Disaster recovery data synchronization system and method for dual-homing in different places

Abstract

The application provides a kind of including monitoring management platform, MPP data warehouse task scheduling platform, DNS relay server, MPP client server, main cluster, backup cluster and external service interface;MPP data warehouse task scheduling platform sends task scheduling instruction to DNS relay server;DNS relay server receives instruction and forwards script call instruction to MPP client server;MPP client server responds to instruction and sends data processing instruction to main cluster and backup cluster for data processing;Monitoring management platform monitors the scheduling state of MPP data warehouse task scheduling platform in real time, re-regulates when difference task appears, simultaneously monitors MPP client server service state, main and backup cluster state and switches external service interface.The application guarantees data integrity, consistency, availability, realizes double-live system service no-sense switching, solves the problem of data service unavailability in disaster scenario, and improves enterprise business continuity level.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular, to a remote dual-active disaster recovery data synchronization system and method. BACKGROUND

[0002] MPP (Massively Parallel Processing) is a large-scale parallel processing. In a non-shared database cluster, each node has an independent disk storage system and memory system. Business data is divided into various nodes according to the database model and application characteristics. Each data node is connected to each other through a dedicated network or a commercial general-purpose network, and cooperates with each other to provide database services as a whole. The non-shared database cluster has the advantages of complete scalability, high availability, high performance, excellent cost performance, and resource sharing.

[0003] In simple terms, MPP is to distribute tasks in parallel to multiple servers and nodes. After the calculation is completed on each node, the results of each part are combined to obtain the final result

[0004] Disaster backup refers to using scientific technical means and methods to establish a systematic data emergency mode in advance to deal with disasters. Its content includes data backup and system backup, business continuity planning, personnel architecture, communication guarantee, crisis public relations, disaster recovery planning, disaster recovery plan, business recovery plan, emergency response, third-party cooperation institutions and supply chain crisis management, etc.

[0005] The current mainstream single-machine room cluster or city dual-machine room cluster deployment corresponding to the data service disaster recovery scheme has certain limitations, mainly in that the standby service switching affects the business window time for a long time, the backup data has hysteresis, and the fault tolerance is low when dealing with force majeure such as natural disasters and wars.

[0006] Patent document CN109901948A discloses a non-shared database cluster remote dual-active disaster recovery system, which comprises a task scheduler located at the server end of the MPP cluster Server, used for executing different scheduling tasks for the main and backup clusters; a task executor located at the NodeAgent node agent end of the MPP cluster, which is an execution thread of the MPP cluster node agent layer, used for receiving and executing the scheduling tasks distributed by the task scheduler; a message bus module used for providing message interaction function for the non-shared database cluster remote dual-active disaster recovery system; and a DNS proxy service module used for providing service interface for the main and backup MPP clusters, and realizing automatic switching between the main and backup clusters through updating.

[0007] However, although the patent document CN109901948A establishes a dual-active data center, backs up the data of the business system data center to a remote place, and claims to use "incremental replication technology to reduce the use of network bandwidth", in fact, the scheme still pushes the daily change data from the full replication of the master cluster to the backup cluster, and must wait for the master cluster to process the corresponding data before starting synchronization to the backup cluster. There is a large time interval between the data versions of the two, and this scheme undoubtedly increases the network bandwidth consumption of the master cluster, and the synchronization data and query business simultaneously occupy the uplink bandwidth, which threatens the query efficiency and greatly reduces the efficiency of synchronization data. In this scheme, if you want to eliminate the impact of bandwidth occupation, you need to expand the uplink bandwidth by several times. For the dedicated line network of remote and remote transmission, the increased cost will be in the billions of dollars per year.

[0008] In summary, the scheme of patent CN109901948A is greatly limited by the uplink bandwidth of remote transmission, and cannot guarantee the optimal efficiency of external services, and the data version after the master-standby switching will inevitably have a large difference. SUMMARY

[0009] In view of the defects in the prior art, the purpose of the present application is to provide a remote dual-active disaster recovery data synchronization system and method.

[0010] The remote dual-active disaster recovery data synchronization system provided by the present application comprises a monitoring and management platform, an MPP data warehouse task scheduling platform, a DNS relay server, an MPP client server, a master cluster, a backup cluster and an external service interface.

