JP2623968B2 - Commit synchronization control method for journalless files - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a journal-less file commit synchronization control method, and more particularly to a program for processing a database file, such as an update transaction file or a temporary editing file without a journal. Commit synchronization control method when using

[Conventional technology]

Conventionally, when updating a record in a database file or adding a record to a database file, the update or addition to the database file is performed after all input / output of records processed in one transaction is completed. If the actual record is not output, inconsistency may occur in data update.

For this reason, in general, when updating or adding records in a database system, records that have been read from a database file and that have been updated or added are temporarily written to a journal file, and records that are processed in a transaction. A method is adopted in which the records of the database file are collectively updated at the time when all of the input / output operations are completed.

In such a database system, if a record to be updated is used by a plurality of users at the same time, data consistency may not be maintained. Therefore, a commitment control method for managing locks of records in a database file is used. Has been adopted.

For this reason, in the processing program of the conventional database system, in order to update the database file in a batch, a commit command to update the database file and release all record locks and to establish a quiesce point of the database file, A rollback command is provided that invalidates the update request to the database file executed after the immediately preceding still point and returns the state of the database file to the immediately preceding still point.

A processing program in such a database system may use a file having no journal (journalless file) as its input file or output file.

However, in a conventional database system, when a journalless file is used as an input file, a means for saving the current indicator of the journalless file at the time of a commit command for the database file, or a current indicator saved at the time of the commit command at the time of a rollback command There was no way to restore the current indicator of a journalless file using.

Further, when a journalless file is used as an output file, there is no means for outputting the output records in a batch at the time of a commit command, and no means for invalidating the records once output at the time of a rollback command.

Therefore, even if a rollback command that invalidates the update request to the database file is executed and the database file returns to the previous quiesce point, the file current indicator is not returned to the journalless file. In the case of a file, when the database is re-executed from the immediately preceding still point, record processing omission would occur.

On the other hand, in the journalless file in the output mode, the contents of the database file are returned to the state of the previous quiesce point by the rollback instruction, whereas the contents of the journalless file are not returned to the state of the previous quiesce point. There was a conflict with the contents of the database file.

For this reason, for a journalless file in the input mode, the user saves the current indicator of the journalless file by the processing program by himself / herself at the time of the commit instruction, and uses the current indicator of the file saved earlier at the time of the rollback instruction. A method of repositioning the address file has been attempted.

Regarding the journalless file in the output mode, a method of temporarily storing records requested to be output between commit commands by a processing program and then outputting the stored records collectively after a successful commit command has been attempted. Had been.

[Problems to be solved by the invention]

In the conventional database system described above, the commit command and the rollback command guarantee synchronous update for a database file having a journal.
For files that do not have a journal or for files that do not have a journal (floppy disk, magnetic tape file, etc.), it was not possible to save / restore the file current indicator or invalidate the output record.

For this reason, in the journalless file in the input mode, record processing is lost, and in the journalless file in the output mode, the contents of the database file and the journalless file are inconsistent.

Furthermore, in order to create a processing program to avoid these disadvantages, the programmer must take care in programming not to include journalless file I / O between commit instructions. There is also a drawback that it is one of the causes for reducing the productivity of the program.

Further, even if the processing program can cope with it to some extent, there is a drawback that the user must manually save the current indicator of the journalless file, which places a burden on the operator.

SUMMARY OF THE INVENTION In view of the above, it is an object of the present invention to be able to perform synchronous update of a journalless file when updating a database file, to improve data reliability and to save labor in operations and program creation. An object of the present invention is to provide a file synchronization control method.

[Means for solving the problem]

A journalless file commit synchronization control method according to the present invention includes a database file processed by a predetermined processing program, a commit command for performing an actual update of the database file and releasing a record lock to establish a quiesce point of the database file. A lock-back instruction for invalidating data update after the quiesce point immediately before the database file and establishing a quiesce point of the database file, and a commit synchronization control of a journalless file corresponding to a commitment control method for managing a record lock. When the commit command is executed, a file current indicator indicating the position of a record in the journalless file is saved, and when the journalless file is in the output mode, the record is actually output, and the journal is output. A journalless file commit control means for establishing a quiesce point of the nullless file, and when the rollback instruction is executed, the current indicator of the journalless file is returned to the immediately preceding commit instruction execution time, and the journalless file is output. In the case of the mode, journalless file rollback means for invalidating records output since the immediately preceding commit instruction execution and establishing a quiesce point of the journalless file,
A journalless file commit control calling unit that calls the journalless file commit control unit from the processing program; and a journalless file rollback control calling unit that calls the journalless file rollback control unit from the processing program. .

