KR20040076385A - An apparatus and method for mirroring file system of high availability system - Google Patents

Abstract

The present invention relates to a file system mirroring apparatus and a mirroring method of a high availability system for reducing the number of business take-over due to the file system failure and to increase the available time of the service, the file system mirroring of the high availability system according to the present invention The method includes: a first step of performing a requested task based on task state information stored in a memory of each node and state information of each node; A second step of detecting a file state change of each of the file system which is the business component and mirroring the file to a designated file system during the work execution; Determining whether the file system is in an error state and storing a result of the determination in a memory; And a fourth step of handling the failure according to the type of failure of the file system determined to be the failure, improving the operational reliability of the service application by recovering the failure of the file system, and efficiently configuring the disk and response time of the failure. It is a very useful invention that can be shortened to provide the requested service more quickly.

Description

An apparatus and method for mirroring file system of high availability system

The present invention relates to a file system mirroring apparatus and a mirroring method of a high availability system for reducing the number of business take-over due to the file system failure and to increase the available time of the service.

1 illustrates a network configuration of a high availability (HA) system 100 including N server systems (Nodes), wherein the N nodes (A Node, B Node, C Node, .., Each of the N-1 nodes and N nodes is a node manager (NM) 10, a task manager 20, and an information manager IM as shown in FIG. 30, and a shared memory 40 for storing task status information and node information.

The high availability system 100 continuously checks each node's operation status between nodes in order to respond to a service (business) request of an external client, and whether there is a problem with a running service or a problem with a running node. If a problem occurs, it is possible to perform continuous service by switching the requested service right to a designated node, in which case there are three management daemons located in each node, that is, the NM 10 and TM ( 20), it is processed by the IM 30.

2 shows an operation process of a node by the high availability system 100. For convenience of description, only an operation process of a node A and a node B is shown.

As shown in FIG. 2, first, an HA master starts HA, and then steps S1 of activating the NM 10, the TM 20, and the IM 30, respectively, the task collected by the TM 20. Step S2 of storing the state information in the shared memory 40, step S3 of the IM 30 receiving node information of another connected node, and the IM 30 receiving the node information received from the shared memory 40 In step S4, the NM 10 refers to the state information of the local task stored in the shared memory 40, the step S5 refers to the state information of the local task stored in the shared memory 40, Step S6, which leads the state information of the node, NM (10), S7 step of multi-casting the entire information of their own node based on the information read in the steps S5 and S6 (Multi-casting) to other nodes connected, Step S8 in which the TM 20 manages (performs, suspends, monitors) a task according to the information transmitted from the NM 10, and the NM 10 refers to the information of the IM 30 for decision making. Step S9, and when the NM (10) and the IM (30) is disabled, the step S10 is performed by the TM (20).

As described above, when the HA starts, the NM 10, the TM 20, and the IM 30 operate, respectively, and when a problem occurs, the task is taken over in the order specified in the system configuration file. This will be described in detail as follows.

As shown in Fig. 1, in the state where each node performs each assigned task (a to m), if no problem occurs, no task takeover is performed between the nodes. That is, when an external client requests a service, node A knows that it should handle it and responds appropriately to the service request. However, if there is no response to the service request, the high availability system 100 takes over the task (IP, FS, AP) of the node A to another node in the order specified in the configuration file.

Before the task takeover situation occurs, the TM node 20 of the A node checks the status of its own task and kills it. If the problem does not occur, the task takeover does not occur. However, if a situation that cannot be solved such as a network failure occurs, the task is turned over in the order specified in the configuration file.

The TM 20 is always in a monitoring state, and upon receiving a service request from the outside, the TM 20 immediately performs the above-described operation.

If a problem occurs in a node other than a task (regardless of the service request), that is, if the signal is checked first by checking the node active state between nodes through the heartbeat, Since the node is active, no task takeover occurs. However, if the node cannot be confirmed due to the dead state or short-circuit of the heartbeat cable, the node that wants to take over the task is confirmed by pinging secondly. When received, the task is not taken over because the other node is active. And finally, it checks the file system (File System) that is the Task component.