[0011] The MPP data warehouse task scheduling platform sends a task scheduling instruction to the DNS relay server; the DNS relay server receives the instruction and forwards a script calling instruction to the MPP client server; the MPP client server responds to the instruction and sends a data processing instruction to the master cluster and the backup cluster for data processing.

[0012] The monitoring and management platform monitors the scheduling state of the MPP data warehouse task scheduling platform in real time, re-schedules when there are difference tasks, and monitors the service state of the MPP client server, the state of the master-standby cluster and the switching of the external service interface.

[0013] Preferably, the monitoring and management platform is used to refresh the scheduling task state in real time, compare the difference jobs, re-schedule the error jobs, and send the jobs with inconsistent statistical results to the monitoring and operation users for timely analysis and processing.

[0014] Compare the data statistics structure in the cluster to synchronize the differential data, and monitor the running state of the dual-active cluster. When the master cluster fails, the external service interface is switched to the backup cluster in real time.

[0015] Preferably, the MPP data warehouse task scheduling platform is used for configuring, calling and monitoring the developed data loading, processing and comparison jobs.

[0016] Preferably, the DNS relay server is used for forwarding the program running instructions of the MPP data warehouse task scheduling platform to the MPP client server and receiving the information returned by the MPP client server and sending the information to the MPP data warehouse task scheduling platform.

[0017] Preferably, the master cluster and the backup cluster each comprise an MPP cluster scheduler, an MPP data server, an MPP executor and an MPP storage service;

[0018] The MPP cluster scheduler is connected to and parses various sql statements and stored procedures submitted to the cluster from outside the data cluster, optimizes and pushes down the parsed statements, generates an execution plan, and distributes tasks in the cluster to each data server for execution according to the execution plan, while accepting the task return information of the data server, and updating the metadata inside the data cluster;

[0019] The MPP data server is used for building a cluster and providing hardware devices for the cluster;

[0020] The MPP executor is a distributed computing program carried on the data server, which executes the tasks pushed by the MPP cluster scheduler and performs read and write operations on the MPP storage service;

[0021] The MPP storage service is a distributed storage service carried on the data server, which directly provides data reading for the MPP executor according to the instructions of the MPP executor.

[0022] Preferably, the external service interface is integrated by programs, and the external services of the two clusters are encapsulated in the same interface, and the switching of the external service clusters is realized by switching different configuration files.

[0023] The switching action is configured in the monitoring management platform, and is automatically triggered when the original external service cluster fails, and the master and backup states of the cluster are updated.

[0024] According to the off-site dual-active disaster recovery data synchronization method provided by the application, the following steps are included:

[0025] Step S1: The master and backup cluster data are processed by the MPP data warehouse task scheduling platform, and the data processing includes synchronous calling of loading and processing tasks;

[0026] Step S2: the data processing results of the master and standby clusters are compared respectively, whether the data is consistent is judged, if not, an alarm log is sent through the scheduling platform to remind the administrator to manually check; if yes, step S3 is executed;

[0027] Step S3: the monitoring management platform monitors the cluster state in real time, whether the master cluster is abnormal is judged, if yes, according to the configured cluster information, the data service interface is switched to the standby cluster through the external interface service, and the master and standby states are updated and monitored; if not, the cluster state is continuously monitored.

[0028] Preferably, the clusters are established in different geographical plates, and each cluster adopts the same model and version of software and hardware configuration;

[0029] The clusters and the client include fiber special line communication.

[0030] Preferably, the synchronization scheduling in step S1 is independently executed by the master and standby clusters.

[0031] The master and standby clusters each use a set of scheduling configuration and code, and the master and standby cluster codes are checked for consistency when being issued.

[0032] The synchronization scheduling mode is determined according to the timeliness of the job, and only the error task and the task with inconsistent comparison results are re-adjusted and synchronized with data operation, and the timeliness is determined according to the business nature before the job deployment.

[0033] Preferably, the data comparison is executed by the MPP warehouse task scheduling platform to perform the data comparison task, and the master and standby data amounts are checked in real time, and the data content consistency is periodically compared.

