JP2014078067A - Database system, database device, failure recovery method for database and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a database system capable of executing learned failure recovery processing even at the occurrence of failures in a plurality of servers.SOLUTION: In a primary site and a secondary site each including an application server and a database server, each server includes a server state monitoring part. The server state monitoring part includes: a command learning function for storing learning data obtained by extracting a specific pattern from a command input by a maintenance person at the occurrence of a failure in association with an error code; a failure detection function for detecting that the failure has occurred in any server; an error pattern search function for searching whether or not the same error code as the detected failure is stored in the learning data; and a learned command execution function for executing the command of the pattern corresponding to the error code.

Description

  The present invention relates to a database system, a database apparatus, a database failure recovery method, and a program, and more particularly to a database system that learns and automates failure recovery processing.

  In a computer system used in a company or the like, even if it is stopped for a short period of time, a huge loss can occur due to the suspension of the business that occurred during that time. In particular, a database that stores and handles large amounts of data and a company system that operates application software that uses these data to perform processes related to the company's activities must never stop, and data loss There must be no such thing.

  For this reason, when something goes wrong with these computer systems, it is desired to recover as soon as possible. However, the scale of the system itself or the amount of data to be handled is increasing enormously. Therefore, the time and the amount of work required for restoration are also increasing enormously.

  In particular, after the Great East Japan Earthquake that occurred on March 11, 2011, even if such a computer system encounters a large-scale disaster such as an earthquake or storm and flood damage, data loss can be prevented and processing can continue. The importance of so-called disaster recovery has been screaming loudly.

  Therefore, in recent years, development of a technique for reducing a load on a maintenance worker, in particular, a technique for automatically recovering from a failure without the intervention of the maintenance worker has been actively promoted. There are the following technical documents as related documents. Among them, Patent Document 1 describes a failure recovery apparatus that registers a recovery procedure performed on a computer system in preparation for the next occurrence of the same failure.

  Patent Document 2 describes a management system that automatically executes a failure recovery program when a failure occurs in a job network. Patent Document 3 describes an automatic monitoring and recovery system in which processing services provided by a plurality of servers are monitored and recovered by a single monitoring device.

Japanese Patent Laid-Open No. 10-091475 JP 2004-318863 A JP 2002-342180 A

  FIG. 13 is an explanatory diagram showing the overall configuration of the database system 901 according to the existing technology. The database system 901 is configured by connecting a primary site 902 (active system) and a secondary site 903 (standby system) via a network 904.

  The primary site 902 includes an application server 911, a database server 912, and a storage device 913. Similarly, the secondary site 903 includes an application server 921, a database server 922, and a storage device 923. Further, the application servers 911 and 921 constitute a server cluster with each other, and similarly, the database servers 912 and 922 also constitute a server cluster with each other.

  The application servers 911 and 921 can monitor the operation of each other, and if one of them is abnormally stopped, only the remaining one can continue the operation. The database servers 912 and 922 can also monitor the operation of each other, and if one of them stops abnormally, only the remaining one can continue the operation.

  In recent years, in particular, with the spread of e-commerce (electronic commerce), an application server that operates application software such as a web server and a database server that handles a large amount of data about products and services traded there are combined. Many structured web systems are now working. In such a case, as in the database system 901 shown in FIG. 13, the application server and the database server are duplicated and monitored each other, and the standby system can take over the process whenever an abnormality occurs in the active system. Structured systems are being used.

  However, in such systems, switching from the active system to the standby system still requires an operation by a maintenance worker. This operation is inevitably complicated, and mistakes due to human errors are likely to occur.

  In particular, due to hardware and software configurations, there are inevitably specific faults that are likely to occur particularly frequently in the system, but the response to those specific faults can be patterned to some extent through experience. However, there is no prior art that forms a pattern that can be automatically executed.

  The above-mentioned Patent Documents 1 to 3 describe that “learning a failure recovery operation and automatically performing a recovery operation for the same failure”, but the configuration of the database system 901 shown in FIG. Applicable items are not listed. In particular, any one of Patent Documents 1 to 3 that can detect and automatically recover when one or more failures occur in both the application server and the database server. Absent.

  An object of the present invention is to provide a database system and a database apparatus capable of executing a learned failure recovery process even when a failure occurs in both one or more application servers and database servers that monitor each other. Another object of the present invention is to provide a database failure recovery method and program.

  In order to achieve the above object, a database system according to the present invention connects a primary site including an active application server and an active database server and a secondary site including a standby application server and a standby database server via a network. Each of the active application server, the active database server, the standby application server, and the standby database server each includes a server status monitoring unit that monitors each other, Each server status monitoring unit records a command entered by the maintainer when a failure occurs in any of the servers that make up the database system, and generates learning data by extracting specific patterns from this log. A command learning function that associates an error code indicating a failure symptom with a storage means prepared in advance and sends this learning data to other servers for storage, and a failure occurs in any server A failure detection function that detects the failure, an error pattern search function that searches whether or not the same error code as the detected failure is stored in the learning data, and an error code that is the same as the detected failure And a learned command execution function for causing the active application server to execute a command having a pattern corresponding to the error code.

  In order to achieve the above object, a database apparatus according to the present invention connects a primary site including an active application server and an active database server and a secondary site including a standby application server and a standby database server via a network. A database system that can function as an active database server or a standby database server, and includes a server status monitoring unit that monitors each other, and the server status monitoring unit An error that indicates the symptom of the failure that generated the learning data that recorded a specific pattern extracted from the log when the command entered by the maintainer when a failure occurred in any of the servers that make up the database system With code A command learning function for storing in a storage means provided in advance and sending this learning data to other servers for storage, and a failure detection function for detecting that a failure has occurred in any of the servers And an error pattern search function for searching whether or not the same error code as the detected failure is stored in the learning data, and when the same error code as the detected failure is stored in the learning data, And a learned command execution function for causing the active application server to execute a command having a pattern corresponding to the error code.

  In order to achieve the above object, a failure recovery method according to the present invention includes a primary site including an active application server and an active database server, and a secondary site including a standby application server and a standby database server via a network. In a connected database system, a command entered by a maintenance person when a failure occurs in any of the servers that make up the database system is recorded as a log by the command learning function of each server that makes up the database system. In addition, the learning data obtained by extracting a specific pattern from the recorded log is associated with the error code indicating the symptom of the failure that has occurred, and stored in the storage means provided in advance with the command learning function of each server. This server also stores this learning data The failure detection function of each server detects that a failure has occurred in any server, and whether or not the same error code as the detected failure is stored in the learning data. When the error pattern search function searches and the same error code as the detected fault is stored in the learning data, the command of the pattern corresponding to the error code is used by the learned command execution function of each server. It is characterized by causing a server to execute.

  In order to achieve the above object, a failure recovery program according to the present invention includes a primary site including an active application server and an active database server, and a secondary site including a standby application server and a standby database server via a network. In a connected database system, a command input by a maintenance person when a failure occurs in any of the servers constituting the database system is recorded as a log in a processor included in each server constituting the database system. The learning data obtained by extracting a specific pattern from the procedure and recorded log is stored in the storage means prepared in advance in association with the error code indicating the symptom of the failure that has occurred, and this learning is also performed in each of the other servers. Send data and remember A procedure for detecting that a failure has occurred in one of the servers, a procedure for searching whether the same error code as the detected failure is stored in the learning data, and the same as the detected failure When the error code is stored in the learning data, a procedure for causing the active application server to execute a command of a pattern corresponding to the error code is executed.

  In the present invention, as described above, both the active and standby application servers and database servers are configured to include the server state monitoring unit that monitors the occurrence of a failure, so that when a failure occurs, The learned failure recovery operation can be executed from any device. As a result, a database system and a database device having an excellent feature that, even when one or more of the application server and the database server have a failure, the learned failure recovery processing can be executed. A database failure recovery method and program can be provided.

It is explanatory drawing shown about the whole structure of the database system which concerns on embodiment of this invention. It is explanatory drawing shown about the more detailed structure of the primary site shown in FIG. It is explanatory drawing shown about the more detailed structure of the secondary site shown in FIG. FIG. 2 is a sequence diagram illustrating a flow of processing for learning processing performed by each server when all the components are operating normally in the database system illustrated in FIG. 1. It is explanatory drawing shown about an example of the error pattern shown to a maintenance person by the process of FIG.4 S110-114, and the learning data memorize | stored. In the database system shown in FIG. 1, all the server state monitoring units are operating normally, but a sequence diagram illustrating an operation flow when a failure occurs in the database server. FIG. 2 is a sequence diagram showing a flow of operations when a failure occurs in a database server and an application server is not operating normally in the database system shown in FIG. 1. FIG. 2 is a sequence diagram showing a flow of operations when a server state monitoring unit of an active database server is stopped in the database system shown in FIG. 1. FIG. 3 is a sequence diagram showing a flow of operations when a server state monitoring unit of a standby database server is stopped in the database system shown in FIG. 1. FIG. 3 is a sequence diagram showing a flow of processing for learning an operation performed when a server state monitoring unit of an active database server is stopped in the database system shown in FIG. 1. FIG. 3 is a sequence diagram illustrating a flow of processing for learning an operation performed when a server state monitoring unit of a standby database server is stopped in the database system illustrated in FIG. 1. FIG. 3 is an explanatory diagram showing processing performed by a command learning function at an initial stage of system startup in the database system shown in FIG. 1. It is explanatory drawing shown about the whole structure of the database system which concerns on the existing technique.

(Embodiment)
Hereinafter, the configuration of an embodiment of the present invention will be described with reference to FIGS.
First, the basic content of the present embodiment will be described, and then more specific content will be described.
In the database system 1 according to this embodiment, a primary site 2 including an active application server 11 and an active database server 12 and a secondary site 3 including a standby application server 21 and a standby database server 22 are connected via a network 4. It is a database system that is connected to each other. The active application server 11, the active database server 12, the standby application server 21, and the standby database server 22 all include server status monitoring units 102, 112, 202, and 212 that monitor each other. Each server status monitoring unit records a command entered by the maintenance person when a failure occurs in any of the servers that make up the database system, and generates learning data by extracting a specific pattern from this log A command learning function 102a that stores in a storage means prepared in advance in association with an error code indicating a symptom of a failure, and sends the learning data to other servers for storage, and a failure occurs in any of the servers. The failure detection function 102b that detects the occurrence of the failure and the same error as the detected failure -An error pattern search function 102c for searching whether or not a code is stored in the learning data, and a command of a pattern corresponding to the error code when the same error code as the detected fault is stored in the learning data And a learned command execution function 102d for causing the active application server to execute the command.

  Here, the command learning function 102a presents the pattern extracted from the log to the user, and stores the pattern selected by the user as learning data. When the failure detection function 102b detects that a failure has occurred in any of the servers, the learning data is sent to the command learning function 102a after detecting that the failed server has recovered. Let

  Further, when the active application server 11 cannot execute the command, the learned command execution function 102d causes the standby application server 21 to execute the command.

With the above configuration, the database system 1 can execute the learned failure recovery process even when a failure occurs in both of the application server and the database server.
Hereinafter, this will be described in more detail.

  FIG. 1 is an explanatory diagram showing an overall configuration of a database system 1 according to an embodiment of the present invention. The database system 1 is configured by connecting a primary site 2 (active system) and a secondary site 3 (standby system) via a network 4. The network 4 is a network having the same subnet mask.

  The primary site 2 includes an application server 11, a database server 12, and a storage device 13. Similarly, the secondary site 3 includes an application server 21, a database server 22, and a storage device 23. The application servers 11 and 21 constitute a server cluster with each other, and the database servers 12 and 22 similarly constitute a server cluster with each other.

  Each of the primary site 2 and the secondary site 3 includes a number of client devices 14a, b, c,... And 24a, b, c,. .. And 25a, b, c,... Are connected to the application servers 11 and 21, respectively. Since these are not particularly necessary for describing the present embodiment, a detailed description thereof will be omitted here.

