JPH0635781A - Recorded data on-line recovery device - Google Patents

Abstract

(57) [Summary] [Configuration] In a mirror disk system accessed by a plurality of computers, a copy read area 31 on the master disk 30 side of the disk controller 20 is set as a lock area 31, and writing from online becomes a lock area. also with run without an error if the corresponding, City broadcast <br/> tell the lock area write status, corresponding to Carlo area is restored more open computer side 10, then the disk control device 20 receives this report Writing to the slave disk 40 side is not performed until an interrupt comes. As a result, it is not necessary to restart the write request from the computer due to the communication between computers or the lock error, and the online recovery copy operation can be speeded up.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to online disk data recovery in a database system, an online transaction system, etc.

[0002]

2. Description of the Related Art In recent years, data reliability has been particularly strongly demanded in database systems, online transaction processing systems, etc., and mirror disk systems have been widely adopted. Copying from a disk to a recovery target slave disk) The increase in processing time has become a problem. In addition, many computer systems include loosely coupled systems such as a dual system, a standby system, and a load balancing system. In these systems, a recovery copy operation of a disk is requested under online access from a plurality of computers. If the order of the recovery copy operation of the stored contents and the online writing (to the master and slave disks) is not regulated, the online write data will be lost. In order to prevent this, conventionally, a computer that performs recovery processing by communication between computers issues a write-protection command to the slave disk side specified area (in units of cylinders, files, etc.) to another computer, and It was necessary to wait for a response that it was established, perform a copy operation, and after copying, contact each other for area lock such as removing write protection of the specified area and moving to the next area. Alternatively, in order to eliminate this inter-computer communication, the slave disk side of the disk controller has an area lock function, online write to the locked area is returned as a lock error, and a write request is sent again from the computer after waiting for the lock area release interrupt. A lock error reporting method of issuing "." Was also considered.

[0003]

Area lock communication may be required for multiple computers, and when a write-protection command for a designated area is issued, each computer has already requested writing to the designated area. If so, the response that the write-protection is established cannot be returned until the completion. Also, in order to reduce the number of communications for this area lock, the size of the area is set larger than the copy size per one time. For these reasons, the communication overhead of the computer and the copy elapsed time increase, which requires a considerably longer time than the copy time in a single computer system. Alternatively, in the lock error reporting method in which the slave disk side of the disk control device has an area lock function, even though it is considerably improved, the number of times the processing routine is started and the number of interrupts are more than normal writing when a lock error occurs, Neither computer load nor elapsed copy time was optimal.

The present invention has been made to solve the above problems, and provides a recorded data online recovery apparatus capable of recovering data at high speed during online operation when a failure occurs in the data in the database. The purpose is to

[0005]

The recorded data online recovery apparatus according to the present invention has the following requirements. (A) a master recording device for recording data, (b) a slave recording device for recording the same data as the data recorded in the master recording device, (c) the same data for the master recording device and the slave recording device Writing means for writing, (d) failure recovery means for recovering a failure of the slave recording device during the operation of the writing means, including the following elements:
(D1) area registration means for registering a copy area of the master recording device for copying data from the master recording device to the slave recording device, (d2) data of the copy area of the master recording device registered by the area registration means Means for copying the data to the slave recording device, and (d3) when writing by the writing device occurs in the copy area of the master recording device registered by the area registration device, the writing to the master recording device is permitted and Write control means for suspending writing to the slave recording apparatus and causing the slave recording apparatus to execute the suspended writing after the copying by the copying means is completed.

[0006]

In the recorded data online recovery device according to the present invention, of the master recording device and the slave recording device,
The master recording device has area registration means for registering a copy area of the master recording device for copying data to be recovered to the slave recording device, and the writing control means for writing to the master recording device, Since all writing is permitted regardless of the copy area registered by the area registration means, it is possible to issue write requests to the master recording device one after another. On the other hand, when the writing means writes to the slave recording device in the copy area registered by the area registration means, the writing control means suspends the writing and executes the copied writing after the copying by the copying means is completed. Therefore, the problem that the latest data is lost by writing the data to the slave recording device by the writing means before the copying by the copying means is eliminated.

