JPS62216055A - Processing system for data base recovery - Google Patents

Abstract

PURPOSE:To attain recovery of a data base without using a recovery function by detecting a faulty area of the data base and recovering the fault partially and for each logic. CONSTITUTION:A hardware trouble if occurs while an application processing part 11 is accessing to a data file 24 is detected by a fault detecting part 14. Then a data base recovering part 15 starts an injured logical unit detecting part 16 and the part 16 accesses successively to the logical units at the head of the file 24. An injured block detecting part 17 allocates a substitute block to an injured logical unit block detected by the part 16. A record control information recovering part 19 outputs the keys of the records deleted out of an erasion record information output part to an output device 22 by defining the record related to the injured logical unit block as a deletion record. Based on said record key, a user makes use of an index file reforming part 21 to apply a record again. Thus the fault data is restored.

Description

Detailed Description of the Invention [Summary] In a data processing system that has the function of managing a database consisting of index files and data files such as VSAM, a part of the data in the database is damaged due to a hardware failure. By directly repairing the damaged part of the database without using the database backup recovery function.

Data other than the damaged part can be used in the latest state.

[Industrial application field]

The present invention provides a database repair processing method that provides a simple repair function for a database whose data has been damaged due to a hardware failure in a data processing system that has a function of managing a database consisting of an index file and a data file such as VSAM. It is related to.

[Conventional technology]

Manage the database with a data processing device.

The so-called VSAM is used as a method for providing an access method.
(Virtual Storage Access M
method) is used. This type of database has an index file and a data file that stores the data pointed to by the index file. Records within the data file are managed in logical units called control intervals (CI). things are being done.

By the way, if some of the data in the database is damaged due to a hardware failure.

Generally, the contents of a database are recovered using bank-up information whose data has been saved in advance and update log information.

However, it is important to take frequent backups.

It takes a lot of time and effort, and it takes a lot of time and effort to obtain update log information sequentially during online service.

In order to increase the system load, 1. There are users who use databases. In such cases.

Currently, if some data in a database is damaged due to a hardware failure, the entire database must be discarded.

[Problem that the invention seeks to solve]

In the case of files with relatively simple organization, such as so-called direct organization files, even if part of the data is damaged, it is easy to restore the other parts so that they can be used as is. However, in the case of a database consisting of index files and data files that include alternative indexes such as VSAM, if some of the data is damaged, inconsistencies with various data management information will occur. Data cannot be used.

The present invention aims to solve the above problems without using the conventional recovery function using backup or the like.

The purpose is to provide a database repair processing method for repairing a database.

[Means for solving problems]

FIG. 1 shows an example of the basic configuration of the present invention.

In FIG. 1, 10 is a processing unit consisting of a CPU and memory, 11 is an application processing unit that executes a program for processing data, 12 is a database management unit that provides database management functions, and 13 is for accessing the database. a database access section; 14 is a database access section;
A failure detection unit 15 detects a hardware failure caused by access; a database repair unit 15 provides a database recovery function according to the present invention; and 16, a database repair unit 16 accesses the failure database in each predetermined logical unit, which is a unit of logical data input/output. a damage logical unit detection unit that detects damage locations;
Reference numeral 17 denotes a damaged block detection unit that accesses the detected damaged logical unit in units of physical blocks to detect damaged blocks. Reference numeral 18 refers to a replacement area allocation unit that allocates replacement blocks or tracks to the detected damaged blocks.

19 is a record management information repair unit that repairs record management information in a damaged logical unit; 20 is a unit that checks consistency between an index file and a data file, and outputs key values or relative record number information, etc. regarding records lost due to failure; 21 is an index file regeneration unit that recreates the index file based on the restored data file; 22 is an output device such as a printer or external storage device; and 23 is an index file 124 such as an alternative index. represents a data file having data records in each logical unit of a predetermined length.

The application processing unit 11 accesses the data file 24 using the macro provided by the database access unit 13,
If a hardware failure occurs at that time, the failure detection unit 1
4, the fault is detected. The database repair unit 15 is activated by the failure detection unit 14 or by input of a repair command from an operator.

The database repair unit 15 activates the damaged logical unit detection unit 16. The damaged logical unit detection section 16 is.

