CN113312147B - Method and system for migrating object storage across cluster mass data - Google Patents

Abstract

The present invention discloses a method and system for object storage cross-cluster mass data migration. The method includes: step S1, receiving a migration task request from a user and establishing a subtask; step S2, generating a migration task based on information related to establishing a subtask Correspond to the configuration file, and store the relevant information of the established subtask and the corresponding configuration file of the migration task in the task queue of the back-end database as an OSS object; step S3, scan the task queue stored in the back-end database according to the predetermined duration, and save the The migration task in the waiting state and the paused state is scheduled and executed; step S4, the invoked migration task is run in any mode of Docker container, K8S job, or process; step S5, start according to the type of the migration task The migration plug-in of the corresponding type performs the migration operation; in step S6, the called corresponding migration plug-in completes the data migration in a multi-level task manner. This method can realize elastic migration of massive data across clusters.

Description

Method and system for mass data migration across object storage clusters

technical field

The invention relates to the field of cloud storage data migration, in particular to an object storage cross-cluster massive data migration system and system.

Background technique

Data migration is an essential product and tool for major cloud service providers, such as Microsoft Azure, Alibaba Cloud, Huawei Cloud, and Tencent Cloud, all of which have their own data migration tools; various storage clusters and frameworks also have their own data migration and balancing solutions .

At present, Alibaba Cloud's DTS (Data Transmission service) solution focuses on data migration between OSS clusters. It uses a cloud-based central service to provide data migration from various OSS storage to Alibaba Cloud OSS, and does not support heterogeneous clusters. Data migration; also does not support data migration between non-Alibaba cloud storage.

HUAWEI CLOUD migrates data in a transactional manner, and rolls back all operations when the migration fails.

Most of the existing cloud storage data migration technologies use a centralized service as the migration center, and most of them are migrations between specific types of data clusters. Most of the migration systems are temporarily built to respond to the needs, and then the system is discarded; at the same time, the existing technology rarely realizes data migration between heterogeneous clusters, or only realizes data migration between heterogeneous clusters with special needs.

Contents of the invention

Based on the problems existing in the prior art, the object of the present invention is to provide an object storage cross-cluster mass data migration system and system, which can solve the existing cloud storage data migration system, most of which do not support data transfer between heterogeneous clusters Migration or data migration between heterogeneous clusters that only support special requirements.

The purpose of the present invention is achieved through the following technical solutions:

The embodiment of the present invention provides a method for object storage cross-cluster mass data migration, including:

Step S1, receiving the migration task request issued by the user and the request for establishing subtasks issued by the migration task request;

Step S2, generating a configuration file corresponding to the created migration task according to the relevant information of the established subtask, and storing the related information of the established subtask and the corresponding configuration file of the created migration task in the backend database as an OSS object;

Step S3, regularly scan the task information stored in the back-end database, and schedule the migration tasks in the waiting state and the paused state to execute the migration tasks;

Step S4, the invoked migration task runs in any mode of Docker container, K8S job, or process;

Step S5, start the task instance with the corresponding type of migration plug-in according to the type of the task to perform the migration operation;

In step S6, the migration plug-in completes data migration in a multi-level task manner.

The embodiment of the present invention provides an object storage cross-cluster mass data migration system for implementing the method described in the present invention, including:

The entry service unit, the migration task scheduling unit, the migration task execution unit and the backend database; among them,

The entry service unit communicates with the back-end database, can receive the migration task request sent by the user and the subtask established by the migration task request, and generate the corresponding configuration file of the migration task according to the information related to the establishment of the subtask , and store the established subtasks and corresponding configuration file related information of the migration task in the task queue of the backend database as OSS objects;

The migration task scheduling unit is connected in communication with the back-end database, and can scan the task queue stored in the back-end database according to a predetermined period of time, and schedule and execute the migration tasks in the waiting state and the paused state. The above-mentioned transferred migration task runs in any mode of Docker container, K8S job, or process;

The migration task execution unit is connected to the migration task scheduling unit and the back-end database respectively, and can start a corresponding type of migration plug-in according to the type of the migration task to perform the migration operation, and the migration plug-in adopts a multi-level task way to complete the data migration.

