JP4911063B2 - Database state monitoring apparatus, method, and program - Google Patents

Description

  The present invention relates to a database state monitoring apparatus, method, and program, and more particularly, to a database state monitoring apparatus, method, and program for monitoring the operating state of a database having a cluster configuration.

  In recent years, in order to improve the availability of a database, an increasing number of databases are used in a cluster configuration (for example, Patent Document 1). In the case of a database having a cluster configuration, when one database fails, processing can be continued by taking over the processing of the failed database by another operating database.

In Patent Document 1, two databases have the same configuration and the same data. One database is accessed at IP address 1 and the other database is accessed at IP address 2. For example, when a failure occurs in the IP address 1 database, IP address 1 is additionally set in the IP address 2 database. According to Patent Document 1, it is possible to perform failover with only an IP address by doing in this way.
JP 2006-338564 A

  When a failure occurs in some databases in a cluster configuration, the processing performed by the failed database must be taken over by another operating database. After detecting this failure and before taking over the processing that was being performed on the failed database, even if a processing request is accepted, the processing cannot actually be executed. Become. In other words, a state that is neither “normal” nor “abnormal” exists. However, in general, the operating state of the database is managed in two types, “normal” and “abnormal”, and a state in which access to the database is awaited (hereinafter, this state is also referred to as “reconfiguring”), Treated as either normal or abnormal.

  The problem described below occurs when “reconstructing” is handled as “normal” or “abnormal”. That is, when the “reconfigured” state is treated as normal, a process that requires high real-time performance is “waited”, resulting in a problem that the responsiveness to the end user is lowered. Further, the subsequent processing request may be accumulated, resulting in a queue overflow. On the other hand, if “Restructuring” is treated as an error, requests cannot be accepted for all business processes, so the process can be terminated abnormally even if it is possible to continue the business, and business continuity is reduced. The problem occurs. In particular, for a process that does not require real-time performance, such as batch processing, there is time to wait until the process is restarted, and thus the process has ended abnormally even though priority is given to process continuity.

  Also in Patent Document 1, processing waits until takeover occurs when setting an IP address and synchronizing a database. In Patent Document 1, even when a part of a database is stopped due to a failure, the data request side continues processing without being aware of failover. Therefore, the occurrence of a database server failure is not detected particularly on the data request side. Therefore, in Patent Document 1, as in the case of handling “being reconfigured” as “normal”, the process is continuously accepted during the failover, and the process is awaited, thereby causing the same problem as described above.

  Conventionally, the operating state of the server is detected with two types of “normal” and “abnormal”, and the “reconfiguring” state cannot be recognized. If it is possible to detect the state of “reconfiguring”, it is possible to perform control to change the response to the processing request in accordance with the nature of the application that issued the request. However, such a control is not possible because the “reconfiguration” state is not detected.

  An object of the present invention is to provide a database state monitoring apparatus, a database system, a method, and a program capable of detecting a state of “reconstructing” in which a database is neither “normal” nor “abnormal”.

In order to achieve the above object, a database state monitoring apparatus of the present invention monitors a database system in which a cluster is constituted by a plurality of database nodes, and controls a process for a database of an application executed on the database system. a condition monitoring apparatus, each of said database nodes, and access to the monitored database, operating state of the database of the database node, indicating the normal operating state "normal", indicating an abnormal operating condition "abnormal" or a monitoring unit for determining which of "during the reconstruction" indicating a state where the process waits occurs database access, the process execution unit queries the operating state of the database based on the processing request from the application When received, supervise the database to be queried. The Get the operating state of the database from the monitoring unit for, and a state information reference unit to be returned to the processing execution section, the processing execution unit receives a processing request from the application, the type of the application It is determined whether it belongs to the first classification or the second classification, and when the operation state of the database to be inquired returned by the state information reference unit is “reconstructing”, the first classification An error process is performed on a process request of an application belonging to the database, and an access process to a processing target database is performed on a process request of an application belonging to the second category .

Database system of the present invention is a database system comprising a database cluster is composed of a plurality of database nodes, each of said database nodes, and access to the monitored database, operating state of the database of each database node but showing a normal operating state "normal", indicating an abnormal operating condition "abnormal", or, a monitoring unit for determining which of "during the reconstruction" indicating the state of pending database access occurs, When kicking process execution unit or we received an inquiry operating state of the processing based on the request database from an application running on the database system, the operating state of the database from the monitoring unit to monitor the database being queried get, and a state information reference unit to be returned to the process execution unit When the processing execution unit receives a processing request from the application, the processing execution unit determines whether the type of the application belongs to the first classification or the second classification, and the inquiry target returned by the state information reference unit When the operating state of the database is “being reconfigured”, error processing is performed for processing requests of applications belonging to the first category, and processing targets are processed for processing requests of applications belonging to the second category The database access processing is performed .

The database status monitoring method of the present invention is a method for controlling access to a database of an application executed on the database system by monitoring a database system in which a cluster is constituted by a plurality of database nodes using a computer. Te, wherein the computer, in each of the database node, to access to the monitored database, operating state of the database of the database node, indicating the normal operating state "normal", indicating an abnormal operating condition "abnormal", or the steps of determining which of "during the reconstruction" indicating a state where the process waits is generated in the database access, the computer, the processing request from the application is generated, the operating state of the processing target database And get the operation Depending on the state, have a, and performing access processing to the error processing or database to the processing request for the application, in the step of performing access processing to the error processing or database, the computer, the processing request It is determined whether the type of the application that has been performed belongs to the first classification or the second classification, and if the acquired database operating state is “reconfiguring”, it belongs to the first classification Error processing is performed on application processing requests, and database access processing is performed on application processing requests belonging to the second category .

The program according to the present invention is a program for causing a computer to monitor a database system including a plurality of database nodes and to execute processing for controlling access to a database of an application executed on the database system. , the computer, in each of the database node, to access to the monitored database, operating state of the database of the database node, "normal" which indicates a normal operating condition, indicating an abnormal operating condition "abnormal", or a process of determining which of "during the reconstruction" indicating a state where the process waits is generated in the database access, the processing request is generated from the application to acquire the operating state of the database to be processed, the Depending on the acquired operating status, A process of performing an access process to the error processing or database for application processing request, is executed, the process for performing access processing to the error processing or database, the computer, the type of application that made the processing request Is determined to belong to the first classification or the second classification, and if the operating state of the acquired database is “reconfiguring”, in response to the processing request of the application belonging to the first classification Error processing is performed, and processing for performing database access processing is executed for processing requests of applications belonging to the second category .

  The database state monitoring apparatus, database system, method, and program of the present invention can detect a state of “reconstructing” in which the database is neither “normal” nor “abnormal”.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 shows a database system including a database state monitoring apparatus according to the present invention. In the present embodiment, the database has a cluster configuration, and will be described as a cluster of n units (n is an integer of 2 or more). The database state monitoring apparatus 10 includes a state information reference unit 12, a state holding unit 13, a monitoring information collection unit 14, and monitoring units 15-1 to 15-n. The database system includes a database state monitoring device 10 and n database nodes 50 (50-1 to 50-n). The functions of the database state monitoring apparatus 10 and the database system can be realized by executing a predetermined program on a computer.

  Each monitoring unit 15 is arranged on each database node 50. The monitoring unit 15 periodically monitors the state of the database on the same node, and transmits the result to the monitoring information collection unit 14. In the monitoring, the alive monitoring of the database process and the response time monitoring of the transaction execution are performed, and “normal”, “abnormal”, and “reconfiguring” are determined from the two results. The monitoring information collection unit 14 collects the state information of each database from the monitoring units 15-1 to 15-n existing in all the database nodes 50-1 to 50-n having a cluster configuration, and the state holding unit 13 is registered.

  The state holding unit 13 centrally manages the information collected by the monitoring information collection unit 14 and holds “normal”, “abnormal”, or “reconstructing” for each database. Also, in response to a request from the state information reference unit 12, a database list and state information are returned. The state information reference unit 12 receives a request from the application program, acquires necessary information from the state holding unit 13, and returns the acquired information to the application program.

  As application programs, an application 51 that performs real-time processing and an application 52 that performs batch processing are considered. The application 51 that performs real-time processing corresponds to a first classification application that requires high real-time processing, and the application 52 that performs batch processing corresponds to a second classification application that does not require much real-time processing.

  The application program acquires the database state from the state information reference unit 12 when accessing the database. The application 51 that performs real-time processing accesses the database and provides a service to the end user only when the acquired database state is “normal”. If “abnormal” or “reconfiguring”, the database is not accessed and terminated abnormally. When the acquired database state is “normal” or “reconstructing”, the application 52 that performs batch processing accesses the database. If “abnormal”, the process is terminated abnormally.

  FIG. 2 shows the configuration of the monitoring unit 15. The monitoring unit 15 includes a state determination unit 21, a process monitoring unit 22, and a DB access state monitoring unit 23. The process monitoring unit 22 monitors a database process group. The DB access state monitoring unit 23 checks the state of the database by accessing the database. The state determination unit 21 determines from the monitoring results of the process monitoring unit 22 and the DB access state monitoring unit 23 whether the operation state of the database is “normal”, “abnormal”, or “reconfiguring”.

  FIG. 3 shows an operation procedure of the monitoring unit 15. The state determination unit 21 inquires of the process monitoring unit 22 about the state of the process in the database (step A1). When an inquiry is received from the state determination unit 21, it is checked whether or not there are processes constituting the monitoring target database. If all the monitoring target processes exist, “normal” is indicated; Return "abnormal". The state determination unit 21 determines whether or not the process monitoring result “abnormal” is returned from the process monitoring unit 22 (step A2). If the process monitoring result “abnormal” is returned, the operation state is set to “abnormal” and the process is terminated (step A3).

  When the process monitoring result “normal” is returned, the state determination unit 21 inquires of the DB access state monitoring unit 23 about the state of access to the database (step A4). FIG. 4 shows a procedure when the DB access state monitoring unit 23 determines the access state. When the DB access status monitoring unit 23 receives an access status query from the status determining unit 21, the DB access status monitoring unit 23 generates a timeout monitoring thread (step B1) and generates a DB access thread (step B2). The timeout monitoring thread is executed in parallel with the DB access thread.

  The DB access state monitoring unit 23 executes the update SQL in the DB access thread (Step B3). The DB access state monitoring unit 23 determines whether or not the result of monitoring the access state to the database is abnormal (step B4). If a response to the executed SQL is not obtained and the process ends abnormally, the access monitoring result “abnormal” is set (step B5). When a response is obtained, the access monitoring result “normal” is set (step B6).

  The timeout monitoring thread notifies the main thread (DB access thread) of the occurrence of timeout after the elapse of a predetermined time, for example, one second after execution of the update SQL (step B7). The DB access state monitoring unit 23 determines whether or not the access monitoring result in the DB access thread is “abnormal” (step B8). If the access monitoring result is “abnormal”, the access state monitoring result “abnormal” is returned to the state determination unit 21 and the process is terminated (step B9).

  If the access monitoring result is “normal”, the DB access state monitoring unit 23 determines whether a response to the update SQL executed in the DB access thread is obtained before the timeout occurs in the timeout monitoring thread. Is determined (step B10). When a response to the update SQL is obtained after the timeout occurs, the access state monitoring result “reconfiguring (timeout)” is returned to the state determination unit 21 (step B11). When a response is obtained before the timeout occurs, the access state monitoring result “normal” is returned to the state determination unit 21 (step B12).