[Action]

In the present invention, when the commit command for the database file is executed, the journalless file commit control means is activated by the journalless file commit control calling means, and the journalless file commit control means applies the command to the journalless file. In the input mode, the file current indicator is saved, and in the output mode, commit control of the journalless file is performed by validating records output from the journalless file after the immediately preceding commit instruction, A quiesce point for a journalless file is established in synchronization with the database file.

When a database file rollback instruction is executed, the journalless file rollback control calling means activates the journalless file rollback control means, and the journalless file rollback control means causes the journalless file to enter its input mode. If this is the case, the file current indicator is restored using the file current indicator saved at the time of the immediately preceding commit command, and in the output mode, the records output after the immediately preceding commit command are invalidated, thereby rolling the journalless file. Back control is performed, and the quiesce point of the journalless file is restored at the time of the immediately preceding commit instruction in synchronization with the database file.

For this reason, the quiesce point and record of the journalless file are always matched with the database file, avoiding record processing omission and content inconsistency at the time of commit command and rollback command of the database file, thereby improving data reliability. As a result, the above-mentioned object is achieved by eliminating restrictions on programs and complexity at the time of operation.

〔Example〕

 Next, the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of a journalless file commit synchronization control method according to the present invention.

The user program 1 processes the database file 14, and includes a commit command module 2 and a rollback command module 5 for executing a commit command and a rollback command.

The commit command module 2 includes a database commit control calling unit 3, and the database file control unit 8 started by the database commit control calling unit 3 controls the commit of the database file 14.

The rollback command module 5 includes a database rollback control calling unit 6. The database file 14 is controlled by the database rollback control unit 9 started by the database rollback control calling unit 6.
Rollback control is performed.

Here, the commit control and the rollback control of the database file 14 are performed using the journal file 14A in which execution records and the like are stored. Further, the conventional file 15 (a file having no journal) is used for processing the database file 14 by the user program 1, and the conventional file 15 is a file management system including data such as the current indicator CPTR of the conventional file 15. Table 12
And an input / output buffer 13 used for inputting and outputting records to and from the user program 1.

On the other hand, the user program 1 of this embodiment adopts the journal-less file commit synchronization control method of the present invention. The rollback instruction module 5 includes a conventional file rollback control calling unit 7 in addition to the database rollback control calling unit 6.

The conventional file commit control calling means 4 activates the conventional file commit control means 10 to cause the conventional file 15 to perform commit control in synchronization with the commit control of the database file 14, and the conventional file rollback control calling means 7 The conventional file rollback control means 11 is started, and the rollback control of the conventional file 15 is performed in synchronization with the rollback control of the database file 14.

Further, the database file 14 stores the volume (Vo
l) It is located in # 1 and the conventional file 15 is the volume (Vo
l) Volume # 2 has a volume list 17 provided in volume # 2, and further has a file management table 12
And a save area 16 for saving data of the input / output buffer 13 and the like.

Next, the operation of the present embodiment thus configured will be described separately for the case where the conventional file 15 is in the input mode and the case where the conventional file 15 is in the output mode.

First, the case where the conventional file 15 is used in the input mode read from the user program 1 will be described.

When the commit command is executed in the user program 1,
The database commit control calling means 3 calls the database commit control means 8.

The database commit control unit 8 updates the database file 14 and releases all record locks, and outputs a journal JNL to the journal file 14A to establish a quiesce point of the database file 14.

At the same time, the conventional file commit control calling means 4 of the commit command module 2 calls the conventional file commit control means 10 and passes control.

Conventional file commit control means 10
The content CBUF 13 and the content CTBL of the file management table 12 are saved in the save area 16. As a result, the current indicator CPTR of the conventional file 15 recorded in the file management table 12 is also saved.

Here, when the user program 1 executes the rollback command, the database rollback control calling means 6
Causes the database rollback control means 9 to be called.

The database rollback control means 9 establishes the quiesce point of the database file 14 by invalidating the data update after the quiesce point immediately before the database file 14 and restoring the database file 14 based on the journal JNL of the journal file 14A. .

At the same time, the conventional file rollback control calling unit 7 of the rollback instruction module 5 calls the conventional file rollback control unit 11 and passes control.

The conventional file rollback control unit 11 restores the content CTBL of the file management table 12 saved in the save area 16 and the content CBUF of the input / output buffer 13 to the file management table 12 and the input / output buffer 13, respectively.

Therefore, the file management table 12 and the input / output buffer 13 are restored to the state at the time of executing the immediately preceding commit instruction.