[0007]

1 is a diagram showing an embodiment of a recorded data online recovery device according to the present invention. 10 in the figure
Is a computer, 20 is a disk controller connected to the computer 10, 30 is a master recording device, and 40 is a slave recording device for recording the same data as the data recorded in the master recording device 30. Reference numeral 11 is a writing unit that records data in the master recording device 30 and also records the same data in the slave recording device 40.
Reference numeral 50 denotes a failure recovery means for recovering the data of the slave recording device by copying the data from the master recording device to the slave recording device because the data of the slave recording device has a failure. 21 is a master recording device 3
Area registration means for registering a specific area of 0 as the copy area 31. 12 is area registration means 21
It is a copy means that reads data in the copy area 31 registered by and writes the data in the corresponding area of the slave recording device 40. When the writing unit 11 tries to write data to the copy area 31 registered by the area registration unit, the reference numeral 22 permits the master recording device to write the data and the writing to the slave recording device. It is a write control unit that holds this and permits the writing unit to write data to the slave recording device after the copying by the copying unit is completed.

The operation will be described below with the master recording apparatus referred to as a master disk, the slave recording apparatus referred to as a slave disk, and the copy area referred to as a lock area. The master disk side of the disk controller is provided with an area lock function, and when reading for copying, the computer commands the disk controller to lock the area. The area registration means registers this area. The writing control means permits normal writing to both the master disk and the slave disk when the writing means writes to areas other than this locked area. Also, when writing online to the master disk side locked area where writing to the slave disk side has not been completed and the area lock release command has not come, direct writing is performed as in the unlocked area, and Is detected, and in the case of overlap, the write area is registered in the disk controller together with a unique sequence number. Then, the lock area write status is transmitted to the computer together with the sequence number when the write completion report to the master disk is reported.

Here, the lock area write status is a flag for indicating that writing of data from the computer corresponds to the lock area. For example, this lock area write status is indicated by 1 bit,
When it is 0, it indicates that the data to be written does not correspond to the lock area, and when it is 1, it indicates that it corresponds to the lock area. The sequence number is a number assigned in sequence to the lock area write status of 1. The write control means 22 sets the above-mentioned lock area write status to 1 and allocates a sequence number, and reports this to the computer together with a data write completion report to the master disk.

On the computer side, the writing means 11 normally issues a write request to the slave disk subsequent to the completion of writing to the master disk, but when the lock area write status is reported from the master disk, it is transmitted. Stores the sequence number and suspends writing to the slave disk.

Next, when the data in the lock area is copied from the master disk to the slave disk by the copy means, the write control means detects the end of the data and the write request in which writing to the slave disk is suspended during copying. To the write means of the computer, which notifies the write means of the computer of the release of the lock area together with the sequence number assigned by the write control means. When the lock area release interrupt having the same sequence number as the stored sequence number is transmitted from the disk control device, the writing means makes a write request corresponding to the sequence number to the slave disk.

As described above, by controlling so as to hold the lock area release interrupt having the same sequence number next from the disk control device without performing writing start to the slave disk, the slave disk on the slave disk is controlled. It is possible to prevent the new online write data of the above from being lost due to the old copy data. This allows each computer to issue write requests to the master disk one after another without waiting for the lock area to be released, and even if a plurality of them correspond to the write to the lock area, there is no problem. Can be dealt with,
There is no need to restart the processing routine as in the lock error reporting method described in the conventional example. In addition, by adopting the method of giving a unique sequence number when registering the writing area, it is possible to write to the slave disk in sequence number order, so the preceding lock area is not released yet but the latter lock area is released first. In this case, the slave disks are written in the order in which the writing actually occurs, and the problem of order inversion does not occur.

According to the above-mentioned method, the disk control device starts the disk as usual, and after the command is found to be writing and the writing to the disk is started, whether the lock area is designated or not Logic for writing judgment, writing area registration, and simple additional processing at the end report, lock area registration,
It can be said that this is an efficient method because it only requires a change command, a release command, and a lock area release process corresponding to them.

The write area in which writing is suspended by the write control means is registered in the disk controller together with the sequence number, but it is necessary to register a plurality of write areas for each computer.

By adopting the method of registering the write area on the side of the disk controller as described above, when the write request is made over both the lock areas of the previous time and this time, the lock of the previous time and the current time is performed. After the area lock is released, the disk controller outputs a lock area release interrupt with that sequence number, so the computer side received a lock area release interrupt as in the past and restarted. It can be eliminated without being hit as a lock error.

As described above, according to the present embodiment, in the mirror disk data online recovery device, the data is copied from the master disk to the slave disk during the online failure recovery of the mirror disk system accessed by a plurality of computers. The master disk side of the control unit has a means to register the area (copy target area) instructed by the computer as a lock area, and the writing command from the computer to the master disk is executed as usual, and the writing area and the lock area are It has a means to detect the overlap of the lock area, register the write area with a unique sequence number in the case of overlap, and report the lock area write status and sequence number at the end of writing. In response to the change command of (a), the lock area release interrupt is reported together with the sequence number for the registered write area that is no longer included in the new lock area or for all the write areas registered for the lock area release command. A disk control device is provided.

A read and lock function (in addition to the normal read function, a function of releasing the previous lock area and a read target area are registered as lock areas in the disk control device as command commands for the copy target lock area). Function).

According to the read & lock function, the area lock request and the release request which are normally performed are not separately required (except the last release request), and the normal read function can be performed by one command command. In addition to this, since it has the function of releasing the previous lock area and the function of registering the read target area as the lock area, it is possible to reduce the number of inputs and outputs to the disk control device. Also, by limiting the lock range to the necessary minimum, the probability of lock is reduced, and it can be said that this is an efficient method with less overhead.

Next, a specific operation example of this embodiment will be described. FIG. 2 shows an example of application of the present invention to a dual computer system, in which a disk controller, mirror disks (master disks A, C, E are slave disks B, D, F).
Correspond to each other). FIG. 3 is a diagram showing the necessity of the write area lock at the time of recovery processing of the mirror disk, and the lines a and b show the input / output to / from the master and slave disks. Here, the recovery process is a0r-b0w, a1r.
Reading from the master disk and writing to the slave disk are repeated in the order of -b1w, .... xa0w-xb0w, ya0w-yb0w, etc. indicate the mirror mode writing from the online program.
The write xb0w corresponding to xa0w whose writing to the area a0 has already finished before the recovery copy read a0r may be executed independently of the write b0w from the copy program, but the write y after a0r
If yb0w corresponding to a0w precedes writing b0w from the copy program, the data will be lost as if it were not written to the B disc. Therefore, in this embodiment, when the lock area write status is reported for the completion of the online writing to the A disk, the copy writing to the B disk is completed and the area lock release interrupt is reported (area lock request, (When the release command is executed independently), or until the area lock release interrupt is reported by the start of execution of the next area copy reading a1r from the A disk (when the read & lock function is used)
Data loss was prevented by substantially suspending the end processing of the A disk and the mirror mode writing to the B disk. Here, the end interrupt processing of ya0w may be delayed for the period X until the end of the copy writing to the B disk, but in the case of the read & lock method, it means until the next copy reading is started in order to reduce the number of questions and answers with the disk device. Or until the area lock is released): Delayed during section S.

Example 2. Although the first embodiment has described the case of including a plurality of computers, it is also effective in the case of a single computer. Further, in the first embodiment, the case where the two disk control devices are provided on the master side and the slave side is considered, but the control device 1 is focused on improving the reliability of the disk medium / device.
There may be a case where a mirror disk system is composed of individual pieces. In this case, even if the writing to the lock area is completed, it is not reported to the computer immediately, the end interrupt is output after waiting for the lock area to be released, and the area other than the recovery copy processing program is not aware of the area lock at all. It is also possible to adopt a system that allows it. However, in this case, it is necessary to accept a plurality of activation requests for each computer and measure the effective use of the disk.

Example 3. In the first embodiment, the case where the medium for recording the data is the disk is shown, but the recording medium is not limited to the disk, and any other magnetic recording device or an optical recording device may be used as the recording device for recording the data. Absent.

Example 4. In the first embodiment described above, the case of recovering data from the master disk to the slave disk is shown. However, the master refers to the person who holds the data, and the slave refers to the person who has a failure. It is not limited to a master or slave relationship when data is recorded in a master-slave relationship during normal use of the computer system. That is,
The one holding the data to be recovered is the master, and the one with the failure is the slave.

[0023]

As described above, according to the present invention, it is not necessary to restart the write request from the computer due to the communication between computers and the lock error, and the online recovery copy operation can be speeded up.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a recorded data online recovery device according to the present invention.

FIG. 2 is a diagram showing a specific example of a recorded data online recovery device according to the present invention.

FIG. 3 is a diagram for explaining the operation of the recorded data online recovery device according to the present invention.

[Explanation of symbols]

 10 computer 11 writing means 12 copying means 20 disk control device 21 area registration means 22 writing control means 30 master recording device (master disk) 40 slave recording device (slave disk) 50 failure recovery means

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] July 12, 1993

[Procedure Amendment 1]

[Document name to be amended] Statement

[Correction target item name] Full text

[Correction method] Change

[Correction content]

[Document name] Statement

Recorded data online recovery device

[Claims]

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to online disk data recovery in a database system, an online transaction system, etc.

[0002]

2. Description of the Related Art In recent years, data reliability has been particularly strongly demanded in database systems, online transaction processing systems, etc., and mirror disk systems have been widely adopted. Copying from a disk to a recovery target slave disk) The increase in processing time has become a problem. In addition, many computer systems include loosely coupled systems such as a dual system, a standby system, and a load balancing system. In these systems, a recovery copy operation of a disk is requested under online access from a plurality of computers. If the order of the recovery copy operation of the stored contents and the online writing (to the master and slave disks) is not regulated, the online write data will be lost. In order to prevent this, conventionally, a computer that performs recovery processing by communication between computers issues a write-protection command to the slave disk side specified area (in units of cylinders, files, etc.) to another computer, and It was necessary to wait for a response that it was established, perform a copy operation, and after copying, contact each other for area lock such as removing write protection of the specified area and moving to the next area. Alternatively, in order to eliminate this inter-computer communication, the slave disk side of the disk controller has an area lock function, online write to the locked area is returned as a lock error, and a write request is sent again from the computer after waiting for the lock area release interrupt. A lock error reporting method of issuing "." Was also considered.

[0003]

Area lock communication may be required for multiple computers, and when a write-protection command for a designated area is issued, each computer has already requested writing to the designated area. If so, the response that the write-protection is established cannot be returned until the completion. Also, in order to reduce the number of communications for this area lock, the size of the area is set larger than the copy size per one time. For these reasons, the communication overhead of the computer and the copy elapsed time increase, which requires a considerably longer time than the copy time in a single computer system. Alternatively, in the lock error reporting method in which the slave disk side of the disk control device has an area lock function, even though it is considerably improved, the number of times the processing routine is started and the number of interrupts are more than normal writing when a lock error occurs, Neither computer load nor elapsed copy time was optimal.

The present invention has been made to solve the above problems, and provides a recorded data online recovery apparatus capable of recovering data at high speed during online operation when a failure occurs in the data in the database. The purpose is to

[0005]

The recorded data online recovery apparatus according to the present invention has the following requirements. (A) a master recording device for recording data, (b) a slave recording device for recording the same data as the data recorded in the master recording device, (c) the same data for the master recording device and the slave recording device Writing means capable of temporarily holding writing to the slave storage device depending on conditions while writing, (d) failure recovery means for recovering a failure of the slave recording device while the writing means is in operation, having the following elements:
(D1) area registration means for registering a copy area of the master recording device for copying data from the master recording device to the slave recording device, (d2) data of the copy area of the master recording device registered by the area registration means Means for copying the data to the slave recording device, and (d3) when writing by the writing device occurs in the copy area of the master recording device registered by the area registration device, the writing to the master recording device is permitted and Writing control means for holding the writing to the slave recording apparatus by the writing means, and executing the held writing to the slave recording apparatus after the copying by the copying means is completed.

[0006]

In the recorded data online recovery device according to the present invention, of the master recording device and the slave recording device,
The master recording device has area registration means for registering a copy area of the master recording device for copying data to be recovered to the slave recording device, and the writing control means for writing to the master recording device, Since all writing is permitted regardless of the copy area registered by the area registration means, it is possible to issue write requests to the master recording device one after another. On the other hand, the writing control means suspends this writing when the writing means writes to the slave recording device in the copy area registered by the area registration means, and executes the suspended writing after the copying by the copying means is completed. Therefore, the problem that the latest data is lost by writing the data to the slave recording device by the writing means before the copying by the copying means is eliminated.

