JPH10133927A - Computer system and file managing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a computer system realizing rolling back at the time of generating a fault without waiting the saving of data before update at the time of updating a file. SOLUTION: When writing, etc., is requested to the file, 'file writing information' is preserved in an unidentified queue 431 to instantly update only a primary file 39. Then after a check point is picked up, 'file writing information' preserved in the queue 431 is moved to an identified queue 432 to reflect to a backup file 41. On the other hand, at the time of recovery, all the pieces of data before update corresponding to data updated after a finally picked check point are read from the file 41 based on 'file writing information' preserved in the queue 431 to recover the file 39 to the point of a check point time by using this read data before update.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a group computing process, a database process, a transaction process, and the like that require high reliability in a network computing environment including a plurality of computers connected to a network. The present invention relates to a computer system and a file management method suitable for application.

[0002]

2. Description of the Related Art The status of a process executed by a CPU, such as an address space, a context, and a file, is periodically collected (this is referred to as a checkpoint). In a system having a failure recovery function such as restoring the state of a point and restarting the execution of the process from that point, there has been a problem regarding external input / output processing. In other words, when a failure occurs, when the process is re-executed from the last collected checkpoint, the state such as the address space of the process and the context of the processor can be restored, but the state of the external input device is not easy to restore. Was.

For example, since it is difficult to cancel writing to a file, when writing to a file, data before writing is read and saved before writing data to the file. Data was being written to the file.

FIG. 15 shows that it is difficult to cancel writing to a file. Therefore, when writing to a file, data before writing is read and saved before writing data to the file. FIG. 1 is a diagram for explaining a mechanism of a conventional system for writing data to a file.

In this example, "X" is written to the first byte at time t2 in a file consisting of 4 bytes of data "ABCD" at the time when the checkpoint is taken at time t1 (1). In this case, conventionally, before writing "X" to the first byte of the file, data "B" of the first byte of the file is read (this is also called an undo log) (2), and thereafter, “X” is written in the first byte of the file (3).

After that, at time t3, the process is rolled back to the last checkpoint state (t1) because a failure has occurred. The first byte of the file is updated to “X” after the checkpoint t1, but the state of the checkpoint t1 is restored by using the undo log collected at the time of the update. This undo log becomes unnecessary at the next checkpoint and is discarded.

Also, for example, the computer is constructed by two computers, one of which (primary computer) is distributed as an active system and the other (backup computer) is divided into a standby system and duplicated, and when a failure occurs in the primary computer, the backup computer There is also a system that increases the availability of the system by taking over the processing. And with such a system,
As described above, if checkpoints are periodically collected, the reliability can be further improved.

[0008]

As described above, the state of the address space, the context, and the file of the process, that is, the checkpoint is periodically collected, and when a failure occurs, the state is finally collected. In a system (whether or not duplexed) having a failure recovery function of restoring the checkpoint state and restarting the process execution from that point, the reliability is improved. However, on the other hand, when updating (for example, writing) a file, the data before updating has to be once read from the file, and then the file must be updated. There were challenges.

The present invention has been made in view of such circumstances, and periodically collects checkpoints and restores the status of the last checkpoint collected when a failure occurs. In a system that has a recovery function from a failure such as restarting the execution of a process, when updating a file, it is not necessary to read the data before updating from the file, greatly improving the file update performance. It is an object of the present invention to provide a computer system and a file management method capable of performing such operations.

[0010]

A computer system according to the present invention is a computer system duplexed by two computers, an active system and a standby system, and has a checkpoint for restarting an interrupted process. In a computer system that periodically collects and saves the files on both the active and standby computers, a file updated by a process executed on the active computer is written on both the active and standby computers. When the file update is instructed by the process, the update information is stored on the standby computer, and only the active file is updated. When the update is completed, Means for notifying the update request source to the update requester, and after the checkpoint is taken , Characterized by comprising and means to reflect the update content indicated in the update information file of the standby system.

