JPH11353125A - Data restoring method for radio device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the data restoring method of the RAID device which can restore a replaced magnetic disk without impeding normal access to a magnetic disk drive after faulty magnetic disk replacement. SOLUTION: By the data restoring method for a magnetic disk which generates error correction information from external write information and stores it, and then corrects error information by using the error correction information in the case of error occurrence and outputs it, repairing operation for restoring a replaced magnetic disk by using write information and read information corresponding to an external normal access request (S13, S14, and S16) after the faulty magnetic disk is replaced or writing information which should be stored correctly on the faulty magnetic disk to the replacing magnetic disk is automatically performed (S15) to restore the replaced magnetic disk.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a redundant disk array (hereinafter referred to as a RAID (Redundant Array of Inexpensive Disk)) in which information is divided and stored on a plurality of magnetic disks, and error correction information is created and stored on the magnetic disks.
s)) After replacing the faulty magnetic disk of the device,
The present invention relates to a data recovery method for a RAID device that restores information that should be originally written on a failed magnetic disk to an exchanged magnetic disk (hereinafter referred to as “repair”).

[0002]

2. Description of the Related Art Conventionally, RAID (redundant disk array)
Are classified into the following five types according to the difference in the method of dividing and storing information on a plurality of magnetic disks. The first is RAID 0, which simply divides information on each magnetic disk in sector units and stores it. The second is RAID 1, which duplicates and stores the same information on a plurality of magnetic disks. RAID2 and RA2, which are divided and stored on a disk, create error correction information at that time, and store them on a dedicated magnetic disk
ID3 (classified by a method of generating error correction information, for example, hamming / parity, etc.), and fourthly, information is divided and stored in each magnetic disk in sector units, and error correction information is generated at that time to create a dedicated error correction information. RAID4 and Fifth, which are stored on the magnetic disk, divide the information in units of sectors and cyclically store the information on each magnetic disk, and the error correction information created at this time is the same as other information (no dedicated magnetic disk for error correction information is provided). RAID5 and the like are known which cyclically store data in each magnetic disk in sector units.

[0003] Next, referring to FIG.
That is, RAID0 to RAID5 will be described in more detail. 6A and 6B show configurations of RAID0 to RAID5, where FIG. 6A shows RAID0 and FIG. 6B shows RAID1.
, (C) is a diagram showing RAID2 and RAID3, (D) is a diagram showing RAID4, (E) is a RAID5
FIG.

In FIG. 6A, reference numeral 7 denotes write information transferred from, for example, a host or the like, and 1 denotes an R which merely divides the input information on a magnetic disk basis in sector units and stores it.
AID0. The RAID 0 1 simultaneously operates a plurality of magnetic disks (disks 51 to 54, a magnetic disk as a hard disk device or an HDD, and is also simply referred to as a disk hereinafter) to input information to the disk 51.
, The input information from the outside can be transferred to the disks 51 to 54 at a high speed. However, the storage information of the failed magnetic disk (hereinafter referred to as a failed magnetic disk) can be transferred. Error correction is not possible.

In FIG. 6B, reference numeral 7 denotes write information, 2
Is an RA that duplicately stores the same input information in sector units on a plurality of magnetic disks (disks 55 and 56).
ID1. RAID12 is also called a mirrored disk, and is a highly reliable storage method. However, since the same information is redundantly stored, the use efficiency of the hard disk is poor.

In FIG. 6C, reference numeral 7 denotes write information, 3 denotes input information divided and stored in a plurality of magnetic disks (disks 57 to 60) in bit units, and error correction information of the stored information is stored. Are RAID2 and RAID3 that are created on the disk 61. RAID2 and R
The AID 3 differs from the AID 3, for example, in the method of creating error correction information for a faulty magnetic disk, such as the parity method and the Hamming method.

In FIG. 6D, reference numeral 7 denotes write information, 4 denotes input information which is divided and stored in a plurality of magnetic disks (disks 62 to 65) in sector units, and storage information for a failed magnetic disk is stored. RAID4 in which the error correction information is created and stored on the disk 66.

