CN108400999A - A kind of load-balancing method and device of data-base cluster mirror nodes - Google Patents

Abstract

The present invention relates to a kind of load-balancing method and device of data-base cluster mirror nodes, this method includes：Obtain the configuration parameter of data-base cluster；Wherein, the configuration parameter includes the host node example quantity of cabinet quantity, the number of servers of each cabinet and each server；Following operation is executed for each cabinet in the data-base cluster：According to the configuration parameter, the corresponding mirror nodes example of all host node examples of current cabinet is evenly distributed in the every other cabinet in addition to current cabinet, to realize the load balancing of data-base cluster mirror nodes.Technical solution provided in an embodiment of the present invention, the corresponding mirror nodes example of the host node example of current cabinet is dispersed in other different cabinets, even if in the case where current cabinet breaks down, network pressure can also be dispersed in different cabinets, and entire cluster can be allowed to be in the state of load balancing.

Description

A load balancing method and device for database cluster mirror nodes

technical field

The invention relates to the technical field of databases, in particular to a load balancing method and device for database cluster mirror nodes.

Background technique

Greenplum is a relational database cluster, which is actually a logical database composed of several independent database services. Different from the share-disking (shared disk) architecture of Oracle RAC, Greenplum adopts the Shared-Nothing (shared nothing) architecture. The entire cluster consists of a control node (Master Host) and multiple data nodes (Segment Host) (here can be understood as storage Multiple machines for data), where multiple databases can run on each data node.

Most Greenplum database clusters are used on relatively cheap X86 architecture PC servers, which will inevitably lead to an increase in hardware failure rate in the process of reducing usage costs. In order to better avoid database service failures caused by software and hardware failures Greenplum also provides the Mirror (mirror) protection mechanism of the database layer, that is, the data of each database node is mirrored synchronously in another node, that is, when the primary node (Primary) is working normally, the cluster manager will master The data of the node is always synchronized to the mirror node (Mirror). When the data node where the master node is located has a software and hardware failure and loses service capabilities, the mirror node immediately takes over all services of the master node, so the failure of a single node does not affect the use of the entire system.

Take a cluster of one hundred servers as an example: the layout of cabinets and servers is shown in Figure 1. Due to the large number of servers, the number of database instances in the Greenplum Database cluster (referred to as GP cluster) will also be relatively large. The above environment There are 98 data nodes, and each data node has 4 master node instances, that is, there are 392 master node instances and mirror node instances respectively. Therefore, the distribution of the mirror nodes of the database cluster also needs to fully consider the physical location of the server. In order to ensure the high availability of the database cluster, the availability of the GP will not be affected if a cabinet fails, and the data node Primary instance and the corresponding Mirror instance are also deployed in separate cabinets. In addition, the Primary instance and the corresponding Mirror instance use two cabinets as a group to form a mirror relationship with each other.

Each server deploys 4 Primary instances and 4 Mirror instances. Primary instances and mirror instances are distributed on servers in different cabinets. For example, the mirror instances corresponding to the primary instances of all servers in cabinet 1 are evenly distributed on all servers in cabinet 2, as shown in Figure 2. Similarly, the mirror instances corresponding to the primary instances of all servers in cabinet 2 are evenly distributed on all servers in cabinet 1. The schematic diagram of the deployment of cabinets 1 and 2 is shown in Figure 3.

As shown in Figure 4, if a server in cabinet 1 fails and goes down, the four primary instances on it will be taken over by the four servers in cabinet 2 respectively. Even if the overall failure of cabinet 1 occurs, all the primary instances on it can be taken over by the server in cabinet 2. However, when a cabinet as a whole fails, such as a cabinet switch failure, the mirrored cabinet will bear twice the network pressure of the original design, and the server may crash. In addition, all mirror nodes need to be manually configured, which is inefficient. In the process of deploying Greenplum clusters, especially large-scale clusters, due to the large number of machines and storage in different racks, the network environment between different racks is different, so how to more effectively design the distribution of mirror nodes to ensure that in Keeping the entire cluster in a balanced state when a failure occurs has become a difficult problem for the DBA (database administrator).

Contents of the invention

The technical problem to be solved by the present invention is to provide a load balancing method for database cluster mirror nodes, so as to solve the defect in the prior art that when a whole cabinet fails, the mirror cabinet will bear twice the network pressure of the original design.