If there is more than a heartbeat or ping, it checks the filesystem of the other node through the SCSI channel. If this filesystem is active, it is simply not able to verify with the heartbeat and ping between nodes. It can be said that normal service is performed. If the file system operation cannot be confirmed above, the node takes over because the node is dead.

On the other hand, the file system failure checking method performed in the high availability system 100 of FIGS. 1 and 2 searches for a file holding the mount information of the node. If the mount point is not mounted, the file system failure is detected. In this case, file system failure is recognized, but task takeover does not always occur. Task takeover occurs only when the unmounted state of the file system affects the operation of the service application, which is recognized as a failure of the service application.

The case of affecting the operation of the service application is a case in which a file system having an executable file of a service application is affected when the daemon is registered to check the state of the service application in a configuration file. It is not mounted.

If the unmounted file system contains only data and this data is recorded by the service application, the task application does not occur because the service application daemon is running. In this case, even if the service application is running, normal service operation is not possible. As described above, the task availability of the high availability system 100 due to the failure of the file system is frequent, so that the service availability time is reduced.

Accordingly, the present invention was created to solve the above problems, and in the high availability system, a high availability system that reduces the number of business takeovers due to file system failures and increases service availability time through file system mirroring. Its purpose is to provide a file system mirroring device and a mirroring method of the system.

Figure 1 shows a network configuration of a high availability system consisting of N server systems,

FIG. 2 illustrates a detailed configuration and operation process of a node constituting the system of FIG.

Figure 3 shows the detailed configuration and operation of the file system mirroring apparatus of the high availability system according to the present invention,

4 is a diagram illustrating an internal configuration and operation process of a file system mirroring unit constituting the file system manager of FIG. 3, a file system reporting unit, a file system failure checking unit constituting the task manager, and a failure management unit constituting the node manager. It is.

※ Explanation of code for main part of drawing

10: Node Manager 11: File status detection unit

12: change file storage unit 13: processing unit

14: mirroring file storage unit 15: mirroring data storage unit

16: mirroring result output unit 20: task manager

21: mirroring result input unit 22: fault diagnosis result input unit

23: Error processing result input unit 24: Result screen output unit

30: Information Manager

31: file system check unit 32: system signal receiving / analysis unit

33: failure determination unit 34: failure diagnosis output unit

40: Shared Memory 41: Fault Input

42: failure type / process determination unit 43: file system reset unit

44: fault handling output unit

50: File System Manager

51: File System Mirroring Part

52: File System Reporting Part

53: File System Fault Checking Part

54: File System Fault Handing Part

100: High Availability System

File system mirroring apparatus of a high availability system according to the present invention for achieving the above object, the task manager for performing a task operation task or a task operation state monitoring task to perform or stop the work of the local node; An information manager for performing a task of receiving multicasted state information from another node, a task of inspecting a received data, and storing the received data in a memory; A node manager which collects state information of the local node and multicasts the state information to another node; And a file system manager for performing a task of mirroring a file system designated as a service file system of a local operation to a designated file system, and outputting the service file system check result and a failure response result of the service file system. It is characterized by being

In addition, the file system mirroring method of the high availability system according to the present invention, the first step of performing the requested task based on the state information for each node and the state information stored in the memory of each node; A second step of detecting a file state change of each of the file system which is the business component and mirroring the file to a designated file system during the work execution; Determining whether the file system is in an error state and storing a result of the determination in a memory; And a fourth step of handling the failure according to the type of failure of the file system determined to be the failure.

Hereinafter, an embodiment of a file system mirroring apparatus and a mirroring method of a high availability system according to the present invention will be described in detail with reference to the accompanying drawings.

First, the file system mirroring apparatus of the high availability system according to the present invention is configured in each node of FIG. 1, and the file system mirroring apparatus configured in each node is shown in FIG. In addition to the manager (NM) 10, the work manager (TM) 20, the information manager (IM) 30, and the shared memory 40, a file system manager (FSM) 50 is further included. It is composed.

Each node of the high availability system 100 is interconnected by the heartbeat for information transmission, the service to the outside is performed through the connected service network, such a connection is made by a network card mounted on the system node.

In addition, the disk of the high availability system 100 requires one extra file system for the duplication of the file system.