  FIG. 2 is an explanatory diagram showing a more detailed configuration of the primary site 2 shown in FIG. The application server 11 and the database server 12 are both general computer devices (server machines). In FIG. 2, the computers are physically separate computers, but the application server 11 and the database server 12 may be configured by the same computer, or the application server 11 or the database server 12 is a combination of a plurality of computers. May be configured virtually.

  The application server 11 includes a processor 11a that is a computer program implementation entity, a storage unit 11b that stores the computer program and operation data, and a communication unit 11c that is connected to the network 4 and performs data communication with other computers. Is provided. Similarly, the database server 12 also includes a processor 12a, which is a computer program executing entity, a storage unit 12b that stores the computer program and operation data, and a communication unit that is connected to the network 4 and performs data communication with other computers. 12c.

  In the processor 11a of the application server 11, an application operation unit 101 that operates an application (specifically, a web server, a business system, etc.) functions based on a processing request from each client device. In the processor 12 a of the database server 12, a database operation unit 111 that stores data processed by an application operating in the application operation unit 101 in the storage device 13 functions.

  The storage device 13 includes a main storage unit 13a using a normal magnetic disk or a semiconductor disk, and a backup storage unit 13b using a tape (QIC, DDS, DLT, etc.) or an optical disk.

  Furthermore, the processor 11a of the application server 11 and the processor 12a of the database server 12 monitor each other's operations, and simultaneously monitor the operations of the application server 21 and the database server 22 at the secondary site 3, and when an abnormality is detected, The server status monitoring units 102 and 112 that recover the server from the abnormality operate simultaneously as server processes.

  The storage means 11b of the application server 11 and the storage means 12b of the database server 12 store learning data 103 and 113 stored by the server state monitoring units 102 and 112, respectively.

  FIG. 3 is an explanatory diagram showing a more detailed configuration of the secondary site 3 shown in FIG. The application server 21 and the database server 22 at the secondary site 3 are both general computer devices (server machines).

  The application server 21, the database server 22, and the storage device 23 all have the same configuration as the application server 11, the database server 12, and the storage device 13 in the primary site 2, but only the reference numbers of the respective elements in terms of hardware. +10, and in terms of software, only the reference number of each element is +100. The names of the elements are all the same. Therefore, the detailed description of those configurations is omitted here.

  The server state monitoring unit 102 of the application server 11 includes various functions such as a command learning function 102a, a failure detection function 102b, an error pattern search function 102c, and a learned command execution function 102d. Since the server status monitoring units 112, 202, and 212 included in the database server 12, the application server 21, and the database server 22 also have these functions, respectively, in this specification, for example, the server status monitoring of the database server 12 will be described hereinafter. The command learning function included in the unit 112 will be referred to as “command learning function 112a”.

  FIG. 4 is a sequence diagram showing the flow of processing for learning the processing performed by each server when all the components are operating normally in the database system 1 shown in FIG. Although FIG. 4 shows an example in which the maintenance person issues a start request to the application server 11 at the primary site 2, this start request may be input by directly operating the application server 11 or on the primary site 2 side. The application server 11 may be accessed remotely from any of the client devices 14 and input. Furthermore, the application server 21 at the secondary site 3 can be operated in the same manner as in FIG. 4 by a similar start request.

  The command learning function 102a of the server state monitoring unit 102 that has received the start request (step S101) starts the recording of the operation log, while the server state monitoring unit of each other server, that is, the primary site 2 side, In the server status monitoring unit 112 and the secondary site 3, a start request is transmitted to each of the server status monitoring unit 202 of the application server 21 and the server status monitoring unit 212 of the database server 22 (steps S102a to c).

  The server status monitoring units 112, 202, and 212 that have received the start request start recording operation logs using the command learning functions 112a, 202a, and 212a, and return a normal response to the start request to the server status monitoring unit 102. (Steps S103a-c). Then, the command learning function 102a of the server state monitoring unit 102 returns a normal response to the maintenance person (step S104).

  With respect to all commands that are input by the user to the application server 11 and appropriately transferred from there to other servers, the command and the operation performed on the command and the response to be returned are respectively The server status monitoring units 102, 112, 202, and 212 record logs (steps S105 to S106). In FIG. 4, detailed descriptions of commands transferred from the application server 11 to other servers and responses from the other servers are omitted.

  Similarly to the start request, when the maintenance person inputs an end processing request (step S107), the server state monitoring unit 102 ends the recording of the operation log, and at the same time, each of the other server state monitoring units 112, 202, and 212. An end processing request is transmitted to the server (steps S108a to c). The other server status monitoring units 112, 202, and 212 that have received the termination processing request each end the recording of the operation log, and transmit the recorded operation log to the server status monitoring unit 102 (steps S109a to S109c).

  The server state monitoring unit 102 extracts an error pattern from the operation log recorded by itself and the operation log received from others and presents it to the maintenance person (step S110). Details of this operation will be described later.

  Then, the maintenance person selects and inputs an error pattern to be learned from among them (step S111). The server status monitoring unit 102 stores the selected error pattern as the learning data 103 and sends the error pattern to each of the other server status monitoring units 112, 202, and 212 (steps S112a to c). . Each server status monitoring unit 102, 112, 202, 212 receives the error pattern, stores it as learning data 113, 203, 213 (step S113a-c), and returns a normal response (step S114). .

  FIG. 5 is an explanatory diagram showing an example of the error pattern presented to the maintenance person in the process of steps S110 to 114 of FIG. 4 and the learning data 103, 113, 203, 213 stored. In the processing shown in step S110, the target device 120a and the error code 120b of the error that has occurred and the command 120c actually input to each device are presented to the maintenance person.

  On the other hand, the maintenance person selects what should be actually input in the command 120c in the process shown in step S111, and also selects elements such as “host name” and “process name” as “target”. Replace it with a variable that indicates "device" or "target process". Thus, the input data becomes a learning pattern 120d, is associated with the error code 120b, and is stored as learning data 103, 113, 203, 213 of each device in the processing of steps S112 to S114.

  Here, the learning data actually created is different for each device. This is because different commands should be executed on the application server and the database server, and different commands should be executed on the primary site and the secondary site as well. However, since the difference in commands to be executed for each apparatus is well known to those skilled in the art, a detailed explanation is omitted here.

  FIG. 6 shows the operation flow when all of the server status monitoring units 102, 112, 202, and 212 are operating normally in the database system 1 shown in FIG. FIG. In the database server 12, since the server state monitoring unit 112 operates as a server process, the failure detection function 112b detects the occurrence of this failure (step S201).

  When the failure detection function 112b detects the occurrence of this failure, the error pattern search function 112c that has responded to this searches for whether or not the same error code 120b is registered in the learning data 113 (step S202). . If not registered, the application server 11 is instructed to stop the process, and the maintenance person is notified of the failure (step S203). In that case, the maintenance person performs the operation shown in FIG. 4 to deal with the failure for each server and to learn about the operation for dealing with the failure.

  If the same error code 120b is registered in the learning data 113, the error pattern search function 112c sends the command of the learning pattern 120d corresponding to the error code 120b to the learned command execution function 102d of the application server 11. In response to this, the learned command execution function 102d reads the learning pattern 120d from the learning data 113 and sends the learned pattern execution function 102d, 112d, 202d, 212d of each server to the learning pattern 120d. Are executed (steps S204 to S205).

  Here, the subject that starts execution of the learned command may be the database server 12, but the application server 11 is assumed as the subject here, assuming that the operation of the database server 12 is stopped. Basically, this execution start subject may be any server.

  Even during the execution, the command learning function 102a, 112a, 202a, 212a records a log in the same manner as in steps S105 to S106 in FIG. 4, so when the execution of the learned command is completed, the learned command execution function is completed. 102d instructs the other command learning functions 102a, 112a, 202a, and 212a to end (steps S206a to S206c), and each command learning function 102a, 112a, 202a, and 212a returns the recorded log accordingly. (Steps S207a-c). Then, the learned command execution function 102d notifies the user of the processing result (step S208).

  FIG. 7 is a sequence diagram showing an operation flow when a failure occurs in the database server 12 and the application server 11 is not operating normally in the database system 1 shown in FIG. In this case, the operation is the same as that in FIG. 6 until step S202. However, in step S203, the error pattern search function 112c instructs the application server 11 to stop processing, or in step S204, the learned command execution function 102d has an error. Although an instruction is given to execute the command of the learning pattern 120d corresponding to the code 120b, a normal response to this is not returned.

  Therefore, if the learning pattern 120d corresponding to the error code 120b is not registered in the learning data 113, the error pattern search function 112c transmits the same processing stop instruction as that in step S203 to the application server 21 of the secondary site 3. Then, the maintenance person is notified of the occurrence of the failure (step S251). When the learning pattern 120d corresponding to the error code 120b is registered in the learning data 113, the same command execution instruction as in step S204 is transmitted, and the command of the learning pattern 120d is sent to the learned command execution function 202d. It is made to perform (step S252-3).

  Thereafter, the learned command execution function 202d of the application server 21 performs the same operation as that after step S205 in FIG. However, since the application server 11 is not operating normally, there is a possibility that the log in step S207a may not be transmitted normally and time out. In that case, the fact is recorded in a log and the maintenance person is notified.

  FIG. 8 is a sequence diagram showing an operation flow when the server state monitoring unit 112 of the active database server 12 is stopped in the database system 1 shown in FIG. Since the failure detection function 102b of the application server 11 periodically transmits a state monitoring signal to the server state monitoring units 112 and 212 of the database servers 12 and 22, there is no response to the state monitoring signal. Thus, the stop of the server state monitoring unit 112 is detected (step S301).

  When the failure detection function 102b detects the occurrence of the failure, the error pattern search function 102c that responds to the failure searches whether or not the same error code 120b as this error is registered in the learning data 103 (step S302). ). If it is not registered, as in step S203 of FIG. 6, its own processing is stopped and the occurrence of the failure is notified to the maintenance person (step S303).

  If the same error code 120b is registered in the learning data 113, the learned command execution function 102d of the application server 11 executes the command of the learning pattern 120d corresponding to the error code 120b, and the steps of FIG. Similarly to S204 and subsequent steps, the other learned command execution functions 112d, 202d, and 212d are caused to execute the command of the learned pattern 120d. The subsequent operation is the same as that shown in FIG.

  FIG. 9 is a sequence diagram showing an operation flow when the server state monitoring unit 212 of the standby database server 22 is stopped in the database system 1 shown in FIG. This is the same operation as that in FIG. 8 except that the failure detection function 102b of the application server 11 detects the stop as the server status monitoring unit 212 (step S351).

  FIG. 10 is a sequence diagram showing a flow of processing for learning an operation performed when the server status monitoring unit 112 of the active database server 12 is stopped in the database system 1 shown in FIG. More specifically, when the maintenance person is notified in step S303 in FIG. 8 that the error code 120b is not registered, it is necessary to learn the processing to be performed for this state.

  Also in that case, the operation performed on each server is learned by the same operation as in FIG. Then, after the learning operation is completed, an operation for monitoring whether or not the server state monitoring unit 112 has been restored to the failure detection function 102b of the application server 11, more specifically, a state monitoring signal is periodically transmitted. The transmission operation is started (step S401).

  If a normal response to this signal is returned, it is determined that the server state monitoring unit 112 has recovered (step S402), and the command learning function 102a sends the error pattern that could not be transmitted in step S112 to the application server 11. (Step S403). Receiving this, the command learning function 112a stores the received error pattern as learning data 113 and returns a normal response (step S404).

  FIG. 11 is a sequence diagram illustrating a flow of processing for learning an operation performed when the server state monitoring unit 212 of the standby database server 22 is stopped in the database system 1 illustrated in FIG. 1. This is the same as FIG. 10 except that the object for detecting whether or not the failure detection function 102b of the application server 11 has been restored is changed to the server state monitoring unit 212 (steps S451 to 454).

  FIG. 12 is an explanatory diagram showing processing performed by the command learning function 102a in the initial stage of system startup in the database system 1 shown in FIG. First, the command learning function 102a of the application server 11 requests configuration information unique to each server from the other command learning functions 112a, 202a, and 212a (steps S501a to S501c). The command learning functions 112a, 202a, and 212a each transmit this configuration information in response to this request (steps S502a to c).