[0034] The algorithm of the data comparison includes numerical type full table sum summation and sampling full field MD5 comparison.

[0035] Compared with the prior art, the present application has the following beneficial effects:

[0036] 1. The present application realizes the double-live system service non-sensing switching, solves the problem of unavailable data service in the disaster scene, and improves the enterprise business continuity level.

[0037] 2. The present application realizes the parallel and data synchronization of the massive data processing system through the comparison and synchronization of the master and standby cluster data, and only the error task and the task with inconsistent comparison results are re-adjusted and synchronized with data operation, which maximally reduces the data version difference between the clusters and the time and bandwidth occupation required by the synchronization version, so that the negative influence of the business switching is infinitely close to zero.

[0038] 3、The application realizes the same task scheduling and the same data processing operation through two sets of clusters, and realizes real-time statistical monitoring of the processing results of both sides and periodic comparison at the minute level (each according to the same algorithm to count data, and only the calculation results are exchanged to check and compare), so that the true sense of off-site dual-active analysis is ensured. BRIEF DESCRIPTION OF DRAWINGS

[0039] Other features, objects, and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments, when read in conjunction with the accompanying drawings:

[0040] Figure 1 The figure is a schematic diagram of the workflow of the application. DETAILED DESCRIPTION

[0041] The application will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the application, but do not limit the application in any form. It should be noted that, for those skilled in the art, without departing from the concept of the application, a number of changes and improvements can be made. These all belong to the protection scope of the application.

[0042] Example 1

[0043] According to the off-site dual-active disaster recovery data synchronization system provided by the application, as shown in the figure, it comprises a monitoring management platform, an MPP data warehouse task scheduling platform, a DNS relay server, an MPP client server, a master cluster, a backup cluster and an external service interface. Figure 1

[0044] The MPP data warehouse task scheduling platform sends a task scheduling instruction to the DNS relay server; the DNS relay server receives the instruction and forwards a script calling instruction to the MPP client server; the MPP client server responds to the instruction and sends a data processing instruction to the master cluster and the backup cluster for data processing; the monitoring management platform monitors the scheduling state of the MPP data warehouse task scheduling platform in real time, re-schedules when there is a difference task, and monitors the service state of the MPP client server, the state of the master and backup clusters and the switching of the external service interface.

[0045] Specifically, the monitoring management platform is used to refresh the scheduling task state in real time, compare the difference jobs, re-schedule the error jobs, send the tasks with inconsistent statistical results to the monitoring and operation users for timely analysis and processing, such as comparing the consistency of the data change quantity involved. Compare the data statistical structure in the cluster to synchronize the differential data, and monitor the running state of the dual-active cluster. When the master cluster fails, the external service interface is switched to the backup cluster in real time.

[0046] ​The MPP data warehouse task scheduling platform is used for configuring, calling and monitoring the developed data loading, processing and comparison jobs.

[0047] The DNS relay server is used for forwarding the program running instructions of the MPP data warehouse task scheduling platform to the MPP client server, and receiving the information returned by the MPP client server and sending the information to the MPP data warehouse task scheduling platform.

[0048] The MPP client server deploys and runs the MPP client, data loading program and data processing script, and simultaneously serves as a data file transfer station to interface various data file exchange services of external systems.

[0049] The master cluster and the backup cluster each include an MPP cluster scheduler, an MPP data server, an MPP executor and an MPP storage service. The MPP cluster scheduler interfaces and parses various sql statements and stored procedures submitted to the cluster from outside the data cluster, optimizes and pushes down the parsed statements, generates an execution plan, and distributes tasks in the cluster to each data server for execution according to the execution plan, while accepting the task return information of the data servers, and updating the metadata in the data cluster. The metadata includes data in the data cluster and cluster service management information. The MPP data server is used for building the cluster and providing hardware devices for the cluster. The MPP executor is a distributed computing program carried on the data server, which executes the tasks pushed by the MPP cluster scheduler and performs read and write operations on the MPP storage service. The MPP storage service is a distributed storage service carried on the data server, which directly provides data reading for the MPP executor according to the instructions of the MPP executor.

[0050] The external service interface is integrated by programs, and the external services of the two clusters are encapsulated in the same interface, and the switching of the external service clusters is realized by switching different configuration files. The switching action is configured in the monitoring management platform, and is automatically triggered when the original external service cluster fails, and the cluster master and backup states are updated.

[0051] Embodiment two

[0052] According to the off-site dual-active disaster recovery data synchronization method provided by the application, the method comprises the steps that:

[0053] Step S1: data processing is performed on the master and backup cluster data through the MPP data warehouse task scheduling platform, and the data processing includes synchronous calling of loading and processing tasks. The method realizes synchronous data processing tasks of the master and backup systems, and maximally maintains the consistency of the dual-cluster data.

[0054] Specifically, the synchronization scheduling is independently executed by the master and standby clusters. The master and standby clusters each use a set of scheduling configurations and codes, and the master and standby cluster codes are checked for consistency when being issued. The synchronization scheduling mode is determined according to the timeliness of the job, and only the jobs with errors and the jobs with inconsistent comparison results are re-scheduled and synchronized with data operations. The timeliness is determined according to the business nature before the job deployment.

[0055] Further, the MPP data warehouse task scheduling platform includes a star flow scheduling platform. For jobs with general timeliness, the master cluster provides data, and the standby cluster provides emergency replacement. Under normal circumstances, the standby cluster does not execute the export transmission job, and only executes the export transmission after an abnormal switch. For the export transmission of the master cluster, it is no longer processed; the standby cluster adds a master-standby data comparison job flow, and the comparison result is recorded in a table. The comparison exception does not perform job layer error reporting, and can be notified and viewed through the management platform. The problem is manually investigated and the standby cluster data is repaired. For jobs with high timeliness, the master and standby clusters are checked for code consistency before code execution. When the codes are inconsistent, the master cluster code is synchronized to the standby cluster before execution.

[0056] It should be noted that the job with high timeliness refers to a query scenario that requires real-time or quasi-real-time update of data. Such a job has a higher requirement for data timeliness, such as FST job. Conversely, if the query business can accept T+0.5 or T+1 or even longer period version delay data, the requirement for data timeliness is general.

[0057] Step S2: The data processing results of the master and standby clusters are compared respectively to determine whether the data is consistent. If not, an alarm log is sent through the scheduling platform to remind the administrator to manually check and determine whether manual intervention is needed. If yes, step S3 is executed. Specifically, the data comparison is performed by the MPP data warehouse task scheduling platform, such as calling a self-developed comparison program to compare the master and standby data in real time, such as increment comparison according to time conditions, and periodic comparison of data content consistency. The algorithm for data comparison includes summing all tables of numerical type and sampling all fields MD5 comparison (based on primary key).

[0058] Step S3: The monitoring management platform monitors the cluster state in real time to determine whether the master cluster is in an abnormal state. If yes, according to the configured cluster information, the data service interface is switched to the standby cluster through the external interface service, and the master and standby states are updated and monitored, including sending log alarms to remind the operation and maintenance personnel to troubleshoot abnormal problems; if not, the cluster state is continuously monitored. The log alarm is displayed through the scheduling platform.

[0059] The multi-center architecture design and the main and standby system deployment in the application specifically include that the clusters are established in different regional plates, each cluster is established by using the same model and version of software and hardware configuration. The communication between the clusters and between the clusters and the clients includes the fiber special line.

[0060] Those skilled in the art know that, in addition to implementing the system, device and each module thereof provided by the application in the form of pure computer readable program code, the same program can be implemented in the form of logic gate, switch, special integrated circuit, programmable logic controller and embedded microcontroller by logically programming the method steps. Therefore, the system, device and each module thereof provided by the application can be considered as a hardware component, and the modules included therein for implementing various programs can also be considered as the structure in the hardware component; the modules for implementing various functions can also be considered as both the software program for implementing the method and the structure in the hardware component.

[0061] The specific embodiments of the application are described above. It needs to be understood that the application is not limited to the specific implementation described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essential content of the application. The embodiments of the application and the features in the embodiments can be combined with each other in any way without conflict.