In the computer system according to the present invention, when a process requests an update of a file, update information indicating the update content is obtained and stored, and
Immediately updates only the files (active files) located on the active computer, and returns the result to the requesting process. Then, after the checkpoint is collected, the update content indicated by the stored update information is reflected in a file (standby file) arranged in the standby computer.

On the other hand, for example, when the process is aborted, the data before update corresponding to the data updated since the last collected checkpoint is stored in the standby file based on the stored update information. All are read out, and the operating system file is restored to the check point using the read out data before update.

That is, in this computer system, when a file is updated as in the prior art, the completion of processing such as reading out the data before update and saving the data is not waited for in the normal processing, and the file at the time of the failure is updated. Recovery can be realized, and the file update performance can be dramatically improved without deteriorating reliability.

Instead of restoring the active file, re-executing the process from the checkpoint using the standby file in which all the updates indicated by the update information stored before the final checkpoint are reflected. It is valid. In other words, continuation of processing when restart using the active file is impossible due to a failure of the active computer or the like is also ensured, and the availability of the system is improved. In this case, if a new standby file is secured in the third computer, the availability of the system can be further improved.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First, the basic principle of the present invention will be described with reference to FIG. As shown in FIG. 1, the computer system of the present invention is based on a system multiplexed with an active system 10 and a standby system 20. Hereinafter, each operation will be described.

(Normal Processing) (1) In the active system 10, the application program 11 issues a Write system call.

(2) The jacket routine 12 is Writ
The e-system call is hooked, a Write system call is issued to the active operating system, and the Write request is transmitted to the standby system 20. However, the write to the standby system 20 is immediately performed.
There is no need to send the request, just send it by the next checkpoint. In addition, the standby system 20 temporarily stores the received Write request in the undetermined queue 211 instead of immediately executing the received Write request.

(3) When a checkpoint process is instructed, the active system 10 must finish transmitting all accumulated Write requests to the standby system 20.

(4) On the other hand, in the standby system 20, the Write request stored in the undetermined queue 211 is moved to the defined queue 212.

(5) W transferred to the fixed queue 212
The write requests are sequentially processed by the operating system of the standby system 20.

That is, in the file update occurring in the normal processing, there is no need to wait for the completion of the processing of reading out the data before update and saving the data.

(Rollback Process) (3) ′ When a failure occurs, both the active system 10 and the standby system 20 are instructed to perform the rollback process.

(4) 'At this time, all Write requests remaining in the active system 10 are transmitted to the standby system 20.
Send to Since the Write request stored in the pending queue 211 of the standby system is issued after the last checkpoint, the data before the update is read out from the standby file 23 by referring to the Write request. The operating system file 14 is rolled back using the read data before update. As a result, both the active file 14 and the standby file 23 are in the state at the time of the last checkpoint.

(5) 'And the standby system 20
All Write requests remaining in the undetermined queue 211 are canceled.

As a result, it is possible to restart from the check point.

Next, an embodiment of the present invention will be described.

(First Embodiment) First, a first embodiment of the present invention will be described.
An embodiment will be described. FIG. 2 shows a system configuration of a computer system according to the first embodiment of the present invention. As shown in FIG. 2, in the computer system of the present embodiment, the computers are duplicated by a primary computer 30 and a backup computer 40, and these are connected by a network 50. The primary computer 30 and the backup computer 40 have both the active system 10 and the standby system 20 described above. When the active system 10 operates in one of them, the standby system 20 on the other side. Operate. Here, the primary computer 30 and the backup computer 40 will be described as the active system 10 and the standby system 20, respectively.

The process 35 is executed on the primary computer 30 and updates a file duplicated with the primary file 39 and the backup file 41. Here, the primary file 39 is located on the primary computer 30 and the backup file 41 is located on the backup computer 40, and are updated via the file system 36 on the primary computer 30 and the file system 48 on the backup computer 40.