Meanwhile, FIG. 3 illustrates an operation process of a node by the high availability system 100. For convenience of description, only an operation process of a node A and a node B is shown.

In the high availability system 100 in which the file system mirroring apparatus according to the present invention is configured, the HA master first starts HA, and the NM 10, TM 20, IM 30, and FSM 50. Step S1 for performing the daemon, step S2 for storing the task state information collected by the TM 20 in the shared memory 40, step S3 for the IM 30 to receive node information of another connected node, IM Step S4 of storing the received node information in the shared memory 40, step S5 of NM 10 referencing state information of a local task stored in the shared memory 40, and NM (30). 10), the step S6 of reading the status information of the other node stored in the shared memory 40, the NM (10) is connected to the entire node information of the own node organized based on the information read in the steps S5 and S6 In step S7 of multicasting to a node, the TM 20 manages a task according to information transmitted from the NM 10 (performs, stops, and monitors). In step S8, the NM 10 refers to the information of the IM 30 for decision making. In the step S9, when the NM 10 and the IM 30 fail, the step S10 performed by the TM 20, TM Step 20 in which S20 stores the result in the shared memory 40 after checking the file system, and step S12 in which the NM 10 processes the failure after checking the file system failure in the shared memory 40, FSM. 50 includes a step S13 of reading a file system related result value from the shared memory 40 and outputting the result.

4 is a file system mirroring part 51 constituting the FSM 50, a file system reporting part 52, and a file system constituting the TM 20. FIG. The internal configuration and operation process of the File System Fault Checking Part 53 and the File System Fault Handing Part 54 constituting the NM 10 are described. Detailed internal configuration and operation process will be described in detail later.

The file system mirroring unit 51 is a part for mirroring (backup) the redundant file system whenever a file system of the high availability system 100 is changed, and the file system failure checking unit 53 is the conventional file described above. The system checks various failures of the file system, including a system check method, and stores the failures in the shared memory 40 accordingly.

In addition, the file system failure management unit 54 is a part for restoring a task to a mirroring file system in the event of a failure by reading whether the file system has failed from the information stored in the shared memory 40, the file system report unit 52 Denotes a part of reporting the results of the file system mirroring unit 51, the file system failure checking unit 53, and the file system failure managing unit 54 to the user.

In addition, the file systems # 1, # 2, # 3, and # 4 shown in FIG. 4 are located in a partial region of the shared hard disk of FIG.

Hereinafter, a file system mirroring method of a high availability system according to the present invention will be described in detail with reference to the configuration and operation of FIGS. 3 and 4.

First, when the HA master operates the NM 10, the TM 20, the IM 30, and the FSM 50, respectively, the managers that operate operate the information transmission and information for operating the high availability system 100. Start the acquisition.

The TM 20 is responsible for the task operation task for performing or stopping the task of the local node according to the command of the NM 10 and the operation state monitoring task of the task, and the IM 30 is multicasting from another node. Responsible for receiving the received status information, the inspection operation of the received data, and storing the received data in the shared memory 40.

In addition, the TM 20 checks the states of the NM 10 and the IM 30 and re-executes the NM 10 and the IM 30 if they are in a dead state. While collecting state information and performing multicasting to other nodes, the state of the TM 20 and the FSM 50 is examined and re-run if the two managers are dead.

The FSM 50 mirrors a file system (for example, file system # 1) designated as a service file system of a local task to a designated extra file system (for example, file system # 2), A file system mirroring unit 51 constituting the FSM 50, a file system reporting unit 52, and the TM An internal configuration and an operation process of the file system failure checking unit 53 constituting the 20) and the failure management unit 54 constituting the NM 10 will be described in detail as follows.

First, the file system mirroring unit 51 includes a file state detection unit 11 for detecting a change state of a file system in real time, and a change file storage unit for storing information received from the file state detection unit 11 ( 12), a processing unit 13 for processing and controlling overall operations related to file backup, a mirroring file storage unit 14 for storing a list of files to be backed up, and mirroring data for storing data of the mirroring file. And a storage unit 15 and a mirroring result output unit 16 for outputting a backup result performed by the processing unit 13 so that the file state detection unit 11 can change the files on the file system. Detects in real time (F1) and transmits the file whose state change is detected to the change file storage unit 12 (F2), and the change file storage unit 12 stores the change file identifier continuously received (F3). . In this state, when the backup operation is started, the processing unit 13 reads the file list from the mirroring file storage unit 14 (F4) and only the file corresponding to the identifier stored in the change file storage unit 12 from the file system. Read it and transmit it to the mirroring data storage unit 15 (F5) to be stored (F6).