With the macro provided by the database access unit 13,
The data file 24 is accessed sequentially from the first logical unit. For that access, if a failure is detected and the injured logical unit is found, the injured block detector 17
call.

The damaged block detection unit 17 detects a damaged physical block by directly accessing the physical block in the logical unit using a so-called EXCP tuna or the like. The replacement area allocation unit 18 allocates a replacement block or a replacement track to the detected physical block.

Secure an area to replace the damaged physical block.

The record management information restoration unit 19 guarantees consistency between control information and record information in a logical unit by converting the records related to the damaged physical block into two so-called deletion records.

The lost record information output unit 20 determines whether or not there is a record pointed to by the index file 23 based on the 9-dimensional index file 23.

A process of checking consistency and outputting a list of key values, relative record number information, etc. regarding records lost due to a failure to an output device 22 such as a printer is executed.

The index file re-creation unit 21 re-creates the index file 23 from the data file 24 from which records related to the failure have been deleted. To recreate index file 23.

Commands provided by the conventional database management unit 12 can be used.

[Effect]

According to the present invention, for the physical block detected by the damaged block detection section 17, the replacement area allocation section 18 allocates two alternative areas, and the record management information restoration section 19 allocates two alternative areas.

Since control information is written to invalidate the record related to the damaged physical block, data other than the damaged portion is saved in the latest state.

Since the lost record information output unit 20 reports information such as key values regarding the lost record to the user, the user can later restore the failed data by re-entering the data of that record. is also possible.

Since the index file regeneration unit 21 guarantees the consistency between the index file 23 and the data file 24, data other than the damaged portion that is saved in the latest state can be accessed normally.

In particular, since there is no need to restore from backup information or update log information, partial repairs can be performed directly while the system is running, and there is no need to stop the system and perform recovery as in the past.

The effects of failure can be kept to a minimum.

〔Example〕

Fig. 2 is an explanatory diagram of the control interval format, Fig. 3 is an explanatory diagram of damage CI detection, and Fig. 4 is an explanatory diagram of the control interval format.
The figure is an explanatory diagram of repair processing in cr units.

FIG. 5 shows a processing explanatory diagram of an embodiment of the present invention.

In the following description of the embodiment, an example will be described in which it is applied to a database processing system using VSAM.

The invention is not limited to VSAM, but can be similarly applied to database processing systems that manage the same type of index files and data files.

In VSAM, as a unit that controls data transfer between input/output cover sofas and VSAM data sets.

A control interval (CI) as shown in FIG. 2 is used. A control interval is a logical unit regarding a data storage area to increase independence from a physical device, and has a record storage area, a control information area of a record definition field (RDFn), and a control interval definition field (CIOF).

3 by accessing the data file 24 shown in FIG.
When a hardware failure is detected, a damage control interval is detected, as shown in FIG.

That is, reading is performed in control interval units sequentially from the beginning of the data section, and when an I10 error or format error occurs, it is determined that it is a damage control interval, and repair processing is started in control interval units. Note that the format error is related to a state in which a replacement block has been allocated, and in this case, a CI format/record format repair process is performed.

Repair processing in control interval units.

As shown in FIG. 4, a damaged physical block is first detected by accessing each physical block sequentially from the beginning of the damage control interval. If a damaged physical block is found, a replacement block (or replacement track) is assigned to that block. After allocating replacement blocks, etc. to damaged physical blocks, a CI format/record format repair process is performed.

Restoring the cr format/record format is a process that eliminates the inconsistency between the data record section and the control information section shown in Figure 2.
Repair methods vary depending on the location of the damage. CIDF, RDF
If the physical block containing n is damaged, all data within the control interval is invalidated. If a data record's physical block is damaged, invalidate the record. That is, the record is treated as a deleted record, for example.

If a physical block with only free space is damaged, there is no need to repair the CI format/record format. Storage attribute is FRE
In the case of E, after the CI format/record format is repaired, all management chains of deleted records in the database are reallocated.

Through the above processing, the data file 24 shown in FIG. 1 has been formally restored, but this alone causes a contradiction with the index file 23. That is, although the index file 23 has information such as keys and positions in the data file 24, some of the data files 24 are missing records due to repair processing of damaged locations.

