that has not initiated voting reads the voting result from the database.

Sending the voting result to a use terminal and/or an arbitration terminal connected to the service cluster that initiated the voting;

When the use terminal and/or the arbitration terminal are connected to the service cluster that has not initiated voting, the voting result is sent to the service cluster that has not initiated voting.

Optionally, after determining whether the service cluster that initiated the voting obtains the predetermined number of votes within a preset time, the method further includes:

If the service cluster that initiated the voting does not obtain the predetermined number of votes within a preset time, the service clusters that have not initiated the voting in the two service clusters initiate voting and return to the service cluster that performs the judgment to initiate the voting to obtain the preset time. The number of tickets booked.

Optionally, the predetermined number of votes is more than half of the votes in the set of voters.

Optionally, the number of votes obtained by the polling business cluster is a weighted sum of the voter votes.

Alternatively, the weight of the voter changes according to its importance.

Optionally, a business cluster that does not have the right to initiate voting does not initiate a vote.

The embodiment of the invention further provides a processing device for a distributed cluster brain split, the device comprising:

a first mode switching unit, configured to switch the two service clusters into a read-only mode when the two service clusters connected to each other in the distributed system cannot sense the running state of the other party;

a first determining unit, configured to determine whether the connection between the two service clusters and the database is normal;

a first determining unit, configured to determine, according to a judgment result that the connection between the two service clusters and the database is normal, at most one service cluster having a voting right from the two service clusters;

a first initiating voting unit, configured to initiate a voting to a set of voters by the service cluster having the right to initiate voting; wherein the set of voters is determined from the using terminal and/or the arbitration terminal according to a preset selection rule The use terminal is a terminal that performs network operation by connecting any one of the two service clusters, where the arbitration terminal is a server other than the use terminal in the network node;

a second determining unit, configured to determine whether the service cluster that initiated the voting obtains a predetermined number of votes within a preset time;

a second mode switching unit, configured to: if the service cluster that initiates voting obtains a predetermined number of votes in a preset time, switch the service cluster that initiates the voting from a read-only mode to a normal service mode;

The notification unit is configured to notify the service cluster that does not initiate voting in the two clusters if the service cluster that initiated the voting obtains a predetermined number of votes within a preset time.

Optionally, the first determining unit includes:

a detecting subunit, configured to: if each of the two service clusters is connected to the data, each of the two service clusters detects whether another service cluster initiates voting;

a first determining subunit, configured to: when the first service cluster of the two service clusters does not detect that another service cluster initiates voting, determining that the first service cluster has the right to initiate voting;

a second determining subunit, configured to determine that a service cluster that is normally connected to the database has the right to initiate voting if only one of the two service clusters is normally connected to the database;

And a third determining subunit, configured to determine that the two service clusters do not have the right to initiate voting if the connection between the two service clusters and the database is abnormal.

Optionally, the first initiating voting unit includes:

Write unit for writing a vote record and a time stamp in the database

Optionally, the detecting subunit includes:

Querying a subunit, configured to query a voting record and a timestamp of the service cluster in the database;

a determining subunit, configured to: when the voting record and the timestamp of the other service cluster exist and the current time does not exceed a preset time starting from the timestamp, determine that the other service cluster initiates voting; when the other When the voting record of the service cluster does not exist or the current time exceeds the preset time from the time stamp, it is determined that the other service cluster does not initiate voting.

Optionally, it also includes:

a third determining unit, configured to determine whether the number of terminals that meet the preset condition in the using terminal reaches a preset number of voting sets;

a second determining unit, configured to determine, from the user, a voter in the set of voters if the number of used terminals satisfying the preset condition exceeds the preset number of voters;

a third determining unit, configured to: if the number of used terminals that meet the preset condition does not exceed the preset number of voter sets, select an arbitration terminal from the arbitration terminal that meets the preset condition to determine with the use terminal of the preset condition The set of voters.

Optionally, the preset condition is:

The terminal is online for more than the preset time and the state is stable.