  Receiving this, the command learning function 102a collects each received configuration information, creates the configuration information of the entire database system 1 and stores it as the learning data 103, and stores it as the command learning function 112a, 202a, 212a. (Steps S503a to S503). The command learning functions 112a, 202a, 212a store these as learning data 113, 203, 213 and return normal responses (steps S504a to S504c). As described above, the learning data actually created is different for each device, but the difference in the command to be executed is well known to those skilled in the art, so a detailed explanation is omitted here. To do.

(Overall operation of the embodiment)
Next, the overall operation of the above embodiment will be described.
In the failure recovery method according to the present embodiment, the primary site 2 including the active application server 11 and the active database server 12 and the secondary site 3 including the standby application server 21 and the standby database server 22 are connected via the network 4. In the database system 4 configured to be connected to each other, a command input by a maintenance person when a failure occurs in any of the servers constituting the database system is used as a log for the command learning function of each server constituting the database system. A command learning function of each server is recorded in advance (FIG. 4, steps S101 to S109), in which learning data obtained by extracting a specific pattern from the recorded log is associated with an error code indicating the symptom of the failure that has occurred. When stored in the storage means In addition, the learning data is sent to other servers for storage (FIG. 4, steps S110 to 114), and the failure detection function of each server detects that a failure has occurred in any server (FIG. 6). Step S201), the error pattern search function of each server searches whether or not the same error code as the detected fault is stored in the learning data (FIG. 6, Step S202), and is the same as the detected fault When the error code is stored in the learning data, the learned command execution function of each server causes the active application server to execute the command of the pattern corresponding to the error code (steps S204 to 205 in FIG. 6).

Here, each of the above operation steps may be programmed so as to be executable by a computer, and these may be executed by a processor of each server that directly executes each of the steps. The program may be recorded on a non-temporary recording medium, such as a DVD, a CD, or a flash memory. In this case, the program is read from the recording medium by a computer and executed.
By this operation, this embodiment has the following effects.

  According to the present embodiment, if a pattern corresponding to the error code of the failure that has occurred is registered, it is possible to automatically execute the pattern and reduce the effort required for recovery. In addition, since the recovery by the pattern can be executed from any of the active and standby application servers and the active database server, the pattern can be executed even when two or more servers are simultaneously failed.

  Therefore, for example, in the present embodiment, the restoration work, which requires about several tens of commands and several hours in the existing technology, can be completed in several minutes by executing several commands at most. it can. Furthermore, particularly frequently occurring symptoms can be learned and automated immediately, and can be automatically restored without requiring any operation by the maintenance personnel. That is, according to the present embodiment, it is possible to significantly reduce the complicated work that causes human error and to smoothly cope with the failure that has occurred.

(Extended embodiment)
The above-described embodiment can be variously expanded without departing from the spirit of the present invention described above. This will be described below.
First, in the above-described embodiment, the configuration in which the primary site (active system) and the secondary site (standby system) each include an application server and a database server has been described. However, the actual device classification need not necessarily be as in this example. Absent. For example, the application server and the database server may be configured by the same computer, or conversely may be configured by a plurality of physical computers. Furthermore, a plurality of secondary sites may be provided.

  The present invention has been described with reference to the specific embodiments shown in the drawings. However, the present invention is not limited to the embodiments shown in the drawings, and any known hitherto provided that the effects of the present invention are achieved. Even if it is a structure, it is employable.

  The present invention can be applied to a web system having a structure in which an application server and a database server are combined. For example, it is particularly suitable for applications that improve availability in e-commerce systems and business systems.

DESCRIPTION OF SYMBOLS 1 Database system 2 Primary site 3 Secondary site 4 Network 11, 21 Application server 11a, 12a, 21a, 22a Processor 11b, 12b, 21b, 22b Storage means 11c, 12c, 21c, 22c Communication means 12, 22 Database server 13, 23 Storage device 13a, 23a Main storage unit 13b, 23b Backup storage unit 14a, 24a Client device 15a, 25a Peripheral device 101, 201 Application operation unit 102, 112, 202, 212 Server status monitoring unit 102a, 112a, 202a, 212a Command learning Function 102b, 112b, 202b, 212b Fault detection function 102c, 112c, 202c, 212c Error pattern search function 10 d, 112d, 202d, 212d learned command execution function 111, 211 database operation unit 103,113,203,213 learning data 120a target device 120b error code 120c command 120d learning pattern

Claims (7)

A database system configured by connecting a primary site including an active application server and an active database server and a secondary site including a standby application server and a standby database server via a network,
Each of the active application server, the active database server, the standby application server, and the standby database server each includes a server state monitoring unit that monitors each other,
Each of the server status monitoring units
A command input by a maintenance person when a failure occurs in any of the servers constituting the database system is recorded as a log, and a symptom of the failure in which learning data is generated by extracting a specific pattern from the log is indicated. A command learning function for storing in a storage means provided in advance in association with an error code, and sending and storing this learning data to each of the other servers;
A failure detection function for detecting that a failure has occurred in any of the servers;
An error pattern search function for searching whether or not the same error code as the detected fault is stored in the learning data;
A learning command execution function for causing the active application server to execute a command of the pattern corresponding to the error code when the error code that is the same as the detected failure is stored in the learning data; A database system characterized by that. The command learning function is
The database system according to claim 1, wherein the pattern extracted from the log is presented to a user, and the pattern selected by the user is stored as the learning data.   When the failure detection function detects that a failure has occurred in any of the servers, the learning data is sent to the command learning function after detecting that the server in which the failure has occurred is recovered. The database system according to claim 1, wherein: The learned command execution function is
The database system according to claim 1, wherein when the active application server cannot execute the command, the standby application server executes the command. A database system comprising a primary site including an active application server and an active database server and a secondary site including a standby application server and a standby database server connected via a network, wherein the active database server or A database device capable of functioning as the standby database server,
A server status monitoring unit that mutually monitors other devices,
This server status monitoring unit
A command input by a maintenance person when a failure occurs in any of the servers constituting the database system is recorded as a log, and a symptom of the failure in which learning data is generated by extracting a specific pattern from the log is indicated. A command learning function for storing in a storage means provided in advance in association with an error code, and sending and storing this learning data to each of the other servers;
A failure detection function for detecting that a failure has occurred in any of the servers;
An error pattern search function for searching whether or not the same error code as the detected fault is stored in the learning data;
A learning command execution function for causing the active application server to execute a command of the pattern corresponding to the error code when the error code that is the same as the detected failure is stored in the learning data; A database apparatus characterized by that. In a database system configured by connecting a primary site including an active application server and an active database server and a secondary site including a standby application server and a standby database server via a network,
The command learning function of each server constituting the database system records a command input by a maintenance person when a failure occurs in any server constituting the database system as a log,
The learning data obtained by extracting a specific pattern from the recorded log is associated with an error code indicating the symptom of the failure that has occurred and stored in a storage means provided in advance with a command learning function of each server, Send this learning data to each other server for storage,
The failure detection function of each server detects that a failure has occurred in any of the servers,
The error pattern search function of each server searches whether or not the same error code as the detected failure is stored in the learning data,
When the same error code as the detected failure is stored in the learning data, the learned command execution function of each server sends the pattern command corresponding to the error code to the active application server. A failure recovery method characterized by being executed. In a database system configured by connecting a primary site including an active application server and an active database server and a secondary site including a standby application server and a standby database server via a network,
In the processor provided in each server constituting the database system,
A procedure for recording a command entered by a maintenance person as a log when a failure occurs in any of the servers constituting the database system;
The learning data obtained by extracting a specific pattern from the recorded log is stored in the storage means provided in advance in association with the error code indicating the symptom of the failure that has occurred, and the learning is also performed in each of the other servers. Procedure to send and memorize data,
A procedure for detecting that one of the servers has failed;
A procedure for searching whether or not the same error code as the detected fault is stored in the learning data;
When the error code that is the same as the detected failure is stored in the learning data, a procedure for causing the active application server to execute the command of the pattern corresponding to the error code is executed. Disaster recovery program.

Images