The file system 36 on the primary computer 30 includes a primary file operation unit 38 and a primary file restoration unit 37. On the other hand, the file system 48 on the backup computer 40
Indicates the backup file operation unit 43 and the undetermined queue 4
31, a confirmation queue 432, a backup file updating unit 44, and a primary file restoration information reading unit 42
Contains.

When the process 35 updates the duplicated file, the update is performed via the primary file operation unit 38 and the backup file operation unit 43. The process 35 determines whether the wr corresponding to the duplicated file is wr.
When the “item” is performed, the primary file 39 is immediately updated as it is, but the backup file 41 is not updated at that time, and “file write information” is transmitted to the backup computer 40 via the backup file operation unit 43. Is stored in the unconfirmed queue 431.

When the process 35 collects checkpoints, the checkpoint control unit 31 issues an instruction to the checkpoint information storage unit 32 and the primary file operation unit 38. Upon receiving the checkpoint collection instruction, the checkpoint information storage unit 32 stores the checkpoint information (address space and processor context) on the primary computer 30 and the backup computer 40 (checkpoint information 34 on the primary computer 30). And checkpoint information 45 on the backup computer 40).

On the other hand, the primary file operation unit 38
Upon receiving the checkpoint collection instruction, the undetermined queue 431 is transmitted via the backup file operation unit 43.
Is moved to the confirmation queue 432. The “file write information” moved to the confirmation queue 432 is used for updating the backup file 41 by the backup file update unit 44 after the checkpoint is collected, and is discarded after the update of the backup file 41. Thus, the same write operation performed on the primary file 39 after the checkpoint is performed on the backup file 41.

When the process 35 causes a failure such as an abort and the process 35 is re-executed on the primary computer 30 from the last collected checkpoint,
The address space and the processor context are stored in the checkpoint information restoring unit 3 on the primary computer 30.
7 restored.

Regarding the file, the backup file 41 has not been updated since the checkpoint since the “file write information” is only stored in the undetermined queue 431 and is not actually updated. is there. However, the primary file 39 is
Since the update has already been performed after the checkpoint, restoration is necessary. Therefore, the undetermined queue 431
The pre-update data of the primary file 39 is read from the backup file 41 on the basis of the “file write information” stored in the backup file 41, and the read pre-update data is written to the primary file 39 to be restored. After that, the “file writing information” stored in the undetermined queue 431 is discarded. If the “file writing information” is stored in the confirmation queue 432, the “file writing information”
After the update to the backup file 41 is completed,
The restoration processing described above is started.

On the other hand, if the primary computer 30 or the operating system that controls the primary computer 30 causes a failure such as a system failure and re-executes the process 35 from the last checkpoint collected on the backup computer 40, the The space and the processor context are restored to the process 47 by the checkpoint information restoring unit 46.

Regarding the file, the backup file 41 has not been updated since the checkpoint since the “file write information” is only stored in the undetermined queue 431 and is not actually updated. is there.

The transfer of the "file write information" from the primary computer 30 to the backup computer 40 can be optimized. If the primary computer 30 did not go down when the failure occurred, the primary file 39 is restored,
The processing from the check point is restarted using the primary file 39. On the other hand, if the primary computer 30 goes down when a failure occurs, the processing is restarted from the checkpoint using the backup file 41.

Therefore, "file write information"
Need not be sent from the primary file operation unit 38 to the backup file operation unit 43 immediately. That is,
Since these "file writing information" need only be sent before the next checkpoint, considering transfer efficiency,
The data is temporarily stored in the primary file operation unit 38, and triggered by the occurrence of events such as "accumulated to a certain capacity", "elapse of a certain time", and "request for checkpoint collection", the backup file operation unit 43 It is possible to send them all together.