Optional, including:

Writing a subunit for writing the voting result to the database;

Reading a subunit for reading the voting result from the database for the business cluster that did not initiate voting.

Optional, including:

a first sending subunit, configured to send the voting result to a using terminal and/or an arbitration terminal connected to the service cluster that initiates the voting;

And a second sending subunit, configured to send the voting result to the service cluster that does not initiate voting when the using terminal and/or the arbitration terminal are connected to the service cluster that does not initiate voting.

Optionally, it also includes:

The second initiating voting unit is configured to: if the service cluster that initiates the voting does not obtain the predetermined number of votes in a preset time, the service cluster that has not initiated the voting in the two service clusters initiates voting and returns to the second determining unit.

Alternatively, the weight of the voter changes according to its importance.

Embodiments of the present invention provide a storage medium on which a program is stored, and when the program is executed by a processor, the distributed cluster splitting processing method is implemented.

An embodiment of the present invention provides a processing device for distributed cluster brain splitting, where the distributed cluster device includes: a memory for storing a program; a processor for running the program, when the processor runs In the program, the processor implements the processing method of the distributed cluster brain split according to the present invention.

In this embodiment, when the two service clusters that are connected to each other cannot sense the running state of the other party, the two service clusters are switched to the read-only mode, and according to whether the two service clusters and the database can be normally connected. As a result, it is determined that at most one service cluster having the right to initiate voting initiates a vote, and the service cluster without the right to initiate voting does not initiate voting, and the service cluster having the right to initiate voting initiates a vote to the set of voters, if within a preset time The predetermined number of votes is obtained, and the service cluster that initiated the voting is switched from the read-only mode to the normal service mode, the voting result is notified to the service clusters in the two clusters that have not initiated the voting, and the work of the service cluster that has not initiated the voting is stopped. In this way, in the case of ensuring that the data can be read, the occurrence of brain splitting is completely avoided; and since the voter in the set of voters is determined by using the terminal as a priority principle, the voting result is closer to the actual use. In addition, the voting result is the result of multi-path transmission, and the voting result can be known even if a part of the link that has not initiated the voting fails.

DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below. Obviously, the drawings in the following description are only It is an embodiment of the present invention, and those skilled in the art can obtain other drawings according to the provided drawings without any creative work.

1 is a schematic flowchart of a method for processing a distributed cluster brain split according to an embodiment of the present invention;

2 is a schematic flowchart of a method for determining a service cluster with voting rights according to an embodiment of the present invention;

3 is a schematic diagram of a cluster initiated voting in a distributed cluster according to an embodiment of the present invention;

4 is a schematic diagram of notifying another cluster after voting is completed in a distributed cluster according to an embodiment of the present invention;

FIG. 5 is a schematic structural diagram of a processing device for a distributed cluster brain split according to an embodiment of the present disclosure;

FIG. 6 is a schematic structural diagram of a processing device for a distributed cluster brain split according to an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

Referring to FIG. 1 , a schematic flowchart of a method for processing a distributed cluster brain split according to an embodiment of the present invention is shown. In this embodiment, the method may include:

S101: When the two service clusters connected to each other in the distributed system cannot detect the running status of the other party, the two service clusters are switched to the read-only mode.

In this embodiment, in a highly available distributed cluster, there may be two service clusters connected to each other, wherein a plurality of service clusters connected to each other may sense the state of each other through a certain mechanism, for example, through a heartbeat chain. The mode of the road senses whether the other party is in the normal running state. When a heartbeat link fault occurs, that is, the two service clusters connected to each other cannot sense the running status of the other party. Otherwise, when either of the two service clusters appears In the case of downtime, the two business clusters can no longer sense the other party's operating status. In addition, further, the two service clusters that are connected to each other may also be two service clusters that are mutually backed up.

In this embodiment, the read-only mode can be understood as that the terminal can only read data from the database through the service cluster, but cannot write data in the database, for example, cannot add or modify data in the database. In the read-only mode, the cluster can filter the service type at the agent entry, filter out other services except the data read service, that is, accept only the service that reads the data, and does not accept the data read service. Other services; or it can also be restricted in the database write service, that is, the business that does not perform database write data.