It can be seen from the above-mentioned technical solution provided by the present invention that the object storage cross-cluster massive data migration system and system provided by the embodiment of the present invention have the following beneficial effects:

By starting the task instance with the corresponding type of migration plug-in according to the connection of the migration task, it can easily realize the migration of massive data between heterogeneous clusters. Because it is a data migration heterogeneous plug-in based on cloud native technology, it supports migration task service components One-click automatic deployment and destruction, support plug-in access for data migration of different storage systems.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings that need to be used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For Those of ordinary skill in the art can also obtain other drawings based on these drawings on the premise of not paying creative work.

FIG. 1 is a flowchart of a method for object storage cross-cluster massive data migration provided by an embodiment of the present invention;

FIG. 2 is a block diagram of an object storage cross-cluster massive data migration system provided by an embodiment of the present invention;

FIG. 3 is a block diagram of an entry service unit of an object storage cross-cluster mass data migration system provided by an embodiment of the present invention;

FIG. 4 is a block diagram of a migration task scheduling unit of an object storage cross-cluster mass data migration system provided by an embodiment of the present invention;

FIG. 5 is a block diagram of a migration task execution unit of an object storage cross-cluster mass data migration system provided by an embodiment of the present invention;

FIG. 6 is a flow chart of the processing of the entry service unit of the object storage cross-cluster mass data migration system provided by the embodiment of the present invention;

FIG. 7 is a processing flowchart of the task scheduling unit of the object storage cross-cluster mass data migration system provided by the embodiment of the present invention;

FIG. 8 is a processing flowchart of the task execution unit of the object storage cross-cluster mass data migration system provided by the embodiment of the present invention;

FIG. 9 is a flow chart of ingress service processing of the object storage cross-cluster massive data migration method provided by the embodiment of the present invention;

FIG. 10 is a schematic diagram of the specific composition of the object storage cross-cluster massive data migration system provided by the embodiment of the present invention;

FIG. 11 is a flow chart of task generation and scheduling in the object storage cross-cluster massive data migration method provided by the embodiment of the present invention;

12 is a schematic diagram of the logic structure of the account-level data migration plug-in of the object storage cross-cluster mass data migration system provided by the embodiment of the present invention;

FIG. 13 is a schematic diagram of the logical structure of the container/bucket level data migration plug-in of the object storage cross-cluster mass data migration system provided by the embodiment of the present invention;

FIG. 14 is a schematic diagram of the logical structure of the FileList-level data migration plug-in of the object storage cross-cluster mass data migration system provided by the embodiment of the present invention.

detailed description

The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the specific content of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention. The content not described in detail in the embodiments of the present invention belongs to the prior art known to those skilled in the art.

Referring to Figure 1, an embodiment of the present invention provides a method for object storage cross-cluster massive data migration, including:

Step S1, receiving the migration task request issued by the user and the request for establishing subtasks issued by the migration task request;

Step S2, generating a configuration file corresponding to the created migration task according to the relevant information of the established subtask, and storing the related information of the established subtask and the corresponding configuration file of the created migration task in the backend database as an OSS object;

Step S3, regularly scan the task information stored in the back-end database, and schedule the migration tasks in the waiting state and the paused state to execute the migration tasks;

Step S4, the invoked migration task runs in any mode of Docker container, K8S job, or process;

Step S5, start the task instance with the corresponding type of migration plug-in according to the type of the task to perform the migration operation;

In step S6, the migration plug-in completes data migration in a multi-level task manner.

In step S6 of the above method, the migration plug-in uses multi-level tasks to complete the data migration in the following manner:

Step S61) Scheduling the corresponding migration task image according to the received user migration task request, if the user needs to migrate the entire account (account), then start the account migration task, the account migration task traverses all buckets (ie container or bucket) and creates bucket migration task;

Step S62) The bucket migration task created is scheduled to traverse all the files under the bucket, and create a file migration task for every n files according to the predetermined number n;

After the file migration task created in step S63) is scheduled, the source file of the source cluster and the target file of the target cluster will be compared. If the source file and the target file are the same, copying and migration of the file will be skipped. If the source file is the same as the target file If the file is different, copy and migrate the file.