For this purpose, the present invention proposes a kind of load balancing method of database cluster mirror node, comprising:

Obtain configuration parameters of the database cluster; wherein, the configuration parameters include the number of cabinets, the number of servers in each cabinet, and the number of master node instances of each server;

Perform the following operations for each cabinet in the database cluster: according to the configuration parameters, evenly distribute the mirror node instances corresponding to all the master node instances of the current cabinet to all other cabinets except the current cabinet, so as to realize the database cluster Load balancing of mirror nodes.

Optionally, the following operation is performed for each cabinet in the database cluster: according to configuration parameters, evenly distribute mirror node instances corresponding to all master node instances of the current cabinet to all other cabinets except the current cabinet, including:

Perform the following operations for each cabinet: according to the number of master node instances in the current cabinet and the number of cabinets, evenly distribute the mirror node instances corresponding to the master node instance of the current cabinet to all other cabinets except the current cabinet;

Perform the following operations for each cabinet: according to the number of mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, evenly distribute the mirror node instances allocated to the current cabinet to different servers in the current cabinet.

Optionally, the following operation is performed for each cabinet: according to the number of master node instances of the current cabinet and the number of cabinets, evenly distribute the mirror node instances corresponding to the master node instance of the current cabinet to all other cabinets except the current cabinet In the cabinet, including:

Number the mirror node instances corresponding to the master node instances of all cabinets;

For each cabinet, perform the following operations: according to the number of the mirror node instance corresponding to the master node instance of the current cabinet and the number of the cabinets, evenly distribute the numbers of the mirror node instances corresponding to all the master node instances of the current cabinet to the other cabinets except the current cabinet In other cabinets, the number data of the mirror node instance assigned to each cabinet is finally obtained.

Optionally, according to the numbers of the mirror node instances corresponding to the master node instances of the current cabinet and the number of cabinets, the numbers of the mirror node instances corresponding to all the master node instances of the current cabinet are evenly distributed to other cabinets except the current cabinet. In the cabinet, the following hash function is used for uniform distribution:

HashFunc(X)=M%(N-1)

Among them, X is the hash address, which represents the cabinet number M is assigned to; M is the number of the mirror node instance corresponding to all the master node instances of the current cabinet, and N is the number of cabinets.

Optionally, the following operations are performed for each cabinet: according to the number of mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, evenly distribute the mirror node instances allocated to the current cabinet to different servers in the current cabinet, including :

For each cabinet, perform the following operations: according to the number data of the mirror node instance assigned to the current cabinet and the number of servers in the current cabinet, evenly distribute the number of the mirror node instance assigned to the current cabinet to different servers in the current cabinet, and finally determine the current cabinet The number data of the mirror node instance assigned by each server.

Optionally, according to the number data of the mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, the numbers of the mirror node instances allocated to the current cabinet are evenly distributed to different servers in the current cabinet, specifically using the following hash function Make an even distribution:

HashFunc(Y)=K%I

Among them, Y is the hash address, which represents the server number K is allocated to; K is the number of the mirror node instance allocated to the current cabinet, and Y is the number of servers in the current cabinet.

On the other hand, the embodiment of the present invention also provides a load balancing device for database cluster mirror nodes, including:

A parameter acquisition module, configured to acquire configuration parameters of the database cluster; wherein the configuration parameters include the number of cabinets, the number of servers in each cabinet, and the number of master node instances of each server;

The uniform distribution module is used to perform the following operations for each cabinet in the database cluster: according to the configuration parameters, evenly distribute the mirror node instances corresponding to all the master node instances of the current cabinet to all other cabinets except the current cabinet, To achieve load balancing of database cluster mirror nodes.

Optionally, the uniform distribution module includes:

The cabinet allocation unit is configured to perform the following operations for each cabinet: according to the number of master node instances of the current cabinet and the number of cabinets, evenly distribute the mirror node instances corresponding to the master node instance of the current cabinet to all other cabinets except the current cabinet in the cabinet;

The server allocation unit is configured to perform the following operations for each cabinet: according to the number of mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, evenly distribute the mirror node instances allocated to the current cabinet to different servers in the current cabinet.

Optionally, the cabinet distribution unit includes:

The numbering subunit is used to number the mirror node instances corresponding to the master node instances of all cabinets;

The cabinet number allocation subunit is used to perform the following operations for each cabinet: according to the number of the mirror node instance corresponding to the master node instance of the current cabinet and the number of cabinets, assign the mirror node instances corresponding to all master node instances of the current cabinet The number is evenly distributed to other cabinets except the current cabinet, and finally the number data of the mirror node instance assigned to each cabinet is obtained;

Optionally, the server allocation unit is specifically used for:

For each cabinet, perform the following operations: according to the number data of the mirror node instance assigned to the current cabinet and the number of servers in the current cabinet, evenly distribute the number of the mirror node instance assigned to the current cabinet to different servers in the current cabinet, and finally determine the current cabinet The number data of the mirror node instance assigned by each server.

The load balancing method and device for database cluster mirror nodes provided by the embodiments of the present invention evenly distribute the mirror node instances corresponding to all the master node instances of the current cabinet to the cabinets other than the current cabinet according to the configuration parameters of the database cluster. In all other cabinets, the mirror node instances corresponding to the primary node instance of the current cabinet are scattered in other different cabinets. Even if the current cabinet fails, the network pressure can be distributed in different cabinets, and will not be single Centralized in the mirror cabinet, when the current cabinet fails, the entire cluster can also be in a load-balanced state; in addition, the allocation of mirror node instances does not depend on the experience of maintenance personnel, and the reasonable distribution of mirror node instances is carried out according to configuration parameters. There will be no human error, etc., which ensures the rationalization of the distribution of mirror node instances and avoids problems.

Description of drawings

The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the invention in any way. In the accompanying drawings:

FIG. 1 is a schematic diagram of placement of existing cabinets and servers;

FIG. 2 is a schematic diagram showing that the mirror node instances corresponding to the master node instance of the cabinet 1 are distributed in the cabinet 2 in the prior art;

FIG. 3 is a schematic diagram of the distribution of mirror node instances corresponding to the master node instance of the prior art cabinet 2 in the cabinet 1;

FIG. 4 is a schematic diagram of the principle of taking over by the cabinet 2 after a certain server in the cabinet 1 fails in the prior art;

5 is a schematic flow diagram of a load balancing method for a database cluster mirror node provided by an embodiment of the present invention;

FIG. 6 is a schematic structural diagram of a load balancing device for a mirror node of a database cluster provided by an embodiment of the present invention;

FIG. 7 is a schematic flowchart of a load balancing method for database cluster mirror nodes provided by another embodiment of the present invention

FIG. 8 is a schematic numbering diagram of mirroring node instances of three cabinets provided by an embodiment of the present invention;

FIG. 9 is a schematic diagram of the results of uniform distribution of mirror node instances based on cabinets according to an embodiment of the present invention;

FIG. 10 is a schematic diagram of the results of uniform distribution of mirror node instances based on the server according to the embodiment of the present invention;

11 is a schematic structural diagram of a load balancing device for a mirror node of a database cluster provided by an embodiment of the present invention;

FIG. 12 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention.

Detailed ways

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

As shown in Figure 5, the embodiment of the present invention provides a load balancing method for database cluster mirror nodes, the method includes the following steps:

S1: Obtain configuration parameters of the database cluster; wherein, the configuration parameters include the number of cabinets, the number of servers in each cabinet, and the number of master node instances for each server;

The load balancing method of the mirror node of the database cluster provided by the embodiment of the present invention may take the load balancing device of the mirror node of the database cluster as the execution subject. As shown in Figure 6, the device can be divided into three layers, task scheduling layer, model layer and analysis layer. The task scheduling layer is mainly responsible for initiating tasks, and judging and processing the input configuration parameters, that is, judging whether the number of cabinets and servers are within a reasonable range, and if unreasonable, terminate the task and report an error. The model layer is mainly responsible for digitizing configuration parameters, converting machines, cabinets, and instances in the production process into digital numbers that can be quantified. After the digital processing is completed, the data analysis layer is responsible for evenly distributing mirror node instances to each cabinet according to the number of cabinets included in the configuration parameters. Inside the cabinet, re-assign mirror node instances evenly to each server.

S2: Perform the following operations for each cabinet in the database cluster: according to the configuration parameters, evenly distribute the mirror node instances corresponding to all the master node instances of the current cabinet to all other cabinets except the current cabinet, so as to realize the database cluster Load balancing of mirror nodes.