S102: Determine whether the connection between the two service clusters and the database is normal, and determine at most one of the two service clusters according to a judgment result of whether the connection between the two service clusters and the database is normal. a business cluster that initiates voting rights;

In this embodiment, it should be noted that a service cluster having the right to initiate voting may initiate a vote, and a service cluster that does not have the right to initiate voting may not initiate a vote.

In this embodiment, it is determined whether the connection between the two service clusters and the database is normal, and it can be understood that the two service clusters can read and write the database normally. Specifically, the two services may be determined. Whether the cluster can read the data in the database normally, and determine whether the two service clusters can normally write data in the database.

In this embodiment, in order to ensure that no brain splitting occurs, only one service cluster having the right to initiate voting can be determined at the same time. Referring to FIG. 2, as an example, S102 may specifically include:

S201: If the connection between the two service clusters and the database is normal, each of the two service clusters detects whether another service cluster initiates voting;

In this embodiment, the service cluster that initiates the voting needs to write the voting record and the timestamp in the database, so that any one of the service clusters can determine whether the service cluster has initiated voting according to the voting records and timestamps of other service clusters. Can include:

The first service cluster of the two service clusters queries the database for the voting records and time stamps of other service clusters;

When the voting records of the other service clusters do not exist, or the voting records and time stamps of the other service clusters exist and the current time exceeds a preset time starting from the time stamp, the other service clusters do not initiate voting.

Distance description: If the two service clusters connected to each other include: cluster A and cluster B, when the first cluster is cluster A, the other clusters are cluster B; if the first cluster is cluster B, the other clusters are cluster A.

The timestamp includes the status of the service cluster, and the running status of the service cluster may include: the health status of the service cluster, that is, whether the service cluster is in a live state, whether a vote is initiated, or the like. Therefore, the two service clusters can learn through the database whether other service clusters have initiated voting.

S202: When the first service cluster in the two service clusters does not detect that other service clusters initiate voting, the first service cluster has the right to initiate voting;

In this embodiment, when the connection between the two service clusters and the database is normal, it indicates that both service clusters can obtain the running status of the other party from the database, where a service cluster detects If other service clusters connected to the service cluster do not initiate voting, the service cluster has the right to initiate voting. If other service clusters connected to the service cluster have already initiated voting, the service cluster does not have the current state. The right to vote.

S203: If only one of the two service clusters is connected to the database, the service cluster that is connected to the database has the right to initiate voting.

In this embodiment, if only one of the two service clusters is in normal connection with the data, it indicates that another service cluster that is not connected to the database is abnormal, and the service cluster cannot be obtained from the database. The status of the interconnected service cluster, and the service cluster normally connected to the database cannot send the running status of the service cluster to the database, so the service cluster normally connected with the database has the right to initiate voting, and cannot The service cluster in which the database is normally connected does not have the right to vote.

For example, suppose that two service clusters connected to each other are cluster A and cluster B. If the cluster can be connected to the database normally, cluster B cannot be connected to the database. In this state, cluster B cannot obtain cluster A from the database. The running state, while cluster B can not send the running status to the database.

S204: If the connection between the two service clusters and the database is abnormal, the two service clusters do not have the right to initiate voting.

In this embodiment, when two service clusters that are connected to each other cannot be normally connected to the database, it indicates that neither of the two service clusters can obtain the running status of the other party from the database, and the two service clusters cannot Perform business normally. At this time, neither of them has the right to initiate a vote, that is, neither of them can initiate a vote.

S103: The service cluster with the right to initiate voting initiates a vote to the set of voters; wherein the set of voters is determined from the use terminal and/or the arbitration terminal according to a preset selection rule; a terminal that connects any one of the two service clusters for network operation, where the arbitration terminal is a server other than the use terminal in the network node;

In this embodiment, only one service cluster having the right to initiate voting is determined in S102, that is, only at most one service cluster having the right to initiate voting can initiate voting, and when a service cluster with voting rights is determined is determined. The service cluster with the right to initiate voting initiates a vote to a preset set of voters.