When an input request is made to the conventional file 15 after the rollback instruction, the record next to the record indicated by the current indicator CPTR of the conventional file 15 saved by the immediately preceding commit instruction is input. That is, after executing the immediately preceding commit command, the current indicator CPTR of the conventional file 15 changed by the input command executed from the user program 1 to the conventional file 15 is always in the state at the time of execution of the commit command. Will be returned.

The processing in the conventional file commit control means 10 and the conventional file rollback control means 11 when such a conventional file 15 is in the input mode will be described more specifically.

As shown in FIG. 2, the input / output buffer 13 has four record areas 13A, 13B, 13C, and 13D and can store four records, respectively. Is designed so that one block, that is, four records, is fetched at a time. On the other hand, with one input command from the user program 1, each record stored in the record areas 13A, 13B, 13C, and 13D is fetched one by one in this order.

Now, it is assumed that the commit command of the user program 1 is executed at the time when the processing of the input record # 3 is completed. At this time, in the record areas 13A to 13D of the input / output buffer 13,
Input records # 1 to # 4 are stored respectively.

Here, the conventional file commit control means 10 started in response to the commit command is the content CBUF of the input / output buffer 13.
And the contents CTBL of the file management table 12 (the file current indicator CPTR indicates record # 3) in the save area
Evacuate to 16.

When the user program 1 makes an input request to the conventional file 15 in this state, the record area 13 in the input / output buffer 13
The content of the record # 4 read into D is passed to the user program 1. Further, when an input request is made, the block B2 of the conventional file 15, that is, the input records # 5 to # 8 are stored in the record areas 13A to 13D of the input / output buffer 13, respectively.
Record # 5 is passed to the user program 1.

If a deadlock or the like occurs in the database file 14 while processing the input record # 7 and a rollback command is issued from the user program 1, the database rollback control means 9 is called to roll the database file 14. Back control is performed.

At the same time, the conventional file rollback control means 11 is called, and the contents CBUF of the input / output buffer 13 at the time of the immediately preceding commit instruction stored in the save area 16 (records # 1 to #
4 is stored), and the content CTBL of the file management table 12 of the conventional file 15 (current indicator CPTR is record #
3), the input / output buffer 13 and the file management table 12 are restored.

Thereafter, when an input request is made to the conventional file 15, the record # 4 stored in the record area 13D is input and restored to the state at the time of the immediately preceding commit instruction.

Next, the case where the conventional file 15 is in the output mode used for writing the output from the user program 1 and the like will be described.

As shown in FIG. 3, when the commit instruction in the user program 1 is executed at the time when the processing of the output record # 3 is completed, the commit control of the database file 14 is performed by the database commit control means 8. .

At the same time, the content CTBL of the file management table 12 and the content CBUF of the input / output buffer 13 are saved in the save area 16 by the conventional file commit control means 10, and the content CBUF of the input / output buffer 13 is written to the conventional file 15. . Also, the block address CBLK of the current indicator CPTR of the conventional file 15 recorded in the file management table 12
Is also stored in the volume list 17 of volume # 2 to which the conventional file 15 is allocated.

When the user program 1 outputs to the conventional file 15 in this state, the record # 4 is stored in the record area 13D of the input / output buffer 13. When there is a further output request, since the input / output buffer 13 is full, the contents CBUF of the input / output buffer 13 are written to the conventional file 15 as one block. When writing, the volume list is taken by the immediately previous conventional file commit control means 10.
Based on the block address CBLK stored in 17, the data is written to the same position as the block B1.

However, since the block address CBLK of the volume list 17 has not been changed, even if there is an error in the user program 1 or the system goes down at this point, the output record # is output.
3 is guaranteed by writing at the time of the immediately preceding commit instruction, but the content of output record # 4 is treated as not being output.

Next, it is assumed that when the output record # 7 is output, a deadlock or the like occurs in the database file 14, and the conventional file rollback control unit 11 is called. First, the file management table saved in save area 16
With the contents CTBL of 12 and the contents CBUF of the input / output buffer 13, the file management table 12 and the input / output buffer 13 are restored to the state immediately after the commit instruction.

Thereafter, when the output record # 8 is output from the user program 1, this output record # 8 is stored at the position of the record area 13D of the restored input / output buffer 13, and when the next record is output, the input / output The contents of the buffer 13 (containing the output records # 1, # 2, # 3, and # 8) are written again into the block B1 of the conventional file 15 as one block.

Therefore, the output records # 4, # 5, # 6, and # 7 output after the immediately preceding commit instruction are invalidated by the rollback instruction.

