JP2009217768A - Method and system for managing partition failure in database - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To permit or prohibit access per partition by managing failure information for each divided partition with respect to a database divided into a plurality of partitions by partitioning. <P>SOLUTION: The method for controlling access to a storage area divided into the plurality of partitions includes: a state detection step of accessing each partition to store a state of each partition in a state holding part; a state confirmation step of confirming, when a data update request to the storage area is received, a state of a partition concerning the request by referring to the state holding part; and a data update step of executing processing concerning the request only when the partition is in a normal state. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a partition failure management method and system in a database. In particular, the present invention relates to a partition failure management method and system in a database divided into a plurality of partitions by partitioning.

  In order to improve database availability, manageability, and query response, a data partitioning method called partitioning that partitions data of one database table into partitions that are sets of a plurality of logical data is known. Multiple partitions divided by this partitioning can be assigned to different disks or logical volumes, and even when a specific partition is damaged, access to other normal partitions and data update continue. Can be done.

  However, even if the table is divided into multiple partitions by partitioning, this table is only seen as one table by the client application, and the client application recognizes what kind of partition exists. Can not do it. For this reason, the client application cannot recognize the partition status individually, and even if a failure occurs in some partitions, it can only recognize whether the entire table is normal or whether the entire table is in error. . The advantage of partitioning is that even if a failure occurs in a specific partition, as long as another normal partition exists, the processing can be continued for the other partition. Therefore, in order to take advantage of this advantage, the client application can ensure that the entire table is healthy in order to allow access to other healthy partitions even if a particular partition fails. It is necessary to continue processing as if there is. However, if the entire table is normal even though a failure has occurred in a specific partition, multiple accesses by multiple transactions may occur in the partition in which the failure has occurred. In this case, there is a problem that error messages are repeatedly generated and other important messages are buried.

In order to cope with such a problem, a file blocking control method that prohibits access to a block area including a failed data is disclosed (for example, Patent Document 1). In this file block control method, the address of the data where the failure has occurred is stored in the memory, and when there is a request for access to the data, it is confirmed whether the address of the data requested to be accessed is stored in the memory. If this address is stored, the access request to the address is rejected.
JP-A-60-258659

  However, this method only prohibits access to a faulty data address, and cannot permit or prohibit access on a partition basis. In addition, since this method does not monitor the failure information of the data area, the client application cannot recognize which memory access is prohibited, and that the failure of the data area has been resolved. I can't recognize it.

  Accordingly, an object of the present invention is to permit or prohibit access in units of partitions by managing failure information for each divided partition in a database divided into a plurality of partitions by partitioning.

  The present invention relates to a method for controlling access to a storage area divided into a plurality of partitions, a state detection step of accessing each partition and storing whether or not it is normal in a state holding unit, and data to the storage area Data for executing a status check step for checking the status of the partition related to the request with reference to the status holding unit when the update request is received, and processing for the request only when the status of the partition is normal And an update step.

  Further, the state detection step is periodically executed, and the latest detection state is stored in the state holding unit.

  Further, the state confirmation step refers to a correspondence table indicating in which partition the data stored in the storage area exists, derives a partition corresponding to the data related to the request, and confirms the state of the partition It is characterized by doing.

  The present invention is a method of controlling access to a storage area divided into a plurality of partitions, and stores each partition in a state holding unit to determine whether it is normal or not, and receives a data update request to the storage area In this case, a partition corresponding to the data related to the request is derived with reference to a correspondence table indicating in which partition the data stored in the storage area exists, the state of the partition is confirmed, and the partition Since the processing related to the request is executed only when the state is normal, access can be permitted or prohibited in units of partitions.

  Further, according to the present invention, since the state detection is periodically performed and the latest detection state is stored in the state holding unit, the state of the storage area can be confirmed almost in real time, Respond quickly to recovery.

  The best mode for carrying out the present invention will be described below in detail with reference to the accompanying drawings.