FIG. 3 shows a schematic configuration of a computer system to which the present embodiment is applied. The computer is duplicated by a primary computer 30 and a backup computer 40, and a disk device 60a is connected to the primary computer 30, and a disk device 60b is connected to the backup computer 40, respectively. The process 35 is executed on the primary computer, and the files accessed by the process 35 are duplicated by the primary file 39 and the backup file 41, and are respectively located in the disk devices 60a and 60b. I have.

The checkpoint information is transmitted to the primary computer 30 (primary checkpoint information 34) and to the backup computer 40 (backup checkpoint information 4).
5) Hold both. In this figure, the check points are stored on the disk device, but may be stored on the memory.

If a failure such as a system failure occurs in the primary computer 30 or the operating system that controls the primary computer 30, the backup computer 40 re-executes the process 47 using the checkpoint information 45. In this case, the process 47 uses the backup file 41.

It is also possible to have a plurality of primary files 39 or backup files 41 to create a triple or more file system. In this case, for example, in the case of a triple file system, a combination of (1) two primary files and one backup file, and (2) one primary file and two backup files can be considered.

FIG. 4 is a diagram showing how a file is updated in this embodiment. In this example, the process 35 running on the primary computer 30 has a duplicated file (a primary file 39 on the primary computer 30,
For the backup file 41) on the backup computer 40, at the time t1, "X" is added to the first byte.
(1). As a result, the primary file 39 is updated immediately, but the backup file 41 is not updated immediately, and only the “file write information” is stored.

Thereafter, a checkpoint is taken at time t2, thereby confirming the execution of the above-mentioned "file write information" (2). Then, after time t2, based on the determined “file write information”,
The backup file 41 is being updated.

FIG. 5 is a diagram showing how a primary file is restored when a failure occurs in this embodiment. In this example, the process 35 running on the primary computer 30 has a duplicate file (a primary file 39 on the primary computer 30 and a backup computer 40) having 4 bytes of data “ABCD”.
In the above backup file 41), "X" is written in the first byte at time t1 (1).
As a result, the primary file 39 is updated immediately, but the backup file 41 is not updated immediately, and only the “file write information” is stored.

Thereafter, a fault has occurred at time t2 (2). That is, the primary file 39 has been updated with the “file write information” at time t1 and needs to be restored, but the backup file 41 has not yet been updated and does not need to be restored. Here, the updated part of the primary file 39 is changed according to the “file writing information” stored at the time t1.
Therefore, in the restoration of the primary file 39, the data at the position indicated by the undetermined “file write information” is read from the backup file 41, and its r
Write the read data to the primary file 39
e, the primary file 39 is restored.

Then, the process 35 is re-executed on the primary computer 30 using the checkpoint taken on the primary computer 30. This re-executed process 35 uses the restored primary file 39.

FIG. 6 is a flowchart showing the flow of processing when the file operation unit is instructed to "write a file". In this case, first, the “file write information” is stored and linked to the undetermined queue 431 (step A1). Next, the primary file 39 is updated according to the "file write information" (step A2). At this point, it is determined that the "file write" operation has been completed, and a notification of completion is given to the requesting side (step A).
3).

FIG. 7 is a flowchart showing the flow of processing when the file operation unit is instructed to "checkpoint collection". In this case, the stored “file write information” is transferred from the undetermined queue 431 to the determined queue 4.
32 (step B1).

FIG. 8 is a flowchart showing the flow of the process of the backup file update unit. In this case, first, it is checked whether “file writing information” is linked to the confirmation queue 432 (step C1).
If they are not linked (N in step C1),
The backup file updating unit 44 continues this check. On the other hand, if linked (Y in step C1), the backup file 41 is updated based on the “file write information” linked to the confirmation queue 432 (step C2). Then, the executed “file writing information” is removed from the confirmation queue 432 (step C3).

FIG. 9 shows that a failure such as abort occurs in the process 35 and the process 35
11 is a flowchart showing the flow of processing when re-executing from a checkpoint that was last collected on 0.