According to the present embodiment, by providing the conventional file commit control unit 10, the conventional file rollback control unit 11, the save area 16, the volume list 17, and the like, synchronization with the establishment of the quiesce point of the database file 14 is achieved. The quiesce point of the conventional file 15 without a journal is also the content CTBL and the input / output buffer of the file management table 12.
The contents of 13 can be established by saving the CBUF to the save area 16.

Therefore, when the database file 14 returns to the state of the previous stationary point by the rollback command, the conventional file
15 also saves file management table 12 saved in save area 16
By restoring the content CTBL of the I / O buffer 13 and the content CBUF of the input / output buffer 13, it is possible to return to the state of the previous stationary point.

Therefore, when the conventional file 15 is in the input mode, it is possible to prevent omission of record processing. In the output mode, since the output records after the commit instruction become invalid as in the database file 14, the database file 14 is invalidated. There is no inconsistency in the contents of the conventional file 15.

These operations on the conventional file 15 are automatically performed by providing the conventional file commit control means 10 and the conventional file rollback control means 11, etc., so that the programmer who creates the user program 1 etc. There is no need to consider not including the input and output of the conventional file 15, and the productivity of the program can be improved.

At the time of a commit command, data saved in the save area 16 includes the contents CTBL of the file management table 12 and the contents of the input / output buffer 13 having four record areas 13A to 13D.
Since only the CBUF is required, only a small storage area is required compared to the journal file 14A, and there is an advantage that a file such as a floppy disk having no journal file can be used.

It should be noted that the present invention is not limited to the configuration and the like of the above-described embodiment, and modifications within a range that can achieve the object of the present invention are included in the present invention.

For example, as data to be saved in the save area 16, not only the content CTBL of the file management table 12 including the file present indicator CPTR and the content CBUF of the input / output buffer 13 but also a stationary point of the conventional file 15 can be established. Other data may be used as long as the data can restore the file 15 to the stationary point state.

The journalless file (conventional file 15) in the present invention is not limited in its use, such as a file used for record update or record editing, etc. In short, the present invention is widely used for a file having no journal file. it can.

〔The invention's effect〕

As described above, the present invention provides a journalless file commit control means for establishing a quiesce point of a journalless file when a commit command is executed, and a new input / output request after a quiesce point immediately before when a rollback command is executed. Journalless file rollback control means to establish a stable quiesce point, and journalless file commit control to start journalless file commit control means and journalless file rollback control means in synchronization with commit control and rollback control for the database By providing the calling means and the journalless file rollback control calling means, the journalless file can be automatically updated in synchronization with the update of the database file, and the operation can be saved. Phi Since the contents of the file do not conflict with the contents of the database file, the reliability of the data can be improved. In addition, since there is no need to perform programming in consideration of the input / output of the journalless file, the productivity of the program can be greatly improved. .

[Brief description of the drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, FIG. 2 is an operation explanatory diagram when a conventional file is in an input mode, and FIG. 3 is an operational explanatory diagram when a conventional file is in an output mode. . 1. User program 2. Commit command module 3. Database commit control calling means 4.
Conventional file commit control calling means, 5: rollback instruction module, 6: database rollback control calling means, 7: conventional file rollback control calling means, 8: database commit control means, 9: database rollback Control means 10, conventional file commit control means 11, conventional file rollback control means 12, file management table 13, input / output buffer 14, database file 14A journal file 15, … Conventional files that are journalless files, 16… Save areas, 17… Volume catalogs.

Claims (1)

(57) [Claims] 1. A database file to be processed by a predetermined processing program, a commit instruction for actually updating the database file and releasing a record lock to establish a quiesce point of the database file, and a quiesce immediately before the database file. And a lockback command for invalidating data update after the point and establishing a quiesce point of the database file, and a commit synchronization control method for a journalless file corresponding to a commitment control method for performing lock management of a record, wherein the commit command Saves the file current indicator indicating the position of the record in the journalless file, and when the journalless file is in the output mode, performs the actual output of the record and establishes the quiesce point of the journalless file Journal And less file commit control means,
When the rollback instruction is executed, the current indicator of the journalless file is returned to the time when the immediately preceding commit instruction was executed, and when the journalless file is in the output mode, records output since the previous commit instruction was executed are invalidated. And a journalless file rollback means for establishing a quiesce point of the journalless file; a journalless file commit control calling means for calling the journalless file commit control means from the processing program; and the journalless file from the processing program. And a journalless file rollback control calling means for calling a rollback control means.