It should be noted that in the prior art, when the whole cabinet 1 fails, all the master node instances on it will be taken over by the servers in cabinet 2 (mirror cabinet). Therefore, cabinet 2 will bear twice the network pressure of the original design, May cause the server in rack 2 to crash. In the embodiment of the present invention, according to the number of cabinets designed for database management, the number of servers in each cabinet, and the number of master node instances in each server, the mirror node instances corresponding to all master node instances in the current cabinet are evenly distributed to all cabinets except the current cabinet. In other cabinets, so that when the whole cabinet 1 fails, the network pressure will not be concentrated in the cabinet 2, but distributed to all other cabinets except the cabinet 1. Even if cabinet 1 fails as a whole, other non-faulty cabinets will not experience a sharp increase in network pressure, ensuring that the entire database cluster is still in a load-balanced state when a fault occurs. For example, when the number of cabinets is 3, since the mirror node instances corresponding to the primary node instance of cabinet 1 are evenly distributed to cabinet 2 and cabinet 3, if cabinet 1 fails, the servers in cabinet 2 and cabinet 3 will Taking over, the network pressure is no longer concentrated in the cabinet 2, thus realizing the load balancing of the mirror nodes of the entire database cluster.

The load balancing method of the mirror node of the database cluster provided by the embodiment of the present invention evenly distributes the mirror node instances corresponding to all the master node instances of the current cabinet to all other cabinets except the current cabinet according to the configuration parameters of the database cluster, thereby Distribute the mirror node instances corresponding to the master node instance of the current cabinet in other different cabinets. Even if the current cabinet fails, the network pressure can be distributed in different cabinets, and will not be concentrated in the mirror cabinet. When the current cabinet fails, the entire cluster can also be in a load-balanced state; in addition, the allocation of mirror node instances does not depend on the experience of maintenance personnel, and the reasonable distribution of mirror node instances is carried out according to configuration parameters, without human errors, etc. This ensures the rational distribution of mirror node instances and avoids problems.

On the basis of the foregoing embodiments, step S2 specifically includes:

S201: Perform the following operation for each cabinet: according to the number of master node instances of the current cabinet and the number of cabinets, evenly distribute the mirror node instances corresponding to the master node instance of the current cabinet to all other cabinets except the current cabinet;

S202: Perform the following operation for each cabinet: according to the number of mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, evenly distribute the mirror node instances allocated to the current cabinet to different servers in the current cabinet.

Specifically, as shown in FIG. 7 , the database administrator defines tasks and inputs configuration parameters (the number of cabinets, the number of servers, and the number of instances designed for each server). The task allocation layer performs task allocation and judges the rationality of the input configuration parameters. How to terminate the task unreasonably and report an error. If it is reasonable, the model layer will digitize the input configuration parameters, that is, the cabinets, servers, and instances can be converted into digital numbers that can be quantified, and the digital numbers can be imported into the database. Afterwards, the data analysis layer can cycle through the assignment operations based on the generated numbers. First, you can evenly distribute based on the cabinet, and for each cabinet in turn, evenly distribute the mirror node instances corresponding to the current cabinet to other cabinets. A distribution list of all image node instances in each cabinet can be generated as needed. Then, for each cabinet in turn, evenly distribute the mirror node instances assigned by the current cabinet to different servers in the current cabinet, and finally complete the uniform distribution of all image node instances, and generate all image node instances in each cabinet and each server as needed The corresponding distribution list can achieve load balancing even when a certain cabinet fails as a whole.

On the basis of the above embodiments, step S201 specifically includes:

S201-1: Number the mirror node instances corresponding to the master node instances of all cabinets;

S201-2: Perform the following operation for each cabinet: according to the number of the mirror node instance corresponding to the master node instance of the current cabinet and the number of cabinets, evenly distribute the numbers of the mirror node instances corresponding to all the master node instances of the current cabinet to In other cabinets except the current cabinet, the number data of the mirror node instance assigned to each cabinet is finally obtained.

For example, suppose there are 9 servers in cabinet 1, and one server has 4 master node instances, the corresponding mirror node instances can be numbered as 001, 002, 003, 004, and cabinet 1 has 001, 002...036 number, and so on. According to the generated number, the distribution is performed cyclically, and the mirror node instances corresponding to this cabinet are evenly distributed to all other cabinets that are not in this cabinet. As a result, each cabinet has a number that is evenly distributed to the mirror node instances in the cabinet. The numbers of the cabinets can be reordered to generate a corresponding list of numbers.