[0007]

1 is a diagram showing an embodiment of a recorded data online recovery device according to the present invention. 10 in the figure
Is a computer, 20 is a disk controller connected to the computer 10, 30 is a master recording device, and 40 is a slave recording device for recording the same data as the data recorded in the master recording device 30. Reference numeral 11 is a writing unit that records data in the master recording device 30 and also records the same data in the slave recording device 40.
Reference numeral 50 denotes a failure recovery means for recovering the data of the slave recording device by copying the data from the master recording device to the slave recording device because the data of the slave recording device has a failure. 21 is a master recording device 3
Area registration means for registering a specific area of 0 as the copy area 31. 12 is area registration means 21
It is a copy means that reads data in the copy area 31 registered by and writes the data in the corresponding area of the slave recording device 40. When the writing unit 11 tries to write data to the copy area 31 registered by the area registration unit, the reference numeral 22 permits the master recording device to write the data and the writing to the slave recording device. It is a write control unit that holds this and permits the writing unit to write data to the slave recording device after the copying by the copying unit is completed.

The operation will be described below with the master recording apparatus referred to as a master disk, the slave recording apparatus referred to as a slave disk, and the copy area referred to as a lock area. The master disk side of the disk controller is provided with an area lock function, and when reading for copying, the computer commands the disk controller to lock the area. The area registration means registers this area. The writing control means permits normal writing to both the master disk and the slave disk when the writing means writes to areas other than this locked area. Also, when writing online to the master disk side locked area where writing to the slave disk side has not been completed and the area lock release command has not come, direct writing is performed as in the unlocked area, and Is detected, and in the case of overlap, the lock area write status is registered in the disk controller for each write and each computer. Then, the lock area write status is notified to the computer when the writing completion report to the master disk is reported.

Here, the lock area write status is a flag for indicating that writing of data from the computer corresponds to the lock area. For example, this lock area write status is indicated by 1 bit,
When it is 0, it indicates that the data to be written does not correspond to the lock area, and when it is 1, it indicates that it corresponds to the lock area. The write control means 22 sets the above-mentioned lock area write status to 1, and notifies the computer together with the data write completion report to the master disk.

On the computer side, the writing means 11 normally issues a write request to the slave disk following the completion of writing to the master disk. However, when the lock area write status is reported from the master disk, the write operation is sent to the slave disk. Write pending.

Next, when the data in the lock area is copied from the master disk to the slave disk by the copy means, the write control means detects the end thereof and locks the computer for which the lock area write status is registered. Outputs area release interrupt. When the lock area release interrupt is transmitted from the disk controller, the writing means collectively cancels the pending write requests to the slave disks and writes the slave disks.

In this way, until the lock area release interrupt is transmitted from the disk controller, by controlling so as to hold without starting writing to the slave disk, new online write data on the slave disk is copied to the old copy. Data can be prevented from being lost. This allows each computer to issue write requests to the master disk one after another without waiting for the lock area to be released, and even if a plurality of them correspond to the write to the lock area, there is no problem. This can be dealt with, and it is not necessary to restart the master disk as in the lock error reporting method described in the conventional example.

According to the above-mentioned method, the disk control device starts the disk as usual, and after the command is found to be writing and the writing to the disk is started, whether the lock area is designated or not Logic for writing judgment, each write request, flag setting for each computer, simple addition processing at the end report, lock area registration, change command, release command and corresponding lock area release processing Therefore, it is an efficient method.

By adopting the method of directly executing the writing on the master disk side as described above, when a writing request is made over both the previous and present lock areas, the computer side is required as in the conventional case. It was restarted after receiving the previous lock area release interrupt, but this time it can be eliminated as a lock error.

As described above, according to this embodiment, in the mirror disk data online recovery device, the data is copied from the master disk to the slave disk during the online failure recovery of the mirror disk system accessed by a plurality of computers. The master disk side of the control unit has a means to register the area (copy target area) instructed by the computer as a lock area, and the writing command from the computer to the master disk is executed as usual, and the writing area and the lock area are Has the means to detect the overlap of the lock area, register the lock area write status in the case of overlap, and report the lock area write status at the end of writing. For computer all that has been registered to the opening command, characterized in that it comprises a disk controller for reporting the lock area open interrupt.

A read & lock function (in addition to the normal read function, a function of releasing the previous lock area and a read target area as a lock area are registered as command commands for the copy target lock area in the disk controller. Function).

According to the read & lock function, the area lock request and the release request which are normally performed are not separately required (except the last release request), and the normal read function can be executed by one command command. In addition to this, since it has the function of releasing the previous lock area and the function of registering the read target area as the lock area, it is possible to reduce the number of inputs and outputs to the disk control device. Also, by limiting the lock range to the necessary minimum, the probability of lock is reduced, and it can be said that this is an efficient method with less overhead.