In FIG. 6 (E), 7 is write information, 6
Is information written on each of the disks 67 to 71, P is error correction information, 5 is divided into a plurality of magnetic disks (disks 67 to 71) in units of sectors and cyclically stores the information, The error correction information of the storage information for the magnetic disk is created and each disk 6
7 to RAID 5 which are stored in the disk 71 in a circular manner.

The above described RAID 0 1 to RAID 5
Among RAID 5, RAID 1 and 2 can use the information of the other disk even if one disk fails,
Further, even if a failure occurs in any one of the disks 57 to 71, the RAID2 3 to RAID5 5 use the error correction information to create information on the failed disk,
The information can be transmitted to an external device (for example, a host or the like). However, RAID 0 1 is capable of high-speed information transfer, but cannot read information if any one of the disks 51 to 54 fails.

As is clear from the above description, RAID
2, RAID3, RAID4, and RAID5 can read information even if one magnetic disk fails, by creating information on the failed magnetic disk based on the error correction information. However, if a read error happens to occur, it can be sufficiently recovered in this way.However, if the read error increases or becomes unreadable for some reason such as deterioration of the magnetic disk, The magnetic disk must be replaced.

[0011]

However, as described above, there is a problem that after replacing the failed magnetic disk, it is necessary to perform a repair operation for restoring all information stored on the failed magnetic disk. This repair operation is a process of creating information to be recorded on the failed magnetic disk from information stored on the other normal magnetic disks and error correction information, and writing the information to the failed magnetic disk. Usually, the magnetic disk drive cannot be used until the work is completed due to the repair work, and the normal write / read ability of the magnetic disk drive is greatly reduced due to the load of the repair work. there were.

The present invention has been made to solve the above-mentioned conventional problems, and has been made in consideration of RAID2, RAID3, and RAID.
(4) Write information and error correction information are divided and written on a plurality of magnetic disks in RAID5 or the like, and even if a failed magnetic disk occurs, information on the failed magnetic disk is created from the error corrected information and reading can be continued. Provided is a method of restoring data in a RAID device, which can execute a repair operation without deteriorating its ability during a normal write or read operation of a magnetic disk device after replacement of a failed magnetic disk. The purpose is to:

[0013]

According to the data restoration method of the present invention, error correction information is created from externally written information, the written information and the error correction information are divided into a plurality of magnetic disks and written, In the data recovery method for a magnetic disk, the information of the failed magnetic disk is created and read from the error correction information when an error occurs due to a failure of the magnetic disk in response to the read request.
After replacing the failed magnetic disk, the replaced magnetic disk is repaired by using write information or read information (information after error correction) for a normal external write or read request using the repair management table, Or, if there is a normal write or read request from the outside, the priority is given to the request.If there is no normal write or read request from the outside, the information that should be correctly stored in the failed magnetic disk is exchanged with the magnetic disk. The repair operation for writing is automatically executed to repair the replacement magnetic disk.

According to the present invention, the in-use address of the replaced magnetic disk is automatically restored when there is no normal access operation, and the correct information created by the normal access operation is written in parallel with the next read request. By writing to the replaced magnetic disk, a data recovery method for the magnetic disk can be obtained in which the repair operation of the replaced magnetic disk is executed without deteriorating the normal access work.

[0015]

According to a first aspect of the present invention, there is provided a data repair method, wherein error correction information is created from externally written information, and the written information and the error corrected information are divided and written on a plurality of magnetic disks. In the data recovery method for a magnetic disk, when an error occurs due to a magnetic disk failure in response to an external read request, information on the failed magnetic disk is created and read from the error correction information. After that, the method includes a step of restoring the replaced magnetic disk by using write information or read information for an external normal write or read request, and when there is an external normal write or read request, the write information or The system uses the read information to automatically repair the replaced magnetic disk. After the disk is replaced, the read information created for the replaced magnetic disk is automatically written into the replaced magnetic disk in a normal read request, so that the read information particularly for the replaced magnetic disk is created. There is no need to perform the repair, the effect of the repair work on the normal access work is small, and the repair can be performed efficiently.