In order to resolve this contradiction, the index file 23 is regenerated by using the ALTER function of the so-called CRTDB command provided by the VSAM management unit. This is a process that collects the contents to be stored in the index file 23 based on the data file 24 and recreates the index file 23.

FIG. 5 shows three or more processes in the form of a flowchart. Processing numbers ■ to ■ in the following explanation correspond to numbers ■ to 0 shown in FIG.

■ A hardware failure occurred in the data file 24.
For example, when a repair process is started by specifying the database name, system name, etc. by an operator command, the information necessary for repair such as the key length/record length, storage attributes, etc. of that database is first obtained from the catalog. .

■ The entire control interval (C
Determine whether access to I) has ended. When finished, control is transferred to process [phase].

■ VS in CI units sequentially from the beginning of the data file 24
Accessed by AM macro.

■ Determine whether an error occurred in accessing the CI. If it is normal, control is returned to process (2).

(2) Determine whether access to all physical blocks in the secondary process (2) has been completed for the CI in which the error occurred. When the process is completed, control is transferred to process ■.

■ Execute EX on all physical blocks in the damaged CI in order from the beginning.
Accessed by CP macro.

■ Determine whether an error occurred. If no error occurs, return to process (2) and repeat the process for the tertiary physical block.

■ If an error occurs, allocate a replacement block or track and return to process ■.

■ Execute the CI format/record format repair process.

Thereafter, the process returns to process (2) and the same process is repeated for the secondary CI.

[Phase] Once the restoration of the data file 24 is completed.

According to the index file 23, consistency between the index file 23 and the data file 24 is checked.

■ There are key values, etc. in the index file 23.

If the data file 24 does not have a corresponding record (a record that existed in the damaged part), it is extracted as a lost record.

@ Outputs a list of lost record information to a specified device such as a printer, display, or external storage device.

■ Index file 23 with CRTDB command A
For example, by using the LTER function.

Re-create.

Records lost due to hardware failure.

Since the user is notified by the list output of the above process @, the user can then completely restore the database by re-entering the data of that record. For example, if it is an employee management database, as a key value.

A list of employee numbers or names will be output.

Based on the key value, the employee's data including department, age, address, and other information can be re-entered.

〔Effect of the invention〕

As described above, according to the present invention, a database can be easily repaired without using a conventional recovery function that uses backup and update logs. Since there is no need to acquire logs for recovery, DASD resources during service, etc.
CPU resources can be used effectively. Directly about the damaged area.

Partial repair is performed without stopping the system.
Repair operations can be performed.

[Brief explanation of drawings]

Figure 1 is an example of the basic configuration of the present invention, Figure 2 is an explanatory diagram of the control interval format, Figure 3 is a diagram to explain detection of damaged CIs, and Figure 4 is an explanation of repair processing in units of CIs. FIG. 5 shows a process explanatory diagram of an embodiment of the present invention. In the figure, 10 is a processing device, 11 is an application processing section, 12 is a database management section, 13 is a database access section, 14
15 is a failure detection unit, 15 is a database repair unit, 16 is a damaged logical unit detection unit, 17 is a damaged block detection unit, 18 is a replacement area allocation unit, 19 is a record management information recovery unit, 20
2 is a lost record information output unit, 21 is an index file regeneration unit, 22 is an output device, and 23 is an index file. 24 represents a data file. Patent Applicant Fujitsu Ltd. Representative Patent Attorney Hiroshi Mori 1) (1 other person) 1 Figure 2 Figure 2 Damaged work & photo 3 Country Figure 5

Claims (1)

[Claims] The index file (23) and the index file (
data file (24) pointed to by (23)
Regarding the data file (24) in which a failure occurred in a data processing system that has the function of managing a database consisting of:
means (16) for sequentially accessing each predetermined logical unit, which is a logical data input/output unit, from the beginning to detect a damaged location; and means (16) for accessing the detected damaged logical unit in physical block units. , means (17) for detecting a damaged block; means (18) for allocating a replacement area for the detected damaged block; means (19) for restoring record management information in the logical unit; and the index file (23). ) and the data file (24) repaired by the above-mentioned means (20), and means (20) for outputting key value or relative record number information regarding the record lost due to a failure; and the above-mentioned index file (23). means (21) for re-creating the data file (24) after restoration.
) A database repair processing method characterized by comprising: