JPS60142445A - Recovery processing system of resident data file - Google Patents

Abstract

PURPOSE:To reduce the number of times of I/O to/from a direct access device by developing a data file or the whole dump information of the direct access device on a memory at the generation of a fault, recovering the data in accordance with its updating history and writing the recovered result in the direct access device. CONSTITUTION:If a fault is generated in the data file 2, a resident discrimination part 17 in a recovery processing part 16 checks its property while referring a data file definition body 12, and when there is the resident property, a memory development part 18 loads original data to a memory 13' and a record recovery part 19 reads out log information which is the record of the image updated in the order of log generation from a log file 5 to recover the contents of the data file is each record on the memory 13'. Then, a data file storage part 20 collectively writes the contents of the data file on the memory 13' in the direct access device 1'.

Description

DETAILED DESCRIPTION OF THE INVENTION (5) Technical Field of the Invention The present invention provides a recovery processing method for resident data files, particularly a failure recovery processing for a database/dataset that is made resident in memory during use. The present invention relates to a recovery processing method for resident data files that enables a significant reduction in recovery time by performing the recovery processing on a record-by-record basis.

The same conventional technology and problems To improve the stone processing efficiency regarding data files that have a relatively small capacity and are accessed frequently.

A so-called resident database/resident data set is known as a system that allows high-speed data processing by making the contents resident in virtual memory during use and omitting input/output access. These resident association data files Ic-') are generally updated frequently, and therefore a large amount of logs are generated as a history of updates. therefore.

The log is not a block, which is the unit of input/output (2) It is usually acquired in units of records.

FIG. 1 shows a diagram for explaining the problems of the conventional method. In the figure, 1 represents a direct access device such as a magnetic disk device, 2 represents a data file, 3 represents a block, 4 represents a record, 5 represents a log file, and 6 represents log data.

Data file 20 input/output to direct access device 1 is as follows:
This is done in blocks of three. Each block 3 usually consists of two or more records 4. The log file 5 is a file that records and saves update history information of each record 4. As log data 6 for transaction T1 that updates record "A1" to record "A 1'J",
The updated image rA1'J is recorded together with its record position information. Similarly, log data 6 of record rA 2'J is obtained for transaction T2, and log data 6 of record [Al1] is obtained for transaction T3.

If data file 2 becomes a failure.

Information is restored as follows. First, the data file 2 at the time of the full dump is restored from a full dump of the data file 2 that has been taken as a backup in advance. Then, restoration is performed using the log data 6 of the log file 5 collected after that.

An example of the conventional method will be explained as follows, focusing only on the block "AJ."

■READ block rAJ (on memory) (to tsufa).■ Read record "A1'J" on the buffer.
1”. ■WRITE block rAJ. ■READ block rAJ. ■Record “
Overwrite person 2/J. ■WRITB block “A”
do. ■READ block "A".

■Overwrite record unit 1”J. ■Block rAJ
WRITE.

In other words, according to the conventional method, block input/output is required when restoring records.

A total of six input/output activations are required: processing ■, ■, ■, ■, ■, ■. The above processing is for uniform processing.

Even if data file 2 is a normal data file,
Resident Φ data files are treated in the same way. Therefore, according to the conventional method, input/output is performed in blocks for each log data during recovery.
There was a problem that recovery time was long.

(Objective and Structure of the Invention The present invention aims to solve the above-mentioned problems, and reduces the number of input/output operations to a direct access device in the restoration process when a failure occurs in the resident data file, thereby shortening the recovery time and speeding up the recovery process. It is intended to be able to be restored to.

The resident data file recovery processing method of the present invention is a data file that has original data on a direct access device, and which is made resident in memory when the data file is used. The processing method includes a memory expansion unit that expands the data file on the direct access device or all dump information of the data file onto memory in the event of a failure, and a log file in which update history information is stored when updating the data file. a restoring means for restoring data on the memory in which the contents of the data file are expanded record by record; and data for writing the restored contents of the data file on the memory onto the direct access device after the restoration by the restoring means is completed. It is characterized by having a file storage means. Embodiments will be described below with reference to the drawings.

(G) Embodiment of the Invention FIG. 2 shows the configuration of an embodiment of the invention, and FIG. 3 is a process explanatory diagram of the recovery processing section shown in FIG. 2.

In the figure, numerals 1, 2, and 5 correspond to those in FIG. 1, 10 is an application processing section, 11 is a data management section, 12 is a data file definition body, 13 is a memory, 14 is a magnetic tape for backup, and 15 is a Log management section, 16 is Rika/< IJ processing section, 1
7 represents a resident determination section, 1B a memory development section, 19 a record restoration section, and 20 a data file storage section.

The application processing unit 10 refers to or updates data files 2 such as so-called databases/data sets, and executes processes desired by the user. When performing data processing of data file 2, data management section 1
1 issues macro instructions provided by the data management unit 11, and under the control of the data management unit 11.

Access data file 2. The data management unit 11 executes processes such as access control and failure management for the data file definitions 12 and data files 2. The data file definition body 12 holds the storage structure, logical structure and other attributes of the data file 2, position information of the data file 2 on the direct access device 1, and the like.

In particular, the data file definition body 12 includes data file 2
When a file has a resident attribute to the memory 13, its definition information is stored in advance.