The above method also includes:

Step S7, failure recovery, all tasks that are called up will regularly insert checkpoints (ie checkpoints) into the back-end database, and re-run from the last completed checkpoint when rescheduling after a failure;

The checkpoint is an internal event, which will trigger the database writing process to write out the dirty data blocks in the data buffer to the data file after the event is activated.

Referring to Fig. 2, the embodiment of the present invention also provides an object storage cross-cluster mass data migration system for implementing the above method, including:

Entrance service unit 1, migration task scheduling unit 2, migration task execution unit 3 and backend database 4; wherein,

The entry service unit communicates with the back-end database, can receive the migration task request sent by the user and the subtask established by the migration task request, and generate the corresponding configuration file of the migration task according to the information related to the establishment of the subtask , and store the established subtasks and corresponding configuration file related information of the migration task in the task queue of the backend database as OSS objects;

The migration task scheduling unit is connected in communication with the back-end database, and can scan the task queue stored in the back-end database according to a predetermined period of time, and schedule and execute the migration tasks in the waiting state and the paused state. The above-mentioned transferred migration task runs in any mode of Docker container, K8S job, or process;

The migration task execution unit is connected to the migration task scheduling unit and the back-end database respectively, and can start a corresponding type of migration plug-in according to the type of the migration task to perform the migration operation, and the migration plug-in adopts a multi-level task way to complete the data migration.

Figure 2 is a logical architecture diagram of the entire system: it is divided into three main logical units: the entry service unit 1, the migration task scheduling unit 2, and the migration task execution unit 3; among them,

The entry service unit is used to receive and execute instructions from users or during task execution;

The migration task scheduling unit schedules migration tasks according to a certain scheduling algorithm to achieve concurrent execution of tasks;

The migration task execution unit is the executor of the real migration task, which visits the source and target clusters to collect information and execute data migration copy.

Referring to Fig. 3, in the above system, the entrance service unit 1 includes:

Task creation request module 11, task request sub-module 12, generating configuration information module 13, inserting queue module 14, database operation module 15, task scheduling module 16, adjusting container cloud resource module 17 and starting container module 18; wherein,

The task creation request module is configured to receive a migration task request from a user, and create a migration task according to the migration task request;

The task request submodule is communicatively connected with the task creation request module, and is used to establish a corresponding subtask according to the migration task;

The generating configuration information module is respectively connected to the task creation request module and the task request sub-module, and is used to generate the corresponding configuration file of the migration task according to the relevant information of establishing sub-tasks;

The insertion queue module is respectively connected to the task creation request module, the task request sub-module and the configuration information generation module, and inserts the established sub-task and the corresponding configuration file related information of the migration task into the task queue;

The database operation module is communicated with the insertion queue module, and stores each task of the task queue in the back-end database as an OSS object;

The task scheduling module is connected to the migration task scheduling unit, the database operation module, and the container cloud resource adjustment module respectively, and can send a scheduling task request to the migration task scheduling unit and receive the migration task scheduling unit. The migration tasks scheduled by the task scheduling unit are sent to the database operation module and the adjustment container cloud resource module respectively;

The module for adjusting container cloud resources is connected in communication with the task scheduling module, and is used to adjust container cloud resources according to the migration task sent by the task scheduling module;

The starting container module is connected in communication with the adjusting container cloud resource module, and is used to start and run the invoked migration task in any mode of Docker container, K8S job, or process.

Fig. 6 schematically shows the processing flow of the above-mentioned entry service unit.

Referring to FIG. 4, in the above system, the migration task scheduling unit 2 includes:

task scanning judging module 21 and migration task determining module 22; wherein,

The task scanning judgment module is communicated with the migration task determination module, and can scan the task queue stored in the back-end database according to a predetermined duration, and send the migration tasks in the waiting state and paused state to the migration task task determination module;

The migration task determining module schedules and executes the migration task sent by the task scanning and judging module.

FIG. 7 schematically illustrates the processing flow of the above-mentioned migration task scheduling unit.

Referring to FIG. 5, in the above system, the migration task execution unit 3 includes:

Migration plug-in selection module 31 and multiple migration plug-in modules 32, 32, 33... 3n; wherein,

The plug-in selection module is communicated with each migration plug-in module respectively, and can start a corresponding type of migration plug-in according to the type of the migration task to perform the migration operation;

Each migration plug-in module can complete data migration in a multi-level task manner after being activated by the migration plug-in selection module.

In the above system, the migration task execution unit further includes: a failure recovery module, which is respectively connected to the back-end database and each migration plug-in in communication, and is used for the plug-in module to perform all scheduled file migration tasks according to preset Insert a checkpoint (i.e. checkpoint) to the back-end database at a fixed time, and when the file migration task is rescheduled after a failure, it will start to run again from the last completed checkpoint; the checkpoint is an internal event, and this event After activation, the backend database write process will be triggered to write out the dirty data blocks in the data buffer to the data file.

FIG. 8 schematically illustrates the processing flow of the above-mentioned migration task execution unit.

The data migration system of the present invention is a completely flexible plug-in cross-cluster data migration system. By setting different types of migration plug-ins that can be selected according to the type of migration task, it can be flexibly scheduled to realize that the migration system does not work when there is no work. Occupies resources, schedules necessary resources when work is required, and releases used resources after work is completed; different migration requirements are realized through migration plug-ins, and data migration between various heterogeneous clusters is supported; when tasks fail, they can be transferred from the process continue at the point where it was interrupted.

The embodiments of the present invention will be further described in detail below.

Referring to Figures 1, 9, 10 and 11, the object storage cross-cluster massive data migration method of the present invention includes the following steps:

Step S1, constructing a service on InterFaceServer for receiving the migration task request sent by the user. In addition to receiving the creation task request from the user, the service also receives the request from the task to create a subtask;

Step S2, InterFaceServer writes the relevant information of the created task into the back-end database, generates the configuration file of the task to be created and uploads it to the unified storage of the back-end, and OSS object storage is used in this embodiment;

Step S3, the task scheduler regularly scans the task queue in the back-end database, and schedules the tasks in the waiting and pause states to execute the migration task;

In step S4, the scheduled migration task can run in a Docker container, can also be started and run as a K8S job, or can generate a process running task on a physical machine;

Step S5, the task scheduler will schedule different plug-ins to start the task instance according to the type of task; the example in this embodiment is to download the image of the corresponding Docker container according to the task type, and start the job instance to perform the migration operation;

Step S6, specifically, implement the migration plug-in in a multi-level task manner:

Step S61) After receiving the user task, schedule the corresponding migration task image according to the requirements of the user task, if the user needs to migrate the entire account, start the account migration task, the account migration task traverses all containers, and creates a container migration task;

Step S62) After the created container migration task is scheduled, it will traverse the files under the container, and create a file migration task for every n files according to the specified number of files (assumed to be n);

Step S63) After the created file migration task is scheduled, it will compare the source file of the source cluster with the target file of the target cluster. If the source file is the same as the target file, skip the copy migration of the source file. If the source file If it is different from the target file, copy and migrate the file;

Step S7, fault recovery, all scheduled tasks will regularly insert checkpoints into the database, and when rescheduled after a fault, rerun from the last completed checkpoint (see Figure 9).

Each step of the above method is specifically implemented as follows: Figure 9 illustrates the execution process of InterFaceServer: user creation of tasks and creation of sub-tasks, execution process and scheduling of operation and maintenance instructions, and Figure 10 illustrates the object storage cross-cluster mass data migration system specific composition;

Implement and deploy InterFaceServer; InterFaceServer can use RestAPI to receive user requests, and can be based on RPC or CMD (command line) mode;

(1) After InterfaceServer receives the request to create a character, it writes the task information to be created into the database or writes the task to be created into a storage or memory block that can be read by the public, and sets the state to wait;

(2) The scheduler schedules the corresponding type of task process in the physical machine according to the task information in the database or public read storage or memory block, and specifies the relevant information required by the task together, and sets the status as running;

(3) The scheduler can also schedule the corresponding tasks or pods in K8s, and set the status to running;

(4) The scheduler can also set the task corresponding to the scheduler in the Docker container, and set the status to running;

(5) After the task ends, close its own process/job/pod, and set the status to end;

(6) Implementation of the migration plug-in:

See Figure 12, (61) account-level data migration plug-in:

(611) After the task is started, checkpoint is written into the database or shared storage;

(612) List all container/bucket information under the account from the source cluster list according to the task configuration information;

(613) Apply to InterfaceServer to create a container level migration task according to the container information in the list;

(614) Create a checkpoint information and write it into shared storage or database every time a certain number of containers are scanned;

See Figure 13, (62) Container level data migration plug-in

(621) Write the checkpoint to the database or shared storage after the task starts;

(622) List all file information under the container from the source cluster list according to the task configuration information;

(623) Every 1000 (or other numbers) of the file list that list goes out is organized into database table or serialized file form information and applies to InterfaceServer to create FileList level migration task;

(624) Create a checkpoint information and write it into shared storage or database every time a certain number of FileLists are scanned;

See Figure 14, (63) FileList level data migration plug-in:

(631) Write the checkpoint to the database or shared storage after the task starts;

(632) Obtain the database table or serialization file corresponding to the task according to the task configuration information;

(633) copy the file from the source cluster to the target cluster according to the obtained file list;

(634) End itself when done.

The migration system and method of the present invention have at least the following advantages:

(1) The present invention adopts the popular containerization scheme, which can easily implement tasks based on K8S orchestration, thereby realizing fully flexible scheduling, and the design is closer to the characteristics of migration tasks: tasks are relatively concentrated, and no system resources are needed when there are no tasks ;

(2) The present invention adopts a plug-in expansion scheme, which can easily expand new requirements and migrate between different types of clusters, instead of the traditional data migration that can only be performed in the same cluster;

(3) The present invention realizes automatic fault recovery based on the checkpoint method, which has the advantages of fast recovery speed, easy tracking of fault recovery points, less record information of recovery points, and simple and convenient implementation method.

Those of ordinary skill in the art can understand that all or part of the process in the method of the embodiment of the present invention can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, the program During execution, it may include the processes of the embodiments of the above-mentioned methods. Wherein, the storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM) or a random access memory (Random Access Memory, RAM) and the like.

The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person familiar with the technical field can easily conceive of changes or changes within the technical scope disclosed in the present invention. Replacement should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the protection scope of the claims.

Claims (8)

1. A method for migrating object storage across cluster mass data is characterized by comprising the following steps:

step S1, receiving a migration task request sent by a user and a subtask established by the migration task request;

step S2, generating a corresponding configuration file of the migration task according to the relevant information of the established subtasks, and storing the established subtasks and the relevant information of the corresponding configuration file of the migration task into a task queue of a back-end database by an OSS object;

s3, scanning the task queue stored in the back-end database according to a preset time length, and scheduling and executing the migration tasks in the waiting state and the pause state;

s4, running the called migration task in any one mode of a Docker container, a jobs of K8S and a process;

s5, starting a migration plug-in of a corresponding type according to the type of the migration task to perform migration operation;

s6, the called migration plug-in completes data migration in a multi-level task mode;

in step S6, the migration plug-in completes data migration in a multi-level task manner as follows:

step S61) scheduling a corresponding migration task mirror image according to the received request of the user migration task, if the user needs to migrate the whole account, starting the account migration task, and creating a bucket migration task after the account migration task traverses all the buckets;

step S62) created the barrel migration task is scheduled and then traverses all files under the barrel, and a file migration task is created according to every n files, wherein n is a preset number;

and S63) after the created file migration task is scheduled, comparing the source file of the source end cluster with the target file of the target end cluster, skipping the copying and migration of the source file if the source file and the target file are the same, and copying and migrating the source file to the target end cluster as the target file if the source file is different from the target file.

2. The method for migrating object storage across cluster mass data according to claim 1, further comprising:

s7, recovering the fault, inserting check points into the back-end database according to preset time length for all scheduled file migration tasks, and when the file migration tasks are rescheduled after the fault, restarting from the last completed check point;

the check point is an internal event, and the event triggers a back-end database writing process to write out the dirty data blocks in the data buffer into the data file after being activated.

3. The method for migrating the object storage across the cluster mass data according to claim 1, wherein in the step S5, the migration plug-in of the corresponding type includes: any of an account level data migration plug-in, a bucket level data migration plug-in, and a file list level data migration plug-in.