The file system report unit 52 may include a mirroring result input unit 21 for receiving a result of the file system mirroring unit 51 and a failure diagnosis result for receiving a result of the file system failure checking unit 53. The input unit 22, a failure processing result input unit 23 for receiving a result of performing the file system failure operation unit 54, and the inputs of the input units 21, 22, and 23 are processed and a result screen is output. And a result screen output section 24, wherein each of the input sections 21, 22, and 23 periodically approaches the corresponding location of the shared memory 40 to read out the stored execution results (R1 to R3). The result screen output unit 24 outputs the execution result transmitted from each of the input units 21, 22, and 23 (R4).

The file system failure checking unit 53 continuously checks the current mount state of the file system (C1), and checks the file system-related error signal transmitted by the operating system (C2). A failure determination unit 33 for determining a failure of the file system (C3) based on the signal reception / interpretation unit 32, the result values of the file system inspection unit 31 and the system signal reception / interpretation unit 32, and And a failure diagnosis output section 34 for storing the failure determination result transmitted from the failure determination unit 33 (C4) to the shared memory 40 (C5).

The fault management unit 54 continuously checks the fault state of the file system from the shared memory 40 (H1) and the fault input unit 41, and determines the type and method of the fault (H2). A type / process determination section 42, a file system reset section 43 for resetting the file system to a mirroring file system (H3), and a fault for storing a result of the failure processing in the shared memory 40 (H4) The process output part 44 is comprised.

In the above, for example, when any one node is executing a task including file systems "/ app" and "/ data" as components, the file system mirroring unit 51 is assigned "/ app" and " / data "Detect changes in each file state and apply the same change to the extra file system specified to mirror the file.

The file system failure checking unit 53 is a part for checking a service file system of a local node, and a method for checking a mount state of a file system and a method for analyzing and analyzing a system signal generated by an operating system. The determination unit 33 determines whether or not the inspection result is a disabled person, and the determination result is stored in the shared memory 40 through the failure diagnosis output unit 34.

The file system failure management unit 54 performs a task of identifying a type of a failure and a corresponding operation with respect to the service file system determined to be a failure, wherein the failure of the file system is solved by replacing the mirroring file system. If so, the file system reset unit 43 instructs the TM 20 to reset the service file system. In this case, the file system reset unit 43 resets the mirroring file system (for example, file system # 2) to the service file system. do.

If the mirrored file system is not useful, such as a disk cable failure, it responds to take over the task to another node.

As described above, the file system mirroring according to the present invention is a function that can be provided by using physical RAID, and if the software of the high availability system 100 has its own function according to the present invention, the mirroring target You can optionally configure the filesystem as the primary filesystem.

In addition, if only the service application daemon is alive, the software of the conventional high availability system 100 determines that the service is a normal service without considering a fault of the data file system in which the service application records the actual data. In the present invention, the failure of the data file system is detected and restored to the mirroring file system when a failure occurs, so that the daemon of the service application is alive and normal operation of the service application is guaranteed.

In addition, in the prior art, the software of the high availability system 100 detects a failure and responds to the process of the NM 10 detecting a failure of a local task (about 5 seconds), and stops the task in the TM 20. Command (approximately 5 seconds), process TM 20 stops the task (approximately 15 seconds), negotiate another node to perform a local task (approximately 10 seconds), negotiation is set Since another node has to perform the task (about 15 seconds) to start the task, it takes about 50 seconds to take over the task to another node due to the failure of the task. However, in the present invention, in order for the software of the high availability system 100 to detect and respond to a failure, a process of detecting a failure of a local task (about 5 seconds) and a process of resetting a task on the TM 20 (about, 10) Second) only need to be performed, the time required compared to the conventional 70% is reduced.