Taking three cabinets and three servers in each cabinet as an example, according to the obtained configuration parameters, number the mirror node instances corresponding to the master node instances of all cabinets in sequence, as shown in Figure 8, the master node instance numbers are 001 and 002 , 003...036, and the corresponding mirror node instance numbers are 001, 002, 003...036. According to the number of cabinets, you can distribute the IDs of all mirror node instances in cabinet 1 to cabinets 2 and 3. That is, the numbers (001-012) of the 12 mirror node instances corresponding to cabinet 1 are evenly distributed to cabinet 2 and cabinet 3. By analogy, evenly distribute the number (013-024) of the mirror node instance corresponding to cabinet 2 to cabinet 1 and cabinet 3, and evenly distribute the number (025-036) of the mirror node instance corresponding to cabinet 3 to cabinet 1 and cabinet 2 on.

Optionally, in step S201-2, the following hash (HASH) function may be used to achieve uniform distribution:

HashFunc(X)=M%(N-1)

Among them, X is the hash address, which represents the cabinet number M is assigned to; M is the number of the mirror node instance corresponding to all the master node instances of the current cabinet, and N is the number of cabinets.

Specifically, the Hash function is used for uniform distribution. The primary key M is the number corresponding to the mirror node instance, N is the number of cabinets, and the primary key is used as the sharding attribute. Take the remainder obtained after dividing the keyword by a number not greater than the length N of the hash table as the hash address. For example: the number of cabinets here is N=3, if M is 001, the hash address is 1, and HashFunc(1)=001%(3-1). Taking the three servers in cabinet 1 as an example, the mirror node instance numbers are (001 002 003 004) (005 006 007 008) (009010 011 012). When cabinet 1 goes down, cabinet 2 and cabinet 3 are numbered respectively 1, 0, instance 001 obtains the hash address 1 through the Hash function, so instance 001 is assigned to number 1 (cabinet 2). By analogy, after the even distribution of mirror node instances in cabinet 1 is completed, the loop is executed until the mirror node instances of each cabinet are evenly distributed in all other cabinets, and the result is shown in Figure 9. It can be seen from Figure 9 that the mirror node instances corresponding to the master node instance of cabinet 1 are evenly distributed to other cabinets (cabinet 2 and cabinet 3), so that when cabinet 1 loses service capability, cabinet 2 and cabinet 3 take over and jointly Undertake network pressure and achieve load balancing.

Among them, after the mirror node instances of all cabinets are distributed, the mirror node instances stored in each cabinet can be sorted again to form a corresponding numbered list. Taking cabinet 1 as an example, a corresponding sorted list is formed as shown in Table 1 below.

Table 1 Numbering of mirror node instances assigned to cabinet 1

On the basis of the above embodiments, step S202 may specifically include:

For each cabinet, perform the following operations: according to the number data of the mirror node instance assigned to the current cabinet and the number of servers in the current cabinet, evenly distribute the number of the mirror node instance assigned to the current cabinet to different servers in the current cabinet, and finally determine the current cabinet The number data of the mirror node instance assigned by each server.

It should be noted that after step S201, each cabinet has the number of mirror node instances evenly assigned to this cabinet. In each cabinet, since there are multiple servers, taking cabinet 1 as an example, the The numbers of the mirror node instances in the cabinet are assigned to different servers in the cabinet in sequence according to the order in Table 1 above. And so on, until the number of the mirror node instance in each cabinet is evenly distributed among different servers in the cabinet.

Optionally, according to the number data of the mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, the numbers of the mirror node instances allocated to the current cabinet are evenly distributed to different servers in the current cabinet. Specifically, the following hash function can be used to implement Even distribution:

HashFunc(Y)=K%I

Among them, Y is the hash address, which represents the server number K is allocated to; K is the number of the mirror node instance allocated to the current cabinet, and Y is the number of servers in the current cabinet.

Specifically, the Hash function is used for uniform distribution, the primary key K is the number corresponding to the mirror node instance, I is the number of servers in the current cabinet, and the primary key is used as the fragmentation attribute. After the cabinet-based image node instance allocation is completed as shown in Figure 9, take cabinet 1 as an example, number the three servers in cabinet 1 as 0, 1, and 2, and refer to the sorting in Table 1, and sequentially assign the numbers in cabinet 1 013, 015, 017.....035 are evenly distributed to different servers according to the above hash function. For example, HashFunc(1)=013%3, the mirror node instance numbered 013 is assigned to the server numbered 1, such as HashFunc(2)=017%3, the mirror node instance numbered 017 is assigned to the server numbered 2 server. By analogy, the mirror node instances in each cabinet are evenly distributed among the servers in the cabinet, and the result is shown in Figure 10.