When a failure occurs in the process 35, first,
Instruct the checkpoint information restoring unit 33 on the primary computer 30 to restore the address space and the processor context (step D1), and then instruct the primary file restoring unit 33 to restore the primary file (step D1). Step D2).

FIG. 10 is a flowchart showing the flow of processing when the checkpoint information restoring unit on the primary computer 30 is instructed to "restoring address space and processor context". In this case, first, the address space of the process 35 is restored (step E).
1). Next, the processor context state at the time of checkpoint collection of the process 35 is restored (step E).
2).

FIG. 11 shows the primary file restoring section 37.
Is a flowchart showing the flow of processing when "restore primary file" is instructed. In this case, first, the “file write information” is stored in the undetermined queue 431.
Is linked or not (step F
1). When the “file writing information” is linked (Y in step F1), according to the “file writing information” linked to the undetermined queue 431,
The updated data in the primary file 39 is read from the backup file 41, and the read data is written to the primary file 39.
to restore the updated data of the primary file 39 (step F).
2). Then, the “file write information” used for restoration is removed from the undetermined queue 431 (discarded).
(Step F3). This process is repeated until there is no more “file write information” linked to the undetermined queue 431.

When a failure such as a system failure occurs in the primary computer 30 or the operating system that controls the primary computer 30,
The process 35 is re-executed on the backup computer 40 from the last collected checkpoint. In this case, the processing is taken over by the backup file 41. FIG.
2 is the backup file 4 when a failure occurs
FIG. 2 is a diagram showing a state of taking over the process in step S1.

In this example, the primary computer 30
The process 35 that operates on the above-described process has four bytes of data “AB
At the time t1, "X" is written to the first byte of the duplicate file having the "CD" (the primary file 39 on the primary computer 30 and the backup file 41 on the backup computer 40).
e (1). This makes the primary file 3
9 is updated immediately, but the backup file 41 is not updated immediately, and only “file write information” is stored.

Thereafter, a failure has occurred in the primary computer 30 at time t2 (2). in this case,
The process 47 is re-executed on the backup computer 40 using the checkpoint taken on the backup computer 40. At this time, process 47
Continues the processing using the backup file 41, but at time t1, the primary file 39 has been updated, but since the backup file 41 has not been updated yet, the process 47 is re-executed on the backup computer 40. In, the backup file 42 can be used as it is.

When the backup file is separated due to the occurrence of a failure, the initial state as shown in FIG. 1 can be reproduced again by creating a new backup file thereafter. However, the recovery process can be performed.

When the backup file takes over the processing due to the occurrence of a failure and the processing is re-executed from the checkpoint, a new backup file is created with the backup file as a primary file, as shown in FIG. A simple initial state can be reproduced, and recovery processing can be performed even if a failure occurs again. To create the backup file again, there are the following two methods.

(1) The update information and data of the primary file after the separation of the backup file are stored, and when the backup file is reconnected, the update information and the data of the primary file after the separation are added to the backup file. To reflect.

(2) Copy the primary file to the backup file. However, if the primary file is continuously updated during the copy, the update information and data of the file are reflected in the backup file at the same time as the copy is started.

Further, the following method combining these two methods is also effective.

(3) Assuming that the detached backup file (or the primary file before the occurrence of the failure) is reconnected, after the backup file is detached so that the method of (1) can be performed until a certain time elapses. The update information and data of the primary file are stored. After a certain period of time, the method (1) is closed, the storage of the update information and data of the primary file after the backup file is separated is stopped, and the method (2) is adopted. Also, when reconnecting with a file other than the disconnected backup file, saving the update information and data of the primary file after the disconnection of the backup file is stopped and the method (2) is adopted.