On the other hand, the application of the present invention to write data to the file system at the same time may occur twice the overhead than the current due to the occurrence of double input and output, but this is the buffering (Buffering) and the methodology (methodology) It can be reduced by applying the same function.

The above-described preferred embodiments of the present invention are disclosed for purposes of illustration, and those skilled in the art can improve, change, and substitute various other embodiments within the technical spirit and scope of the present invention disclosed in the appended claims below. Or addition may be possible.

The file system mirroring apparatus and the mirroring method of the high availability system according to the present invention configured as described above improve the operational reliability of the service application by failback of the data file system, reduce the effective configuration of the disk and reduce the response time of the failure. It is a very useful invention that can provide the requested service more quickly.

Claims (10)

In a high availability system consisting of multiple nodes, A task manager performing a task operation task for performing or stopping a task of a local node and a task operation state monitoring task; An information manager for performing a task of receiving multicasted state information from another node, a task of inspecting a received data, and storing the received data in a memory; A node manager which collects state information of the local node and multicasts the state information to another node; And And a file system manager configured to mirror a file system designated as a service file system of a local business to a designated file system, and output a result of the service file system check and a result of responding to a failure of the service file system. File system mirroring device in a high availability system. The method of claim 1, And each of the plurality of nodes is interconnected by a heartbeat for information transmission. The method of claim 1, The task manager comprises a file system failure checking means for checking a service file system of a local node to determine whether there is a failure and storing the determination result in the memory. The node manager is configured to include a file system failure management means for determining a processing method for the file system determined as the failure and storing a result of the failure processing according to the determined processing method in the memory. The file system manager may further include: file system mirroring means for continuously detecting a file state change of each file system, mirroring the file system, and storing the mirroring result in the memory; And And a file system reporting means for receiving the processing results of the file system mirroring means, the file system failure checking means, and the file system failure operating means through the memory, processing them, and outputting a result screen. File system mirroring device in a high availability system. The method of claim 3, wherein The file system failure checking means may include: a file system checking unit for continuously checking a current mount state of the file system; A system signal reception / interpretation unit for checking a file system related error signal transmitted from an operating system; A failure determination unit that determines whether or not the file system is damaged from the inspection result values of the file system checking unit and the system signal receiving / interpreting unit; And And a failure diagnosis output unit for storing the determination result of the failure determination unit in the memory. The method of claim 4, wherein The file system fault management means may include a fault input unit configured to continuously check whether the file system fails from the memory; A failure type / process determination unit for determining a failure type and a processing method according to the file system determined as a failure according to the confirmation; A file system reset unit for resetting the file system to the mirroring file system according to the determined processing method; And File system mirroring apparatus of the high availability system, characterized in that it comprises a failure processing output unit for storing the result of the failure processing according to the determined processing method in the memory. The method of claim 5, The file system mirroring means may include: a file state detection unit configured to detect a change state of the file system in real time; A change file storage unit for storing the information transmitted from the file state detection unit; A mirroring file storage unit for storing a list of files to be backed up; A mirroring data storage unit for storing data of the mirroring file; A processing unit controlling overall details of the file backup; And And a mirroring result output unit for outputting a backup result performed by the processing unit. The method of claim 6, The file system report means may include: an input unit configured to receive, through the memory, a result of performing the file system mirroring means, the file system failure checking means, and the file system failure operating means; And And a result screen output unit for processing the received results and outputting the result screens. In the service method of a high availability system composed of a plurality of nodes, A first step of performing a requested task based on task state information stored in a memory of each node and state information of each node; A second step of detecting a file state change of each of the file system which is the business component and mirroring the file to a designated file system during the work execution; Determining whether the file system is in an error state and storing a result of the determination in a memory; And And a fourth step of handling the failure according to the failure type of the file system determined as the failure. The method of claim 8, In the fourth step, in the case where the failure type of the file system determined to be a failure is solved by replacing a mirroring file system, the file system of the high availability system comprises resetting the file system to the mirrored file system. Mirroring method. The method of claim 8, In the fourth step, if the failure type of the file system determined to be a failure is a disk cable failure, the file system mirroring method of the high availability system, characterized in that the task is handed over to another node.