Among them, after the uniform distribution of all mirror node instances in each cabinet and each server is completed, a new master node instance and mirror node instance corresponding to each server can be generated, as shown in Table 2 below. Taking cabinet 1 as an example, corresponding to 3 The master node instance numbers of each server are: 001, 002, 003...012, and the mirror node instance numbers are: 013, 019, 025, 031, 015, 021, 027, 033, 017, 023, 029, 035 .

Table 2 List of mirror node instance numbers assigned to different servers in cabinet 1

According to the obtained number of cabinets N and the number of servers I installed in each cabinet, all mirror node instances can be configured for load balancing, and the generated corresponding configuration list can be provided to the GP database, and initialization can be performed according to the configuration list distribution, so as to achieve the effect of load balancing after a cabinet failure.

In the load balancing method of the mirror node of the database cluster provided by the embodiment of the present invention, when the current cabinet fails, the network pressure can also be dispersed in different cabinets, and will not be concentrated in the mirror cabinet alone. When the current cabinet fails , and can also make the entire cluster in a load-balanced state; automatically determine the distribution list of new mirror node instances according to the results of multiple Hash operations in different cabinets and the number of servers in the cabinet, without manually relying on experience to determine the distribution of mirror node instances. Based on the experience of the maintenance personnel, the reasonable distribution of mirror node instances is carried out according to the configuration parameters, and there will be no human error, etc., which ensures the reasonable distribution of mirror node instances and avoids problems.

On the other hand, as shown in FIG. 11 , the embodiment of the present invention also provides a load balancing device for a mirror node of a database cluster, which can adopt the load balancing method for a mirror node of a database cluster described in the above embodiment, including: Connected parameter acquisition module 111 and uniform distribution module 112;

Wherein, the parameter obtaining module 111 is used to obtain the configuration parameters of the database cluster; wherein, the configuration parameters include the number of cabinets, the number of servers in each cabinet, and the number of master node instances of each server;

The uniform distribution module 112 is used to evenly distribute the mirror node instances corresponding to all master node instances of the current cabinet to all other cabinets except the current cabinet according to the configuration parameters, so as to realize load balancing of the database cluster mirror nodes.

Specifically, the parameter acquisition module 111 acquires the configuration parameters of the database cluster; wherein, the configuration parameters include the number of cabinets, the number of servers in each cabinet, and the number of master node instances of each server; A cabinet performs the following operations: according to the configuration parameters, evenly distribute the mirror node instances corresponding to all master node instances of the current cabinet to all other cabinets except the current cabinet, so as to realize load balancing of the database cluster mirror nodes.

The load balancing device of the mirror node of the database cluster provided by the embodiment of the present invention evenly distributes the mirror node instances corresponding to all the master node instances of the current cabinet to all other cabinets except the current cabinet according to the configuration parameters of the database cluster. In the cabinet, the mirror node instance corresponding to the master node instance of the current cabinet is scattered in other different cabinets. Even if the current cabinet fails, the network pressure can be distributed in different cabinets, and will not be concentrated in a single In the mirror cabinet, when the current cabinet fails, the entire cluster can also be in a load-balanced state; in addition, the distribution of mirror node instances does not depend on the experience of maintenance personnel, and the reasonable distribution of mirror node instances according to configuration parameters will not In the event of human error, etc., the distribution of mirror node instances is ensured to be rationalized and problems are avoided.

On the basis of the foregoing embodiments, the even distribution module 112 includes:

The cabinet allocation unit is configured to perform the following operations for each cabinet: according to the number of master node instances of the current cabinet and the number of cabinets, evenly distribute the mirror node instances corresponding to the master node instance of the current cabinet to all other cabinets except the current cabinet in the cabinet;

The server allocation unit is configured to perform the following operations for each cabinet: according to the number of mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, evenly distribute the mirror node instances allocated to the current cabinet to different servers in the current cabinet.