When the application processing unit 10 first applies OPEN to the data file 2 having the resident attribute, the data management unit 11 refers to the data file definition body 12,
To know its attributes and speed up subsequent access,
Virtual space intercommunication area (C8A, ) K used as a common area of the memory 13, data file 2
Load the contents as is. Access from the application processing unit 10 to the data file 2 is thereafter performed on the memory 13, so that processing efficiency is improved.

Data file 2 between direct access device 1 and memory 13
The consistency of the contents of the memory 13 is ensured by writing back the contents of the close (CLO8F, ) M1, raccoon smell, and memory 13 onto the direct access device 1, for example, when the application processing unit 10 declares the end of use.

The backup magnetic tape 14 records and retains all dump information of the data files 2 on the direct access device 1 periodically, such as once a day or once a month. The log management unit 15 manages the log file 5 and executes a process of writing history information regarding updates of the data file 20 to the log file 5 each time based on notifications from the data management unit 11.

When the data file 2 becomes faulty, the data management unit 11 records the fault information in the data file definition body 12.

In the event of a failure, the recovery processing section 16 is activated automatically or by an operator. Recovery processing section 16
executes processing to restore the contents of the data file 2 to the state before the failure. The recovery processing unit 16 first calls the resident determination unit 17. The resident determination unit 17 determines whether or not the data file 2 to be restored has a resident attribute by referring to the data file definition body 12. When it has the resident attribute, the memory expansion unit 18 is activated.

The memory expansion unit 18 is a magnetic chip 14 for backup.
or from data file 2 on direct access device 1,
The original data is loaded into the memory 13' and expanded. This process is.

It can be considered that this is performed as part of the initial open processing by the data management unit 11. Once loaded onto the memory 13', control is transferred to the record restoration section 19.

The record restoration unit 19 reads log information, which is a post-updated image record, from the log file 5 in the order of log occurrence, and restores the contents of the data file record by record on the memory 13'. When processing for all logs is completed, control is transferred to the data file storage section 20. The data file storage section 20 executes a process of collectively writing out the contents of the data files on the memory 13' onto the direct access device 1'. The processing at this time can be considered to be performed as part of the normal closing processing in the data management unit 11. by this.

The restored data file 2' is transferred to the direct access device 1'
will be generated above.

The processing by the recovery processing unit 16 is expressed in the form of a flowchart as shown in FIG. That is,
The recovery processing unit 16 first reads the data file definition 12 to be recovered. Then, the fault book information and the like are taken in from the data file definition body 12, and it is determined whether the affected data file has the resident attribute. If it does not have the resident attribute,
It can be considered that the same restoration process as before is performed.

For residents, load the original data file into memory. Then, on the memory, an address is determined from the block number and record position based on the log in the log file 5, and restoration is performed record by record.

The logs are sequentially processed in the order in which they occur, and when the restoration on the memory using the logs is completed, the contents of the data file on the memory are finally written to the direct access device 1'.
E.

Finish the recovery process.

By performing the above processing, if the recovery processing is executed according to the present invention in a log environment as shown in FIG. 1, for example, block rAJ will be restored as follows. ■ EA the block rAj
Do D.

■ In memory, record r is in record “A1”.
A. Overwrite 1'J.

■ Overwrite record l'-A2Jk record [A 2 /J on the memory.

■ Overwrite record rA1'J with record "Al1ri" on memory.

■ Block “WRITE AJ to direct access device.

That is, in this example, input/output, which was conventionally required six times, only needs to be performed two times (2) and (2) above. When focusing on one block, there are 9 and above, but in particular, the writing process by the data file storage unit 20 shown in FIG. 2 can be performed continuously from the beginning of the entire data file, It is possible to minimize the movement of the data, and furthermore, it becomes possible to improve the efficiency of input and output.

(G) According to the present invention, as described in detail, when each record is restored using log information, the block containing the record is accessed from the direct access device each time.
There is no need for WR, ITE.

Since the number of inputs and outputs for restoration is reduced, recovery processing can be performed efficiently and the time required for recovery can be significantly shortened. In particular, since data files with resident attributes are expected to be accessed frequently and have a large amount of logs, the effect of reducing input/output is significant.

[Brief explanation of drawings]

FIG. 1 is a diagram for explaining the problems of the conventional method. FIG. 2 shows the configuration of an embodiment of the present invention, and FIG. 3 is a process explanatory diagram of the recovery processing section shown in FIG. 2. In the figure, 1 is a direct access device, 2 is a data file, and 5 is a direct access device.
is a log file, 13 is a memory, 14 is a magnetic tape, 1
Reference numeral 5 represents a log management section, 16 a file processing section, 18 a memory development section, 19 a record restoration section, and 20 a data file storage section. Patent applicant: Fujitsu Limited.

Claims (1)

[Claims] In a recovery processing method in the event of a failure of a resident data file that has original data on a direct access device and is made resident in memory when the data file is used, the data file on the direct access device is Or, the contents of the data file are expanded record by record by referring to the memory expansion section that expands all the dump information of the data file onto the memory, and the log file that stores the update history information of the update of the data file. The present invention is characterized by comprising a restoring means for restoring data on the memory, and a data file storage means for writing the contents of the restored data file on the memory onto the direct access device after the restoration by the restoring means is completed. Recovery processing method for resident association data files.