  FIG. 1 is a schematic diagram showing the overall configuration of a database failure management system according to the present invention. In this figure, a partition state monitoring apparatus 10, a client application 20, and a database node 30 are shown. This system is, for example, a data server connected to a network, and the client application 20 receives a data processing request from a client PC and performs processing on the disk 33 of the database node 30. The partition status monitoring device 10 holds a monitoring information registration unit 11 that collects monitoring information from the database node 30, an information holding unit 12 that holds information collected by the monitoring information registration unit 11, and a predetermined partition information. The partition definition holding unit 13 and the status information reference unit 14 for the client application 20 to refer to the partition status are provided.

  FIG. 2 is a diagram showing the monitoring unit 31 of the database node 30 in more detail. The database node 30 includes a monitoring unit 31 and a database 32, and the database 32 includes a failure monitoring table 321 and a business table 322. The monitoring unit 31 refers to the failure monitoring table 321 of the database 32 to determine whether a partition failure has occurred, and the determination result of the determination unit 312 is used as the monitoring information registration unit of the partition state monitoring device 10. 11 and a communication means 311 for transmitting to the terminal 11.

  FIG. 3 is a diagram illustrating the relationship between the failure monitoring table 321 and the business table 322 included in the database 32 of the database node 30 and the disk 33. As shown in this figure, since five disks are used in this embodiment, the failure monitoring table 321 is divided into five partitions 3211 to 3215 so as to correspond to these disks. In this embodiment, the partitions are allocated in units of physical disks, but may be allocated in units of logical volumes instead. The business table (A) 322a is assigned to the four disks 331 to 334, and the business table (B) 322b is assigned to the two disks 334 and 335.

  FIG. 4 shows a definition example of the failure monitoring table 321, and FIG. 5 shows an initial state of the failure monitoring table 321 according to this definition example. In this definition example, an ID column is set as a primary key so that IDs do not overlap. Further, the records of ID 1 to 5 are defined by the list partition so as to uniquely correspond to the partitions 3211 to 3215 of the failure monitoring table 321. These partitions correspond to the disks 331 to 335, respectively. As a result, it is possible to confirm the presence or absence of a failure in units of partitions by issuing a SQL UPDATA statement or the like to each record of ID1 to ID5 and confirming the result.

  Next, the processing operation of the present invention will be described with reference to FIG. First, the determination unit 312 of the monitoring unit 31 in the database node 30 issues an SQL SELECT statement to the failure monitoring table 321 (A1), and a list of IDs (ID1 to ID5) stored in the failure monitoring table 321. Is acquired (A2). These IDs correspond to the partitions and disks in the failure monitoring table 321. Next, the determination unit 312 of the monitoring unit 31 issues an SQL UPDATA statement to a record corresponding to each ID (A3) and receives an update result (A4) in order to confirm whether there is a partition failure. Here, if the SQL statement ends normally, it is determined that the corresponding partition is normal, and if the SQL statement ends abnormally, it is determined that the corresponding partition is abnormal. Next, the communication unit 311 of the monitoring unit 31 transmits the update result to the monitoring information registration unit 11 of the partition state monitoring apparatus 10 (A5), and the monitoring information registration unit 11 transmits the update result to the state holding unit 12. (A6). FIG. 7 is a diagram illustrating an example of the state holding unit 12. The state holding unit 12 holds these update results for each ID and disk. The monitoring unit 31 of the database node 30 repeats these processes regularly, for example, every 1 second. Thereby, the state of a partition is regularly updated, and the state of each partition can be confirmed almost in real time.

  Next, the partition definition holding unit 13 will be described with reference to FIGS. As shown in these drawings, the partition definition holding unit 13 corresponds to the table / partition key correspondence table 131, the key value / disk name correspondence table 132a for the business table A, and the key value / disk name correspondence for the business table B. And a table 132b. In the table / partition key correspondence table 131, the partition keys of the business tables A and B are registered as shown in FIG. In the key value / disk name correspondence table 132a for the business table A, as shown in FIG. 9, the range of the partition key “date” and the disk name related thereto are registered. In the key value / disk name correspondence table 132b for the business table B, as shown in FIG. 10, the range of the partition key “ID” and the disk name related thereto are registered.

  Next, a process in which the client application 20 updates the business table will be described with reference to FIG. In the following description, the business table to be updated is “business table A”, and the key value is “2007.1.1”.

  First, when the client application 20 receives a business table update request from a client PC (not shown) (B1), the status information reference unit 14 of the partition status monitoring device 10 receives the table name “ "Business table A" and information to be updated are transmitted (B2). Next, the status information reference unit 14 refers to the table / partition key correspondence table 131 of the partition definition holding unit 13 and acquires the partition key “date” related to the received table name “business table A” ( B3). Next, based on the acquired partition key “date” and the value “2007.1.1” of the state, the status information reference unit 14 uses the key value / disk name correspondence table for the business table A of the partition holding unit 13. Referring to 132a, the corresponding disk name “disk 333” is acquired (B4). The status information reference unit 14 refers to the status holding unit 12 shown in FIG. 7 to acquire the status information of “disk 333” (B5), and transmits this status information to the client application 20 (B6). The client application 20 performs the database update process only when the state information is normal (B7). As described above, since the client application 20 checks the state of the partition before performing the database update process, unnecessary update of the database can be avoided.

  Here, only the process when there is one update target table has been described, but when there are a plurality of update target tables, these processes are repeated. Similarly in this case, the update process is performed only when the partition to be updated is normal.

  While the preferred embodiment of the invention has been illustrated and described, various modifications can be made without departing from the scope of the invention.

  The present invention can be used for all systems using database partitioning. In particular, it is useful for an online system having a large-scale database with a large number of transaction processes.

FIG. 1 is a schematic diagram showing the overall configuration of a database failure management system according to the present invention. FIG. 2 is a detailed diagram of the monitoring unit 31 of the database node 30. FIG. 3 is a diagram showing the failure monitoring table 321 and business table 322 of the database 32 and the disk 33. FIG. 4 is a diagram illustrating a definition example of the failure monitoring table 321. FIG. 5 shows an initial state of the failure monitoring table 321 according to the definition example of FIG. FIG. 6 is a diagram showing the processing operation of the present invention. FIG. 7 is a diagram illustrating the state holding unit 12 of the partition state monitoring apparatus 10. FIG. 8 is a diagram showing a table / partition key correspondence table 131 for managing the partition keys of the business tables A and B. FIG. 9 is a diagram showing a key value / disk name correspondence table for the business table A that manages the range of the partition key value “date” and the disk names related thereto. FIG. 10 is a diagram showing a key value / disk name correspondence table for the business table B that manages the range of the partition key value “ID” and the disk names related thereto. FIG. 11 is a diagram illustrating processing in which the client application 20 updates the business table.

Explanation of symbols

10 partition status monitoring device 20 client application 30 database node

Claims (6)

A method of controlling access to a storage area divided into a plurality of partitions,
A state detection step for accessing each partition to store whether it is normal or not in the state holding unit, and when receiving a data update request to the storage area, the state of the partition related to the request with reference to the state holding unit An access control method, comprising: a state confirmation step for confirming the data; and a data update step for executing the processing related to the request only when the state of the partition is normal.   The access control method according to claim 1, wherein the state detection step is periodically executed, and a latest detection state is stored in the state holding unit.   3. The method according to claim 1, wherein the status check step refers to a correspondence table indicating in which partition the data stored in the storage area exists, and selects a partition corresponding to the data related to the request. An access control method characterized by deriving and confirming the state of the partition. In a storage device comprising a storage area divided into a plurality of partitions,
A monitoring unit that accesses each partition and stores whether it is normal or not in the state holding unit, and a state information reference unit that checks the state of the partition related to the request by referring to the state holding unit when a data update request is received And
An apparatus characterized by executing the process related to the request only when the state of the partition is normal.   5. The apparatus according to claim 4, wherein the monitoring unit periodically accesses each partition to update the state holding unit.   6. The apparatus according to claim 4, further comprising a correspondence table indicating in which partition the data stored in the storage area exists, and the state information reference means refers to the correspondence table. And identifying a partition corresponding to the data related to the update request and checking the state.

Images