4. An object storage cross-cluster mass data migration system for implementing the method of any one of claims 1 to 3, comprising:

the system comprises an entrance service unit, a migration task scheduling unit, a migration task execution unit and a back-end database; wherein,

the entrance service unit is in communication connection with the back-end database, can receive a migration task request sent by a user and a subtask established by the migration task request, generates a corresponding configuration file of the migration task according to related information for establishing the subtask, and stores the established subtask and the related information of the corresponding configuration file of the migration task into a task queue of the back-end database by an OSS object;

the migration task scheduling unit is in communication connection with the back-end database, can scan a task queue stored in the back-end database according to preset time length, schedules and executes the migration tasks in a waiting state and a pause state, and the scheduled migration tasks run in any mode of a Docker container, a job of K8S and a process;

the migration task execution unit is respectively in communication connection with the migration task scheduling unit and the back-end database, and can start a migration plug-in of a corresponding type according to the type of the migration task to perform migration operation, and the migration plug-in completes data migration in a multi-level task mode; the migration plug-ins of the corresponding type include: any of an account level data migration plug-in, a bucket level data migration plug-in, and a file list level data migration plug-in.

5. The system for migrating object storage across cluster mass data according to claim 4, wherein said portal service unit comprises:

the system comprises a task creating request module, a task request submodule, a configuration information generating module, an insertion queue module, a database operation module, a task scheduling module, a container cloud resource adjusting module and a container starting module; wherein,

the task creation request module is used for receiving a migration task request sent by a user and establishing a migration task according to the migration task request;

the task request submodule is in communication connection with the task creation request module and is used for creating corresponding subtasks according to the migration tasks;

the configuration information generating module is respectively in communication connection with the task creating request module and the task request submodule and is used for generating a corresponding configuration file of the migration task according to the relevant information for establishing the subtasks;

the inserting queue module is respectively in communication connection with the task creating request module, the task request submodule and the configuration information generating module, and inserts the related information of the corresponding configuration files of the established subtasks and the migration tasks into a task queue;

the database operation module is in communication connection with the insertion queue module and stores each task of the task queue into a back-end database by using an OSS object;

the task scheduling module is respectively in communication connection with the migration task scheduling unit, the database operation module and the adjustment container cloud resource module, can send a scheduling task request to the migration task scheduling unit, receives the migration task scheduled by the migration task scheduling unit, and respectively sends the migration task request to the database operation module and the adjustment container cloud resource module;

the adjustment container cloud resource module is in communication connection with the task scheduling module and is used for adjusting container cloud resources according to the migration tasks sent by the task scheduling module;

the starting container module is in communication connection with the adjusting container cloud resource module and is used for starting and running the called migration task in any one mode of a Docker container, a jobs of K8S and a process.

6. The system for migrating the object storage across the cluster mass data according to claim 4 or 5, wherein the migration task scheduling unit comprises:

the task scanning and judging module and the migration task determining module; wherein,

the task scanning and judging module is in communication connection with the migration task determining module, can scan the task queue stored in the back-end database according to preset time, and sends the migration tasks in a waiting state and a suspension state to the migration task determining module;

and the migration task determining module schedules and executes the migration task sent by the task scanning and judging module.

7. The system for migrating the mass data of the object storage across the clusters according to claim 4 or 5, wherein the migration task execution unit comprises:

the system comprises a migration plug-in selection module and a plurality of migration plug-in modules; wherein,

the migration plug-in selection module is respectively in communication connection with each migration plug-in module and can start the migration plug-in modules of corresponding types according to the types of the migration tasks to perform migration operation;

and each migration plug-in module can complete data migration in a multi-level task mode after being started by the migration plug-in selection module.

8. The system for migrating object storage across cluster mass data according to claim 7, wherein the migration task execution unit further comprises:

the fault recovery module is respectively in communication connection with the back-end database and each migration plug-in, and is used for inserting check points into the back-end database according to preset time length when the plug-in module migrates all the scheduled files, and when the file migration tasks are rescheduled after the fault, restarting operation from the last finished checkpoint;

the check point is an internal event, and the event triggers a back-end database writing process to write out the dirty data blocks in the data buffer into the data file after being activated.

Images