For example, as shown in Figure 3, two service clusters connected to each other are cluster A and cluster B. When cluster A and cluster B have heartbeat link failures, that is, the two parties cannot sense the other party's running status, first determine the cluster. The connection between A and the database and the connection between the cluster B and the database are normal; if both cluster A and cluster B can be connected to the database, both cluster A and cluster B have the right to initiate voting; then, cluster A and Cluster B determines whether the other party has initiated voting. If cluster B detects that cluster A has not initiated voting, cluster B is the cluster with the right to initiate voting, and then cluster B can initiate voting to the preset set of voters. Where, when determining the set of voters, if User5 and User6 are not frequently online users, User5 and User6 are considered to be users without voting rights, and therefore do not belong to the range considered by the voter set, as can be seen from the figure. The set of voters is User1, User2, User3, User4, Abiter1, Abiter2, and Abiter3, but User3 in the voter set may have a single point of failure, so voting cannot be performed, but even if there are a small number of terminals that cannot vote, it does not affect. The result of the vote.

In this embodiment, the voter in the set of voters may be pre-configured, for example, may be preset by a technician, or may be dynamically selected according to the user's situation; however, in either case, The priority of the user is required to be followed. Specifically, in S103, the set of voters is determined from the using terminal and the arbitration terminal according to a preset selection rule, including:

Determining a voter in the set of voters from the use terminal if the number of user terminals satisfying the preset condition exceeds the preset number of voter sets;

If the number of used terminals satisfying the preset condition does not exceed the preset number of voters, the arbitration terminal is selected from the arbitration terminal satisfying the preset condition to determine the vote together with the use terminal satisfying the preset condition. Collection.

In this embodiment, it should be noted that the preset condition may be that the online time of the use terminal exceeds a preset time and the state is stable. For example, it can be understood that the use terminal is often online and stable in state, wherein the frequent online can be understood as a network operation often performed by two interconnected service clusters mentioned in S101, such as a fixed collection station of a public security system, a bank. Server, etc.

In this embodiment, when the online time of the terminal exceeds the preset time and the number of states is stable, the set of voters may be determined from the use terminal; when the online time of the use terminal exceeds the preset time and the number of states is stable, For the voter set determination, one part may be determined from the user terminal, and some part is determined from the arbitration terminal; when the use terminal whose online time exceeds the preset time and the state is stable, all the voter sets may be determined from the arbitration terminal. . Therefore, even if a part of the use terminal fails, the execution of the embodiment is not affected, that is, even if a part of the use terminal fails, the service cluster can initiate voting.

In this embodiment, it should be noted that the user terminal may be a terminal that performs network operations by connecting any one of the two service clusters mentioned above, for example, it may be understood that the terminal is used to log in to the two service clusters. Any one of the terminals that provide the network service by the service cluster in which the terminal is logged in. In addition, the network node also includes many frequently-on-line servers, but these servers do not need to be connected to any of the two service clusters for network operations. Such servers can be arbitrator terminals, voting in the voter set. The person can also be determined from the arbitrator terminal.

S104: Determine whether the service cluster that initiated the voting obtains a predetermined number of votes within a preset time.

In this embodiment, the predetermined number of votes may be the majority of the votes in the set of voters, and more specifically, as an example, may be more than half of the votes in the set of voters. Moreover, the number of voters in the set of voters may be an odd number or an even number. When the voting is initiated, the business cluster that initiated the voting may obtain more than half of the votes in the set of voters within a preset time. Successful for voting.

Specifically, the number of votes obtained by the polling business cluster is a weighted sum of the voter votes, and the weight of the voter changes according to importance, and the importance here is the importance of the voter.

In practical applications, after a business cluster with voting rights initiates a vote, some of the voters may fail. These failed business clusters no longer participate in voting, but as long as the failed business cluster is larger than the voter set. More than half of them will not affect the voting results.

S105: If the service cluster that initiates voting obtains a predetermined number of votes within a preset time, the service cluster that initiates the voting is switched from the read-only mode to the normal service mode, and the voting result is notified to the two clusters. A business cluster that did not initiate a vote.

For example, as shown in FIG. 3, if cluster B initiates voting and obtains more than half of the votes in the set of voters within a preset time, it indicates that the voting is successful. At this time, cluster B can be considered as the primary cluster. Cluster A is the backup cluster of cluster B. It can be understood that after B initiates a successful vote, B is a cluster that provides normal services by using the terminal, and A stops the service at this time.

In this embodiment, for the normal service mode, it can be understood that the cluster that successfully initiates the voting can connect to the terminal normally, provide network services for the use terminal, and can perform read and write operations in the database.

In this embodiment, the service cluster that notifies the two clusters that the voting is not initiated may specifically include the following two implementation manners:

Embodiment 1: Write the voting result into the database; the service cluster that has not initiated voting reads the voting result from the database.

Embodiment 2: transmitting the voting result to a use terminal and/or an arbitration terminal connected to the service cluster that initiates voting; when the use terminal and/or the arbitration terminal are connected to the service cluster that does not initiate voting The voting result is sent to the service cluster that has not initiated voting.

In this embodiment, the voting result is sent to a voter connected to the service cluster that initiated the vote and a use terminal other than the voter, wherein the voter uses the terminal and/or the arbitration to vote for the service cluster that initiated the vote. terminal.

In this embodiment, when the operation of notifying the service cluster in the two clusters that are connected to each other is performed, the solution of the first embodiment may be executed, and the solution of the second embodiment may be executed. Embodiment 1 and Embodiment 2 are simultaneously performed. However, in this embodiment, a preferred embodiment is to perform both the first embodiment and the second embodiment. In this way, when the service cluster that has not initiated the voting can connect to the server, the result of the voting can be known by reading the running status of the service cluster that initiated the voting in the server, even if the party that did not initiate the voting is unable to connect with the database normally. It is also possible to know the voting result of the other party from the use terminal on the service cluster connection that has not initiated the voting.

For example, as shown in Figure 4, if cluster B is a service cluster that initiates voting, cluster A is a service cluster that does not initiate voting. After voting ends, cluster B can inform the database of the voting result and the voters connected to cluster B. User terminals other than voters, where User1, User2, User3, User4, Arbiter1, Arbiter2, and Arbiter3 are voters, and User5 and User6 are non-voting user terminals, for example, can be used other than voters. Terminal, where User3 may be a use terminal that is not connected to cluster B, and the reason for not connecting with cluster B may be a single point of failure, or a network failure, and then a database that can receive the voting result The voter or the use terminal other than the voter informs the cluster A of the vote result. However, if the party that initiated the voting times out, the service cluster that initiated the voting does not obtain more than half of the votes in the set of voters within a preset time, and the voting service cluster is not initiated in the two service clusters. The set of voters initiates a vote and returns to execution S104. It should be noted that the service clusters that have not initiated voting in the two service clusters can initiate voting. The premise that the service clusters that have not initiated voting can be normally connected with the database.

Referring to FIG. 5, a schematic structural diagram of a processing device for a distributed cluster brain split according to an embodiment of the present invention is shown. The device may include:

The first mode switching unit 501 is configured to switch the two service clusters into a read-only mode when the two service clusters connected to each other in the distributed system cannot detect the running state of the other party;

The first determining unit 502 is configured to determine whether the connection between the two service clusters and the database is normal;

a first determining unit 503, configured to determine, according to a judgment result that the connection between the two service clusters and the database is normal, at most one service cluster having the right to initiate voting from the two service clusters;

a first initiating voting unit 504, configured to initiate a voting to a set of voters by the service cluster having the right to initiate voting; wherein the set of voters is determined from the using terminal and/or the arbitration terminal according to a preset selection rule The use terminal is a terminal that performs network operation by connecting any one of the two service clusters, and the arbitration terminal is a server of the network node other than the use terminal;

The second determining unit 505 is configured to determine whether the service cluster that initiated the voting obtains a predetermined number of votes within a preset time period;

The second mode switching unit 506 is configured to switch the service cluster that initiates the voting from the read-only mode to the normal service mode, if the service cluster that initiates the voting obtains a predetermined number of votes within a preset time;

The notification unit 507 is configured to notify the service cluster that does not initiate voting in the two clusters if the service cluster that initiated the voting obtains a predetermined number of votes within a preset time.

Optionally, the first determining unit includes:

a second determining subunit, configured to: if only one of the two service clusters is normally connected to the database, determine that the service cluster that is normally connected to the database has the right to initiate voting;

Optionally, the first initiating voting unit includes:

Write unit for writing a vote record and a time stamp in the database

Optionally, the detecting subunit includes:

Optionally, it also includes:

Optionally, the preset condition is:

The terminal is online for more than the preset time and the state is stable.

Optionally, the notification unit includes:

Writing a subunit for writing the voting result to the database;

Optionally, the notification unit includes:

Optionally, it also includes:

Alternatively, the weight of the voter changes according to its importance.

With the apparatus of the embodiment, in the case where the data can be read, the occurrence of brain splitting is completely avoided; and since the voter in the set of voters is determined based on the principle of using the terminal, the result of the voting More close to the actual use situation; in addition, the voting result is the result of multi-path transmission, even if the part of the link that has not initiated the voting fails, the voting result can be known.

FIG. 6 is a schematic structural diagram of a processing device for a distributed cluster brain splitting according to an embodiment of the present invention. In this embodiment, the device includes a memory 601 and a processor 602.

Wherein, the memory 601 is configured to store a program;

The processor 602 is configured to run the program. Specifically, when the processor executes the program, the following steps are implemented:

Optionally, the preset condition is:

The terminal is online for more than the preset time and the state is stable.

Writing the voting result to the database;

Optionally, the notifying the voting result to the service cluster in the two clusters that does not initiate voting, including:

Alternatively, the weight of the voter changes according to its importance.

The device in this embodiment may be a server, a PC, a PAD, a mobile phone, or the like.

The memory in the device of this embodiment may include a non-permanent memory, a random access memory (RAM), and/or a non-volatile memory in a computer-readable storage medium, such as a read-only memory or a flash memory (falsh). RAM), the memory includes at least one memory chip.

With the device of the embodiment, in the case where the data can be read, the occurrence of brain splitting is completely avoided; and since the voter in the set of voters is determined based on the principle of using the terminal, the voting result is More close to the actual use situation; in addition, the voting result is the result of multi-path transmission, even if the part of the link that has not initiated the voting fails, the voting result can be known.

It should be noted that each embodiment in the specification is described in a progressive manner, and each embodiment focuses on differences from other embodiments, and the same similar parts between the embodiments are referred to each other. can.

The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments are obvious to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but the scope of the invention is to be accorded

Claims

A method for processing a distributed cluster brain split, characterized in that the method comprises:

When the two service clusters connected to each other in the distributed system cannot detect the running status of the other party, the two service clusters are switched to the read-only mode.

Determining whether the connection between the two service clusters and the database is normal, and determining, according to a judgment result of whether the connection between the two service clusters and the database is normal, determining at most one of the two service clusters a business cluster that initiates voting rights;

The service cluster having the right to initiate voting initiates a vote to the set of voters; wherein the set of voters is determined from the use terminal and/or the arbitration terminal according to a preset selection rule; a terminal that performs network operation in any one of two service clusters, where the arbitration terminal is a server other than the use terminal in the network node;

Determining whether the business cluster that initiated the voting has obtained a predetermined number of votes within a preset time;

If the polling service cluster obtains a predetermined number of votes within a preset time, the polling service cluster is switched from the read-only mode to the normal service mode, and the voting result is notified that the two clusters are not initiated. Voting business cluster.
The method according to claim 1, wherein at least one of the two service clusters is determined to have the right to initiate voting according to a judgment result of whether the connection between the two service clusters and the database is normal. Business clusters, including:

If the connection between the two service clusters and the database is normal, each of the two service clusters detects whether another service cluster initiates voting;

When the first service cluster of the two service clusters does not detect that other service clusters initiate voting, the first service cluster has the right to initiate voting;

If only one of the two service clusters is properly connected to the database, the service cluster that is connected to the database has the right to initiate voting;

If the connection between the two service clusters and the database is abnormal, the two service clusters do not have the right to initiate voting.
The method of claim 1, wherein the service cluster having the right to initiate voting initiates a vote to the set of voters, including:

The business cluster that initiated the vote writes a voting record and a timestamp in the database.
The method according to claim 3, wherein if the connection between the two service clusters and the database is normal, each of the two service clusters detects whether another service cluster is initiated. Vote, including:

The first service cluster of the two service clusters queries the database for the voting records and time stamps of other service clusters;

When the voting records and time stamps of the other service clusters exist and the current time does not exceed the preset time starting from the time stamp, the other service clusters initiate voting;

When the voting records of the other service clusters do not exist, or the voting records and time stamps of the other service clusters exist and the current time exceeds a preset time starting from the timestamp, the other service clusters do not initiate voting.
The method according to claim 1, wherein determining the set of voters from the use terminal and/or the arbitration terminal according to a preset selection rule comprises:

Determining whether the number of used terminals satisfying the preset condition in the using terminal reaches a preset number of voting sets;

Determining a voter in the set of voters from the use terminal if the number of use terminals satisfying the preset condition exceeds the preset number of voter sets;

If the number of used terminals that satisfy the preset condition does not exceed the preset number of voters, the arbitration terminal is selected from the arbitration terminals that satisfy the preset condition to determine the voter together with the use terminal that meets the preset condition. set.
The method of claim 5 wherein said predetermined condition is:

The terminal is online for more than the preset time and the state is stable.
The method according to claim 1, wherein the notifying the voting result to the service cluster that does not initiate voting in the two clusters comprises:

Writing the voting result to the database;

The service cluster that has not initiated voting reads the voting result from the database.
The method according to claim 1, wherein the notifying the voting result to the service cluster that does not initiate voting in the two clusters comprises:

Sending the voting result to a use terminal and/or an arbitration terminal connected to the service cluster that initiated the voting;

When the use terminal and/or the arbitration terminal are connected to the service cluster that has not initiated voting, the voting result is sent to the service cluster that has not initiated voting.
The method according to claim 1, wherein the determining whether the service cluster initiating the voting has obtained the predetermined number of votes within a preset time further comprises:

If the service cluster that initiated the voting does not obtain the predetermined number of votes within a preset time, the service clusters that have not initiated the voting in the two service clusters initiate voting and return to the service cluster that performs the judgment to initiate the voting to obtain the preset time. The number of tickets booked.
The method of claim 1 wherein said predetermined number of votes is more than half of said votes in said set of voters.
The method of claim 1 wherein the number of votes obtained by the polling service cluster is a weighted sum of said voter votes.
The method of claim 11 wherein the voter's weight changes according to its importance.
The method of claim 1 wherein the business cluster that does not have the right to initiate voting does not initiate voting.
A distributed cluster brain splitting processing device, the device comprising:

a first mode switching unit, configured to switch the two service clusters into a read-only mode when the two service clusters connected to each other in the distributed system cannot sense the running state of the other party;

a first determining unit, configured to determine whether the connection between the two service clusters and the database is normal;

a first determining unit, configured to determine, according to a judgment result that the connection between the two service clusters and the database is normal, at most one service cluster having a voting right from the two service clusters;

a first initiating voting unit, configured to initiate a voting to a set of voters by the service cluster having the right to initiate voting; wherein the set of voters is determined from the using terminal and/or the arbitration terminal according to a preset selection rule The use terminal is a terminal that performs network operation by connecting any one of the two service clusters, where the arbitration terminal is a server other than the use terminal in the network node;

a second determining unit, configured to determine whether the service cluster that initiated the voting obtains a predetermined number of votes within a preset time;

a second mode switching unit, configured to: if the service cluster that initiates voting obtains a predetermined number of votes in a preset time, switch the service cluster that initiates the voting from a read-only mode to a normal service mode;

The notification unit is configured to notify the service cluster that does not initiate voting in the two clusters if the service cluster that initiated the voting obtains a predetermined number of votes within a preset time.
The apparatus according to claim 14, wherein the first determining unit comprises:

a detecting subunit, configured to: if each of the two service clusters is connected to the data, each of the two service clusters detects whether another service cluster initiates voting;

a first determining subunit, configured to: when the first service cluster of the two service clusters does not detect that another service cluster initiates voting, determining that the first service cluster has the right to initiate voting;

a second determining subunit, configured to: if only one of the two service clusters is normally connected to the database, determine that the service cluster that is normally connected to the database has the right to initiate voting;

And a third determining subunit, configured to determine that the two service clusters do not have the right to initiate voting if the connection between the two service clusters and the database is abnormal.
The apparatus according to claim 14, wherein said first initiating voting unit comprises:

A write unit for writing a vote record and a timestamp in the database.
The device according to claim 16, wherein the detecting subunit comprises:

Querying a subunit, configured to query a voting record and a timestamp of the service cluster in the database;

a determining subunit, configured to: when the voting record and the timestamp of the other service cluster exist and the current time does not exceed a preset time starting from the timestamp, determine that the other service cluster initiates voting; when the other When the voting record of the service cluster does not exist or the current time exceeds the preset time from the time stamp, it is determined that the other service cluster does not initiate voting.
The device according to claim 14, further comprising:

a third determining unit, configured to determine whether the number of terminals that meet the preset condition in the using terminal reaches a preset number of voting sets;

a second determining unit, configured to determine, from the user, a voter in the set of voters if the number of used terminals satisfying the preset condition exceeds the preset number of voters;

a third determining unit, configured to: if the number of used terminals that meet the preset condition does not exceed the preset number of voter sets, select an arbitration terminal from the arbitration terminal that meets the preset condition to determine with the use terminal of the preset condition The set of voters.
The device according to claim 18, wherein said preset condition is:

The terminal is online for more than the preset time and the state is stable.
The device according to claim 14, wherein the notification unit comprises:

Writing a subunit for writing the voting result to the database;

Reading a subunit for reading the voting result from the database for the business cluster that did not initiate voting.
The device according to claim 15, wherein the notification unit comprises:

a first sending subunit, configured to send the voting result to a using terminal and/or an arbitration terminal connected to the service cluster that initiates the voting;

And a second sending subunit, configured to send the voting result to the service cluster that does not initiate voting when the using terminal and/or the arbitration terminal are connected to the service cluster that does not initiate voting.
The device according to claim 14, further comprising:

The second initiating voting unit is configured to: if the service cluster that initiates the voting does not obtain the predetermined number of votes in a preset time, the service cluster that has not initiated the voting in the two service clusters initiates voting and returns to the second determining unit.
The apparatus according to claim 14, wherein said predetermined number of votes is more than half of said votes in said set of voters.
The apparatus of claim 14 wherein the number of votes obtained by the polling service cluster is a weighted sum of said voter votes.
The apparatus of claim 24 wherein the weight of the voter changes according to its importance.
The apparatus of claim 14 wherein the service cluster that does not have the right to initiate voting does not initiate voting.
A computer readable storage medium having stored thereon a computer program that, when executed by a processing device, implements the method of claims 1-13.
A processing device for a distributed cluster brain split, characterized in that the device comprises:

Memory for storing programs;

a processor for running the program, the processor implementing the distributed cluster splitting processing method of claims 1-13 when the processor runs the program.