Claims

1. A remote active-active disaster recovery data synchronization system, characterized in that, The system comprises a monitoring management platform, an MPP data warehouse task scheduling platform, a DNS relay server, an MPP client server, a primary cluster, a backup cluster and an external service interface. The MPP data warehouse task scheduling platform sends a task scheduling instruction to the DNS relay server; the DNS relay server receives the instruction and forwards a script calling instruction to the MPP client server; The MPP client server responds to the instruction and sends a data processing instruction to the primary cluster and the backup cluster for data processing; The monitoring management platform monitors the scheduling state of the MPP data warehouse task scheduling platform in real time, re-schedules when there is a difference task, and monitors the service state of the MPP client server, the state of the primary and backup clusters and the switching of the external service interface; The MPP data warehouse task scheduling platform is used to configure, call and monitor developed data loading, processing and comparison jobs; The DNS relay server is used to forward a program running instruction of the MPP data warehouse task scheduling platform to the MPP client server and receive information returned by the MPP client server and send it to the MPP data warehouse task scheduling platform; The primary cluster and the backup cluster each comprise an MPP cluster scheduler, an MPP data server, an MPP executor and an MPP storage service; The MPP cluster scheduler interfaces with and parses various sql statements and stored procedures submitted to the cluster from outside the data cluster, optimizes and pushes down the parsed statements, generates an execution plan, and distributes tasks in the cluster to each data server for execution according to the execution plan, while accepting task return information from the data servers and updating the metadata inside the data cluster; The MPP data server is used to build a cluster and provide hardware devices for the cluster; The MPP executor is a distributed computing program loaded on the data server, which executes tasks pushed by the MPP cluster scheduler and performs read and write operations on the MPP storage service; The MPP storage service is a distributed storage service loaded on the data server, which performs addition, deletion and modification of data according to the instructions of the MPP executor and directly provides data reading for the MPP executor.

2. The disaster recovery data synchronization system of claim 1, wherein, The monitoring management platform is used to refresh the scheduling task state in real time, compare difference jobs, re-schedule error jobs, and send tasks with inconsistent statistical results to monitoring and operation users for timely analysis and processing; The data statistical structure in the cluster is compared, the differential data is synchronized, and the running state of the dual-active cluster is monitored. When the primary cluster fails, the external service interface is switched to the backup cluster in real time.

3. The disaster recovery data synchronization system of claim 1, wherein, The external service interface is integrated by programs, and the external services of the two clusters are encapsulated in the same interface. The switching of the external service cluster is realized by switching different configuration files. The switching action is configured in the monitoring management platform and automatically triggered when the original external service cluster fails, and the cluster primary and backup states are updated.

4. A method for synchronizing data of a disaster recovery system with dual active sites, characterized in that, The system comprises a monitoring management platform, an MPP data warehouse task scheduling platform, a DNS relay server, an MPP client server, a primary cluster, a backup cluster and an external service interface. Step S1: The MPP data warehouse task scheduling platform is used to process data of the primary and backup clusters, which comprises loading and processing tasks for synchronous calling; Step S2: The data processing results of the master and standby clusters are compared respectively to determine whether the data is consistent. If not, an alarm log is sent through the scheduling platform to remind the administrator to manually check. If yes, step S3 is executed. Step S3: The monitoring management platform monitors the cluster state in real time to determine whether the master cluster is in an abnormal state. If yes, the data service interface is switched to the standby cluster through the external interface service according to the configured cluster information, and the master and standby states are updated and monitored. If not, the cluster state is continuously monitored.

5. The method of claim 4, wherein, The clusters are established in different geographical plates, and each cluster uses the same model and version of software and hardware configuration. Communication between clusters and between clusters and clients includes fiber special line.

6. The method of claim 5, wherein, The synchronous call in step S1 is independently executed by the master and standby clusters. The master and standby clusters each use a set of scheduling configurations and codes, and the codes are checked for consistency when they are issued. The synchronous call mode is determined according to the timeliness of the job. Only error tasks and tasks with inconsistent comparison results are re-adjusted and synchronized for data operation. The timeliness is determined according to the business nature before the job is deployed.

7. The method of claim 5, wherein, The data comparison is performed by the MPP data warehouse task scheduling platform to compare the master and standby data in real time and periodically compare the data content consistency. The data comparison algorithm includes sum summation of all tables of numerical type and sampling full field MD5 comparison.

Images