(Second Embodiment) Next, a second embodiment of the present invention will be described.
An embodiment will be described. In the first embodiment, a duplicated computer system has been described.
It is also effective when applied to a file system on a computer that is not duplicated. Therefore, in the present embodiment, 2
An example in which the present invention is applied to a file system on an unduplicated computer will be described. FIG.
FIG. 11 is a configuration diagram when applied to a file system on a non-duplicated computer. In this system, the computers are not duplicated, and only the computer 30 exists. The process 35 is executed on the computer 30 and updates a file duplicated by the primary file 39 and the backup file 41. That is, the primary file 39 and the backup file 41 are both located on the computer 30 and are updated via the file system 36.

File System 3 on Computer 30
Reference numeral 6 includes a primary file operation unit 38, a primary file restoration unit 37, a backup file operation unit 43, an unconfirmed queue 431, a confirmation queue 432, a backup file update unit 44, and a primary file restoration information read unit 42.

The process 35 updates the duplicated file via the primary file operation unit 38 and the backup file operation unit 43.
Process 35 writes wr to this duplicated file.
When the “item” is performed, the primary file 39 is updated as it is, but the backup file 41 is not updated, and “file write information” is stored in the undetermined queue 431 via the backup file operation unit 43.

When the process 35 collects a checkpoint, the checkpoint control unit 31 issues an instruction to the checkpoint information storage unit 32 and the primary file operation unit 43. Checkpoint information storage 3
When receiving the checkpoint collection instruction, 2 performs the address space and the processor context on the computer 30 (checkpoint information 34).

On the other hand, the primary file operation unit 38
Upon receiving the checkpoint collection instruction, the “file writing information” stored in the undetermined queue 431 is moved to the defined queue 432 via the backup file operation unit 43. The “file write information” moved to the confirmation queue 432 is used for updating the backup file 41 by the backup file update unit 44 after the checkpoint is collected, and is discarded after the update of the backup file 41. Thus, the write operation is performed on the backup file 41 in the same manner as performed on the primary file 39 after the checkpoint.

When a failure such as an abort occurs in the process 35 and the process 35 is re-executed from the last checkpoint collected on the computer 30, the address space and the processor context are stored in the checkpoint information restoring unit 33 on the computer 30. Restored by

Regarding the file, the backup file 41 has only the “file write information” stored in the undetermined queue 431 whose update since the checkpoint has not yet been performed. is there. However, the primary file 39 is
Since the update has already been performed after the checkpoint, restoration is necessary. Therefore, the undetermined queue 431
The pre-update data of the primary file 39 is read from the backup file 41 based on the “file write information” stored in the backup file 41, and the read pre-update data is written to the primary file 39 to restore the data. After that, the “file writing information” stored in the undetermined queue 431 is discarded. If the “file writing information” is stored in the confirmation queue 432, the “file writing information”
After the update to the backup file 41 is completed,
The restoration processing described above is started.

FIG. 14 shows a schematic configuration of a computer system to which the present embodiment is applied. The system of the present embodiment operates only by the computer 30 and is not duplicated. The disk device 60a and the disk device 60b are connected to the computer 30. Process 35
Is executed on the computer 30, and the file accessed by the process 35 is the primary file 3
9 and the backup file 41 are duplicated,
They are arranged in the disk device 60a and the disk device 60b, respectively.

As described above, by applying the present invention, the execution (checkpoint information) such as the address space of the process and the context of the processor is continuously performed while being periodically saved, and the last saved when a failure occurs. In a system that takes measures against a failure by re-executing the process from the specified checkpoint, when updating the file, it is no longer necessary to read the pre-update data from the file, and the file update performance has been greatly improved. You.

The file management method described in the above embodiment can be distributed by storing it in a recording medium such as a floppy disk, an optical disk, or a semiconductor memory as a computer-executable program.

[0074]

As described above in detail, according to the present invention, when a process requests a file update, update information indicating the update content is acquired and stored, and only the primary file is immediately updated. Then, after the checkpoint is collected, the update content indicated by the stored update information is reflected in the backup file.
Then, for example, when the process aborts, based on the stored update information, all the pre-update data corresponding to the data updated since the last collected checkpoint is read from the backup file, and The primary file is restored to the time of the checkpoint using the read data before update, and the process is started again (the process can be started again using the backup file).

That is, in this computer system, when a file is updated as in the prior art, the completion of processing such as reading out the data before update and saving the data is not waited for in the normal processing, and the file at the time of the failure is updated. Recovery can be realized, and the file update performance can be dramatically improved without deteriorating reliability.

[Brief description of the drawings]

FIG. 1 is a conceptual diagram for explaining a basic principle of the present invention.

FIG. 2 is an exemplary view showing a system configuration of a computer system according to the first embodiment of the present invention.

FIG. 3 is an exemplary view showing a schematic configuration of a computer system to which the embodiment is applied.

FIG. 4 is an exemplary view showing how a file is updated in the embodiment.

FIG. 5 is an exemplary view showing how a primary file is restored when a failure occurs in the embodiment.

FIG. 6 is an exemplary flowchart showing the flow of processing when the file operation unit of the embodiment is instructed to “write a file”;

FIG. 7 is an exemplary flowchart showing the flow of processing when the file operation unit of the embodiment is instructed to “checkpoint collection”;

FIG. 8 is an exemplary flowchart showing the flow of the process of the backup file updating unit of the embodiment.

FIG. 9 is an exemplary flowchart showing the processing flow when a failure such as an abort occurs in the process of the embodiment and the process is re-executed from the last checkpoint collected on the primary computer 30;

FIG. 10 is an exemplary flowchart illustrating the flow of processing when the checkpoint information restoring unit on the primary computer according to the embodiment is instructed to “restoring an address space and a processor context”;

FIG. 11 is an exemplary flowchart showing the flow of processing when the primary file restoration unit of the embodiment is instructed to “restore a primary file”;

FIG. 12 is an exemplary view showing how a backup file takes over the processing when a failure occurs in the embodiment.

FIG. 13 is a diagram showing a system configuration of a computer system according to a second embodiment of the present invention.

FIG. 14 is an exemplary view showing a schematic configuration of a computer system to which the embodiment is applied.

FIG. 15 shows that it is difficult to cancel writing to a previous file. Therefore, when writing to a file, data before writing is read and saved before writing data to the file. FIG. 2 is a view for explaining the structure of a conventional system for writing data to a file.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 ... Operating system, 11 ... Application program, 12 ... Jacket routine, 13 ... OS buffer cache, 14 ... Disk device, 20 ... Standby system, 21 daemon, 211 ... Indeterminate queue, 212 ... Confirmed queue, 22 ... OS Buffer cache, 23 disk device, 30 primary computer, 31 checkpoint control unit, 32 checkpoint information storage unit, 33 checkpoint information restoration unit, 34 checkpoint information, 35 process, 36 file system 37, primary file restoring unit, 38, primary file operating unit, 39, primary file, 40
... backup computer, 41 ... backup file, 42 ... primary file restoration information reading unit,
43 backup file operation unit, 431 undetermined queue, 432 fixed queue, 44 backup file update unit, 45 checkpoint information, 46 checkpoint information restoration unit, 47 process, 50 network, 60a, 60b ... Disk devices.

Claims (16)