On the basis of the above embodiments, the cabinet distribution unit includes:

The numbering subunit is used to number the mirror node instances corresponding to the master node instances of all cabinets;

The cabinet number allocation subunit is used to perform the following operations for each cabinet: according to the number of the mirror node instance corresponding to the master node instance of the current cabinet and the number of cabinets, assign the mirror node instances corresponding to all master node instances of the current cabinet The number is evenly distributed to other cabinets except the current cabinet, and finally the number data of the mirror node instance assigned to each cabinet is obtained;

On the basis of the above embodiments, the server allocation unit is specifically used for:

For each cabinet, perform the following operations: according to the number data of the mirror node instance assigned to the current cabinet and the number of servers in the current cabinet, evenly distribute the number of the mirror node instance assigned to the current cabinet to different servers in the current cabinet, and finally determine the current cabinet The number data of the mirror node instance assigned by each server.

For the load balancing device embodiment of the mirror node of the database cluster corresponding to the method, since it is basically similar to the method embodiment, the technical effect achieved is also the same as that of the method embodiment, so the description is relatively simple. For details, please refer to the part description of the method embodiment.

In another aspect, as shown in FIG. 12 , an embodiment of the present invention also provides an electronic device, which may be the device described in the above embodiment, and the electronic device includes at least one processor (processor) 121, a communication interface (Communications Interface) 122, at least one memory (memory) 123 and a bus 124, wherein, the processor 121, the communication interface 122, and the memory 123 complete mutual communication through the bus 124. The communication interface 122 can be used for information transmission between the electronic device and the server. The processor 121 may call the logic instructions in the memory 123 to execute the methods described in the above embodiments, for example, including: acquiring configuration parameters of the database cluster; wherein the configuration parameters include the number of cabinets, the number of servers in each cabinet, and the The number of primary node instances of a server; perform the following operations for each cabinet in the database cluster: according to the configuration parameters, evenly distribute the mirror node instances corresponding to all the primary node instances of the current cabinet to all other cabinets except the current cabinet In order to achieve load balancing of database cluster mirror nodes.

In addition, the above logic instructions in the memory 123 may be implemented in the form of software functional units and when sold or used as an independent product, may be stored in a computer-readable storage medium. Based on this understanding, the essence of the technical solution of the present invention or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including Several instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes. .

An embodiment of the present invention provides a computer program product, the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by a computer The computer can execute the methods provided by the above method embodiments, for example, including: obtaining configuration parameters of the database cluster; wherein, the configuration parameters include the number of cabinets, the number of servers in each cabinet, and the number of master node instances for each server; Perform the following operations for each cabinet in the database cluster: according to the configuration parameters, evenly distribute the mirror node instances corresponding to all the master node instances of the current cabinet to all other cabinets except the current cabinet, so as to realize the mirror node of the database cluster load balancing.

An embodiment of the present invention provides a non-transitory computer-readable storage medium, the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the methods provided in the above method embodiments, For example, it includes: obtaining the configuration parameters of the database cluster; wherein, the configuration parameters include the number of cabinets, the number of servers in each cabinet, and the number of master node instances of each server; perform the following operations for each cabinet in the database cluster: according to the According to the above configuration parameters, the mirror node instances corresponding to all the master node instances of the current cabinet are evenly distributed to all other cabinets except the current cabinet, so as to realize the load balancing of the database cluster mirror nodes.

The device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in One place, or it can be distributed to multiple network elements. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. It can be understood and implemented by those skilled in the art without any creative efforts.

Through the above description of the implementations, those skilled in the art can clearly understand that each implementation can be implemented by means of software plus a necessary general hardware platform, and of course also by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the prior art can be embodied in the form of software products, and the computer software products can be stored in computer-readable storage media, such as ROM/RAM, magnetic discs, optical discs, etc., including several instructions to make a computer device (which may be a personal computer, server, or network device, etc.) execute the methods described in various embodiments or some parts of the embodiments.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent replacements are made to some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims (10)