  Returning to FIG. 3, the state determination unit 21 determines whether or not the access state monitoring result “abnormal” is returned from the DB access state monitoring unit 23 (step A5). If the access status monitoring result “abnormal” is returned, the operation status of the database is set to “abnormal” and the process is terminated (step A6). Next, the state determination unit 21 determines whether or not the access state monitoring result “reconfiguring (timeout)” has been returned (step A7). If the access status monitoring result “reconfiguring (timeout)” is returned, the operation status of the database is set to “reconfiguring” and the process is terminated (step A8). If the access status monitoring result “normal” is returned, the operation status of the database is set to “normal” and the process is terminated (step S9).

  FIG. 5 shows a summary of the relationship between the response from the process monitoring unit 22, the response from the DB access state monitoring unit 23, and the operation state. When the process monitoring result “abnormal” is returned from the process monitoring unit 22, the state determination unit 21 sets the operation state of the database to “abnormal”. When the process monitoring result is “normal” and the access status monitoring result “normal” is returned from the DB access status monitoring unit 23, the status determining unit 21 sets the operation status to “normal”. When the process monitoring result is “normal” and the access status monitoring result is “abnormal”, the operation status is “abnormal”, the process monitoring result is “normal”, and the access status monitoring result is “reconfigure”. When “Medium (timeout)”, the operation state is “Reconfiguring”.

  Each monitoring unit 15 periodically checks the operating state of the corresponding database according to the procedure described above, and sends the determination results of “normal”, “abnormal”, and “under reconstruction” to the monitoring information collection unit 14. Send. The monitoring information collection unit 14 collects the database operating state transmitted by the monitoring units 15-1 to 15-n arranged in the database nodes 50-1 to 50-n having a cluster configuration, and the state holding unit 13 holds Update the status of each database that is running.

  FIG. 6 illustrates data held by the state holding unit 13. The state holding unit 13 holds “normal”, “abnormal”, or “reconstructing” for each database node. When receiving a state reference request from the application, the state information reference unit 12 refers to the state holding unit 13 and returns the specified database state to the application.

  The application 51 that performs real-time processing accesses the database and performs business processing only when the response from the state information reference unit 12 is “normal”. If "abnormal" or "reconfiguring", the database is not accessed and error processing is performed. By handling the error, unnecessary waits for responses are avoided and processing requests from end users are kept without stagnation. On the other hand, the application 52 that performs batch processing with low real-time performance performs error processing only when the response from the state information reference unit 12 is “abnormal”. If “normal” or “reconfiguring”, the database is accessed to continue the business process. In this way, unnecessary abnormal termination can be avoided and business continuity can be improved.

  In the present embodiment, the operating state of the database is managed by “reconstructing” in addition to “normal” and “abnormal”. The status of “Reconfiguring” can be detected by issuing a request to the database, whether or not a response to the request is returned, and whether the response takes a long or short time. is there. In the present embodiment, although a response from the database is returned, it is possible to recognize the “reconstructing” state by detecting a state in which it takes a long time to return the response.

  By detecting “reconfiguring”, the following database system can be controlled. In other words, when "reconfiguring" is in progress, processing requests from processing that requires real-time processing are not accepted, increasing the responsiveness of real-time processing and preventing queue overflow due to unexpected processing waits. can do. In the error processing for the real-time processing at the time of “reconfiguring”, for example, a message “Please wait for a while” is displayed to the end user. In this case, it is possible for the end user to suppress useless request issuance by refraining from request issuance.

  Also, for a process where waiting for several tens of seconds to several minutes does not become a problem as in batch processing, an unexpected abnormal termination can be suppressed by receiving a processing request. In this way, when the operating state is “reconfiguring”, depending on the nature of the process, select whether to accept the process request or not, so that the process that requires real-time processing can be quickly handled and the queue overflowed. A process that can be prevented and that does not require real-time performance can be continued as much as possible. The present embodiment can be used for all systems having a database cluster configuration. This is particularly effective for a system with a large number of processes per second.