Next, a specific operation example of this embodiment will be described. FIG. 2 shows an example of application of the present invention to a dual computer system, in which a disk controller, mirror disks (master disks A, C, E are slave disks B, D, F).
Correspond to each other). FIG. 3 is a diagram showing the necessity of the write area lock at the time of recovery processing of the mirror disk, and the lines a and b show the input / output to / from the master and slave disks. Here, the recovery process is a0r-b0w, a1r.
Reading from the master disk and writing to the slave disk are repeated in the order of -b1w, .... xa0w-xb0w, ya0w-yb0w, etc. indicate the mirror mode writing from the online program.
The write xb0w corresponding to xa0w whose writing to the area a0 has already finished before the recovery copy read a0r may be executed independently of the write b0w from the copy program, but the write y after a0r
If yb0w corresponding to a0w precedes writing b0w from the copy program, the data will be lost as if it were not written to the B disc. Therefore, in this embodiment, when the lock area write status is reported for the completion of the online writing to the A disk, the copy writing to the B disk is completed and the area lock release interrupt is reported (area lock request, (When the release command is executed independently), or until the area lock release interrupt is reported by the start of execution of the next area copy reading a1r from the A disk (when the read & lock function is used)
Data loss was prevented by substantially suspending the end processing of the A disk and the mirror mode writing to the B disk. Here, the end interrupt processing of ya0w may be delayed for the period X until the end of the copy writing to the B disk, but in the case of the read & lock method, it means until the next copy reading is started in order to reduce the number of questions and answers with the disk device. Or until the area lock is released): Delayed during section S.

Example 2. Although the first embodiment has described the case of including a plurality of computers, it is also effective in the case of a single computer. Further, in the first embodiment, the case where the two disk control devices are provided on the master side and the slave side is considered, but the control device 1 is focused on improving the reliability of the disk medium / device.
There may be a case where a mirror disk system is composed of individual pieces. In this case, even if writing to the master disk lock area is completed, it will not be reported to the computer immediately, but will wait for the lock area to be released, complete writing to the slave disk, then output an end interrupt and recover. It is conceivable to adopt a method that does not need to be aware of area lock except for the copy processing program. However, in this case, it is necessary to accept a plurality of activation requests for each computer and to plan efficient use of the disk.

Example 3. In the first embodiment, the case where the medium for recording the data is the disk is shown, but the recording medium is not limited to the disk, and any other magnetic recording device or an optical recording device may be used as the recording device for recording the data. Absent.

Example 4. In the first embodiment described above, the case of recovering data from the master disk to the slave disk is shown. However, the master refers to the person who holds the data, and the slave refers to the person who has a failure. It is not limited to a master or slave relationship when data is recorded in a master-slave relationship during normal use of the computer system. That is,
The one holding the data to be recovered is the master, and the one with the failure is the slave.

[0022]

As described above, according to the present invention, it is not necessary to restart the write request from the computer due to the communication between computers and the lock error, and the online recovery copy operation can be speeded up.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a recorded data online recovery device according to the present invention.

FIG. 2 is a diagram showing a specific example of a recorded data online recovery device according to the present invention.

FIG. 3 is a diagram for explaining the operation of the recorded data online recovery device according to the present invention.

[Explanation of reference numerals] 10 computer 11 writing means 12 copying means 20 disk control device 21 area registration means 22 writing control means 30 master recording device (master disk) 40 slave recording device (slave disk) 50 failure recovery means

[Procedure Amendment 2]

[Document name to be corrected] Drawing

[Name of item to be corrected] Figure 1

[Correction method] Change

[Correction content]

[Figure 1]

Claims (1)

[Claims] 1. A recorded data online recovery device having the following elements: (a) a master recording device for recording data; (b) a slave recording device for recording the same data as the data recorded in the master recording device; c) writing means for writing the same data in the master recording device and the slave recording device; (d) failure recovery means for recovering a failure of the slave recording device while the writing means is in operation, having the following elements:
(D1) area registration means for registering a copy area of the master recording device for copying data from the master recording device to the slave recording device, (d2) data of the copy area of the master recording device registered by the area registration means Means for copying the data to the slave recording device, and (d3) when writing by the writing device occurs in the copy area of the master recording device registered by the area registration device, the writing to the master recording device is permitted and Write control means for suspending writing to the slave recording apparatus and causing the slave recording apparatus to execute the suspended writing after the copying by the copying means is completed.