1. a load balancing method of database cluster mirror node, it is characterized in that, comprising: Obtain configuration parameters of the database cluster; wherein the configuration parameters include the number of cabinets, the number of servers in each cabinet, and the number of master node instances for each server; Perform the following operations for each cabinet in the database cluster: according to the configuration parameters, evenly distribute the mirror node instances corresponding to all the master node instances of the current cabinet to all other cabinets except the current cabinet, so as to realize the database cluster Load balancing of mirror nodes. 2. The method according to claim 1, wherein the following operation is performed for each cabinet in the database cluster: according to the configuration parameters, evenly distribute the mirror node instances corresponding to all the master node instances of the current cabinet into all other enclosures except the current one, including: Perform the following operations for each cabinet: according to the number of master node instances in the current cabinet and the number of cabinets, evenly distribute the mirror node instances corresponding to the master node instance of the current cabinet to all other cabinets except the current cabinet; Perform the following operations for each cabinet: according to the number of mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, evenly distribute the mirror node instances allocated to the current cabinet to different servers in the current cabinet. 3. The method according to claim 2, wherein the following operation is performed for each cabinet: according to the number of master node instances of the current cabinet and the number of cabinets, the mirror node corresponding to the master node instance of the current cabinet Instances are distributed evenly across all racks except the current one, including: Number the mirror node instances corresponding to the master node instances of all cabinets; For each cabinet, perform the following operations: according to the number of the mirror node instance corresponding to the master node instance of the current cabinet and the number of the cabinets, evenly distribute the numbers of the mirror node instances corresponding to all the master node instances of the current cabinet to the other cabinets except the current cabinet In other cabinets, the number data of the mirror node instance assigned to each cabinet is finally obtained. 4. The method according to claim 3, wherein, according to the serial number of the mirror node instance corresponding to the master node instance of the current cabinet and the number of cabinets, the mirror node instances corresponding to all the master node instances of the current cabinet The number of the cabinet is evenly distributed to other cabinets except the current cabinet, and the following hash function is used for uniform distribution: HashFunc(X)=M%(N-1) Among them, X is the hash address, which represents the cabinet number M is assigned to; M is the number of the mirror node instance corresponding to all the master node instances of the current cabinet, and N is the number of cabinets. 5. The method according to claim 3, wherein the following operations are performed for each cabinet: according to the number of mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, the mirror node instance allocated to the current cabinet Evenly distributed to different servers in the current cabinet, including: For each cabinet, perform the following operations: according to the number data of the mirror node instance assigned to the current cabinet and the number of servers in the current cabinet, evenly distribute the number of the mirror node instance assigned to the current cabinet to different servers in the current cabinet, and finally determine the current cabinet The number data of the mirror node instance assigned by each server. 6. The method according to claim 5, wherein, according to the serial number data of the mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, the numbers of the mirror node instances allocated to the current cabinet are evenly distributed to the current cabinet In different servers, the following hash functions are used for uniform distribution: HashFunc(Y)=K%I Among them, Y is the hash address, which represents the server number K is allocated to; K is the number of the mirror node instance allocated to the current cabinet, and Y is the number of servers in the current cabinet. 7. A load balancing device of a database cluster mirror node, characterized in that, comprising: A parameter acquisition module, configured to acquire configuration parameters of the database cluster; wherein the configuration parameters include the number of cabinets, the number of servers in each cabinet, and the number of master node instances of each server; The uniform allocation module is configured to perform the following operations for each cabinet in the database cluster: according to the configuration parameters, evenly distribute the mirror node instances corresponding to all the master node instances of the current cabinet to all other cabinets except the current cabinet In order to achieve load balancing of database cluster mirror nodes. 8. The device according to claim 7, wherein the uniform distribution module comprises: The cabinet allocation unit is configured to perform the following operations for each cabinet: according to the number of master node instances of the current cabinet and the number of cabinets, evenly distribute the mirror node instances corresponding to the master node instance of the current cabinet to all other cabinets except the current cabinet in the cabinet; The server allocation unit is configured to perform the following operations for each cabinet: according to the number of mirror node instances allocated to the current cabinet and the number of servers in the current cabinet, evenly distribute the mirror node instances allocated to the current cabinet to different servers in the current cabinet. 9. The device according to claim 8, wherein the cabinet distribution unit comprises: The numbering subunit is used to number the mirror node instances corresponding to the master node instances of all cabinets; The cabinet number allocation subunit is used to perform the following operations for each cabinet: according to the number of the mirror node instance corresponding to the master node instance of the current cabinet and the number of cabinets, assign the mirror node instances corresponding to all master node instances of the current cabinet The number is evenly distributed to other cabinets except the current cabinet, and finally the number data of the mirror node instance assigned to each cabinet is obtained. 10. The device according to claim 9, wherein the server allocation unit is specifically configured to: For each cabinet, perform the following operations: according to the number data of the mirror node instance assigned to the current cabinet and the number of servers in the current cabinet, evenly distribute the number of the mirror node instance assigned to the current cabinet to different servers in the current cabinet, and finally determine the current cabinet The number data of the mirror node instance assigned by each server.