  As mentioned above, although this invention was demonstrated based on the preferable embodiment, the database state monitoring apparatus of this invention, a method, and a program are not limited only to the said embodiment, From the structure of the said embodiment. Various modifications and changes are also included in the scope of the present invention.

1 is a block diagram showing a database system including a database state monitoring apparatus according to an embodiment of the present invention. The block diagram which shows the structure of a monitoring part. The flowchart which shows the operation | movement procedure of a monitoring part. The flowchart which shows the operation | movement procedure of an access state monitoring part. The determination table which shows the relationship between the response from a process monitoring part, the response from a DB access status monitoring part, and an operation state. The figure which shows the example of data of a state holding part.

Explanation of symbols

10: database status monitoring device 12: status information reference unit 13: status holding unit 14: monitoring information collection unit 15: monitoring unit 21: status determination unit 22: process monitoring unit 23: DB access status monitoring unit 50: database node 51: Real-time processing 52: Batch processing

Claims (14)

A database state monitoring device that monitors a database system in which a cluster is configured by a plurality of database nodes, and controls processing of a database of an application executed on the database system ,
In each of the database node, to access to the monitored database, operating state of the database of the database node, indicating the normal operating state "normal", indicating an abnormal operating condition "abnormal", or the process for database access shows a state where the waiting occurs and monitoring unit for determining which of "during reconstruction",
A state in which when an inquiry about the operating state of a database based on a processing request from the application is received from a processing execution unit, the operating state of the database is acquired from the monitoring unit that monitors the database to be queried and returned to the processing execution unit An information reference part;
Equipped with a,
When the processing execution unit receives a processing request from the application, the processing execution unit determines whether the type of the application belongs to the first classification or the second classification, and the inquiry target returned by the state information reference unit When the operating state of the database is “being reconfigured”, error processing is performed for processing requests of applications belonging to the first category, and processing targets are processed for processing requests of applications belonging to the second category Process to access the database of
Database state monitor you wherein a. Monitoring information collection that acquires the operating state of the database to be monitored from the monitoring unit of each database node, and updates the operating state with the acquired state holding unit that associates and maintains the operating state of the database of the database node for each database node Further comprising
When the state information reference unit receives an inquiry about the operation state of the database from the process execution unit, the state information reference unit refers to the state holding unit, acquires the operation state of the database to be processed, and returns the operation state to the process execution unit.
The database state monitoring apparatus according to claim 1, wherein: The database state monitoring according to claim 1 or 2, wherein the application belonging to the first category is a real-time application requiring real-time property, and the application belonging to the second category is an application not requiring real-time property. apparatus. The monitoring unit, and a database access state monitoring unit that monitors the health of the database to access the database, based on the monitoring results of the database access state monitoring unit, operating state of the database is "normal", "abnormal" , and a state determining section for determining which of "during reconstruction", the database status monitoring apparatus according to any one of claims 1 to 3. Wherein the database access state monitoring unit, based on whether responses to database access can be obtained within a predetermined timeout occurrence time, access status monitoring result "normal", "abnormal", the "time-out occurred" The database state monitoring apparatus according to claim 4 , which is returned to the state determination unit. The database access state monitoring unit returns an access state monitoring result “abnormal” when the response is not obtained, and returns an access state monitoring result “normal” when the response is obtained within the timeout occurrence time. 6. The database state monitoring apparatus according to claim 5 , wherein the database state monitoring device returns the access state monitoring result “timeout occurrence” when the response is obtained after the time-out occurrence time elapses. The monitoring unit further includes a process monitoring unit that monitors a process group of a database, and the state determination unit, when the process monitoring unit determines that there is no process constituting the database, When it is determined that the operating state is “abnormal” and the process monitoring unit determines that a process exists, the operating state of the database is “normal” based on the monitoring result of the database access state monitoring unit, "abnormal", to determine which of "during reconstruction", the database status monitoring apparatus according to any one of claims 4 to 6. A database system including a database in which a cluster is constituted by a plurality of database nodes,
In each of the database node, to access to the monitored database, operating state of the database of the database node, indicating the normal operating state "normal", indicating an abnormal operating condition "abnormal", or the process for database access shows a state where the waiting occurs and monitoring unit for determining which of "during reconstruction",
When kicking process execution unit or we received an inquiry operating state of the processing based on the request database from an application running on the database system, the operating state of the database from the monitoring unit to monitor the database being queried A status information reference unit to obtain and return to the process execution unit ;
With
When the processing execution unit receives a processing request from the application, the processing execution unit determines whether the type of the application belongs to the first classification or the second classification, and the inquiry target returned by the state information reference unit When the operating state of the database is “being reconfigured”, error processing is performed for processing requests of applications belonging to the first category, and processing targets are processed for processing requests of applications belonging to the second category Process to access the database of
A database system characterized by that . Monitoring information collection that acquires the operating state of the database to be monitored from the monitoring unit of each database node, and updates the operating state with the acquired state holding unit that associates and maintains the operating state of the database of the database node for each database node Further comprising
When the state information reference unit receives an inquiry about the operation state of the database from the process execution unit, the state information reference unit refers to the state holding unit, acquires the operation state of the database to be processed, and returns the operation state to the process execution unit.
The database system according to claim 8. The database system according to claim 8 or 9 , wherein the application belonging to the first category is a real-time application that requires real-time property, and the application belonging to the second category is an application that does not require real-time property. A method of monitoring a database system in which a cluster is configured by a plurality of database nodes using a computer, and controlling access to a database of an application executed on the database system ,
The computer is, in each of the database node, to access to the monitored database, operating state of the database of the database node, "normal" which indicates a normal operating condition, indicating an abnormal operating condition "abnormal", or, a step of determining which of showing a state where the process waits is generated in database access "during reconstruction",
When the computer generates a processing request from the application , the computer acquires the operating state of the processing target database, and performs error processing or database access processing on the processing request of the application according to the acquired operating state. and the step of performing,
I have a,
In the step of performing the error processing or database access processing, the computer determines whether the type of the application that requested the processing belongs to the first classification or the second classification, and the acquired database When the operation state is “reconfiguring”, error processing is performed for processing requests of applications belonging to the first category, and access to the database is performed for processing requests of applications belonging to the second category Process,
The database state monitoring method characterized by the above-mentioned. The computer further includes the step of acquiring the operating state of each database, and updating the operating state with the acquired state holding unit that holds the database operating state of the database node in association with each database node;
In the step of performing error processing or database access processing, when a processing request is generated from the application, the computer refers to the state holding unit and acquires the operating state of the database to be processed.
The database state monitoring method according to claim 11, wherein: A program that causes a computer to monitor a database system including a plurality of database nodes and to execute processing for controlling access to a database of an application executed on the database system ,
In the computer,
In each of the database node, to access to the monitored database, operating state of the database of the database node, indicating the normal operating state "normal", indicating an abnormal operating condition "abnormal", or the process for database access a process of determining which of indicating the state of waiting occurs "during reconstruction",
When a processing request from the application is generated, an operation state of a database to be processed is acquired, and an error process or a database access process is performed on the processing request of the application according to the acquired operation state. ,
And execute
In the error processing or processing for accessing the database, the computer determines whether the type of the application that requested the processing belongs to the first category or the second category, and the obtained database operation When the status is “reconfiguring”, error processing is performed on the processing request of the application belonging to the first category, and access processing to the database is performed on the processing request of the application belonging to the second category. Process to do
A program characterized by being executed. The computer further acquires the operating state of each database, and further executes a process of updating the operating state for acquiring a state holding unit that associates and holds the operating state of the database of the database node for each database node,
In the process of performing the error process or the process of accessing the database, when a process request from the application occurs, the operating state of the database to be processed is acquired with reference to the state holding unit.
The program according to claim 13.

Images