[Claims] 1. A computer system duplexed by two computers, an active system and a standby system, wherein a checkpoint for restarting an interrupted process is periodically collected, and In a computer system that saves data on both computers of the system, files updated by a process executed on the computer of the active system are provided in duplicate on both the computer of the active system and the computer of the standby system. When instructed to update the file,
Means for storing the update information on the standby computer and updating only the active file, and when the update is completed, notifying the update request source of the update completion; and Means for reflecting the update content indicated in the update information to the standby file after the update. 2. The system according to claim 1, further comprising means for buffering the update information on the active computer and transferring the update information to the standby computer by the time the checkpoint is collected. The computer system of claim 1, wherein 3. When the process is aborted, data before update for a file update executed since the last checkpoint is read from the standby file using the update information, and the active file is read from the standby file. 3. The computer system according to claim 1, further comprising: means for re-executing the process from the check point after restoring the state at the time of the check point. 4. When the process aborts, deletes the update information stored since the last checkpoint and reflects the update indicated by the update information before the checkpoint on the file of the standby system. 3. The computer system according to claim 1, further comprising means for re-executing said process from said checkpoint on said standby computer. 5. When a failure occurs in the active computer or the operating system controlling the active computer, the update information stored after the last checkpoint is deleted, and the update before the checkpoint is performed. 2. The system according to claim 1, further comprising: a unit configured to re-execute the process from the checkpoint on the standby computer after reflecting the update indicated by the information on the standby file. 3. The computer system according to 2. 6. The system further comprises means for stopping transfer of the checkpoint and update information to the standby computer when a failure occurs in the standby computer or an operating system controlling the standby computer. 3. The computer system according to claim 1, wherein: 7. When a failure occurs in the active file, the update information stored after the last checkpoint is deleted, and the update indicated by the update information before the checkpoint is transferred to the standby file. 3. The computer system according to claim 1, further comprising: means for re-executing the process from the checkpoint on the standby computer after reflecting the process on the standby computer. 8. The system according to claim 1, further comprising means for stopping transfer of the checkpoint and update information to the standby computer when a failure occurs in the standby file. Or the computer system according to 2. 9. The system according to claim 1, further comprising means for newly securing a standby file on the third computer when a standby file is separated. Computer system as described. 10. When the process is re-executed from the checkpoint using a standby file, the standby file is switched to an active file, and a new file is stored on the active computer. 3. The computer system according to claim 1, further comprising means for securing a standby file. 11. A computer which is duplexed by two computers, an active system and a standby system, periodically collects checkpoints for restarting the interrupted processing and stores the checkpoints on the computers of both the active system and the standby system. A file management method for a computer system wherein a file to be stored and updated by a process executed on the active computer is duplicated on both the active and standby computers. When an update is ordered,
Storing the update information on the standby computer, updating only the active file, and notifying the update request source to the update requester when the update is completed; Reflecting the update content indicated in the update information in the file of the standby system after the update. 12. The data before update for a file update executed since the last checkpoint is read from the file of the standby system based on the update information, and the file of the active system is restored to the state at the time of the checkpoint. The method of claim 11, further comprising the step of re-executing the process from the checkpoint afterwards. 13. The method according to claim 12, wherein the update information stored after the last checkpoint is deleted, and the update indicated by the update information before the checkpoint is reflected in the file of the standby system. The file management method according to claim 11, further comprising a step of re-executing from the check point on a computer. 14. A computer which is duplexed by two computers, an active system and a standby system, periodically collects checkpoints for restarting the interrupted processing and stores the checkpoints on the computers of both the active system and the standby system. A program for managing files of a computer system provided to store and multiplex files updated by a process executed on the active computer on both the active and standby computers, When the process instructs to update the file,
The update information is stored on the standby computer, and only the active file is updated.When the update is completed, the update request is notified to the update request source, and the checkpoint is collected. And a computer-readable storage medium storing a program for causing the computer to operate so that the update content indicated in the update information is reflected in the standby file. 15. The program reads data before update for a file update executed since the last checkpoint from the standby file based on the update information, and reads the active file at the time of the checkpoint. The computer-readable medium of claim 14, further comprising causing the computer to re-execute the process from the checkpoint after restoring the state. 16. The program deletes the update information stored since the last checkpoint, reflects the update indicated by the update information before the checkpoint on the file of the standby system, and then executes the process. 15. The computer-readable storage medium according to claim 14, further causing the computer to operate again from the checkpoint on the standby computer.