In the data restoration method according to the present invention, in the step of using the read information, the read information for the exchange magnetic disk created from the read information is written in parallel with the operation of the next generated read request. The writing to the exchange magnetic disk is performed, and the writing of the read information to the exchanged magnetic disk is performed in parallel with the operation of the next read request, thereby preventing the normal read operation. And has the effect that the replaced magnetic disk can be repaired.

According to a third aspect of the present invention, in the data repair method, error correction information is created from externally written information, the write information and the error correction information are divided into a plurality of magnetic disks and written, In a data recovery method for a magnetic disk in which information on a failed magnetic disk is created from error correction information when an error occurs due to a failure of the magnetic disk in response to a read request, the failed magnetic disk is replaced after the failed magnetic disk is replaced. A repair operation for writing information that should be correctly stored in the disk to the replaced magnetic disk, including a step of executing the repair operation when there is no external normal write and read request, and without an external normal write and read request The address of the replaced magnetic disk corresponding to the busy address of the failed magnetic disk. The repaired magnetic disk can be repaired without interrupting the normal access by automatically repairing the replaced magnetic disk when there is no normal external access. It has the effect of being able to.

According to a fourth aspect of the present invention, in the data repair method, the step of automatically repairing the replaced magnetic disk is executed by referring to a repair management table. By using the repair management table for the repair, the address of the replaced magnetic disk that needs repair can be efficiently repaired.

(Embodiment) An embodiment of the present invention will be described below in detail with reference to the accompanying drawings and FIGS. FIG. 1 is a diagram showing an example of the configuration of a magnetic disk drive for implementing a data recovery method according to an embodiment of the present invention.
FIG. 3 is a flowchart showing a flow of a repair operation by the data repair method according to the embodiment of the present invention; FIG. 3 is a diagram showing an example of a repair management table used in the data repair method shown in FIG. 2; FIG. 5 is a diagram showing the timing at which a repair operation is performed by a read command in the data repair method in FIG. 5, and FIG. 5 is a diagram showing the timing at which a repair operation is performed by a write command in the data repair method in the present embodiment.

First, with reference to FIG. 1, an example of the configuration of a magnetic disk drive that implements a data recovery method according to an embodiment of the present invention will be described. In FIG. 1, reference numeral 20 denotes a magnetic disk device, 21 denotes a CPU, 22 denotes a memory, 23 denotes a bus, 24 denotes a VGA, 25 denotes an I / O, 26 denotes an NIC,
Is an SPC, 28 is a CRT, 29 is a keyboard, 30 is a mouse, 31 is a network, 32 is a magnetic disk 1,
2, 3, 4, and 33 are SCSI buses, 34 is a switch 1,
2, 3, 4, and 35 are I / Os. In this example, there are four magnetic disks 1, 2, 3, and 4-32, which are represented by magnetic disks 1-32 through 4-32, respectively. Similarly, the switch 34 has a switch 1-
34 to switch 4-34.

Further, an outline of a detailed configuration and operation of the magnetic disk drive used in the present embodiment will be described with reference to FIG. In FIG. 1, a CPU 21 is a central control device that temporarily stores information in a memory 22. The bus 23 is an information communication path for the CPU 21. Memory 22
The program is written in a part of CPU2.
1. Communication control is executed by the memory 22 and the bus 23. The VGA 24 creates an image to be displayed on the CRT 28. The I / O 25 is an interface for receiving operation inputs from the keyboard 29 and the mouse 30.

The NIC 26 is a network interface card, and is a network 3 as a communication line with the outside.
1 interface. The SPC 27 is an interface of the SCSI bus 33 and executes information writing and reading operations (hereinafter, referred to as access) to and from the magnetic disk 32 via the SCSI bus 33 under the control of the CPU 21. SPC is an abbreviation for a SCSI protocol controller, and the SCSI bus 33 and the CPU bus 23
Interface with CRT 28 is CPU 21
The image or character information is displayed by the instruction of. VGA2
Reference numeral 4 executes information conversion so that the display information specified by the CPU 21 can be displayed on the CRT 28. The keyboard 29 converts the push button operation into an electric signal, sends the push button signal to the CPU 21 via the I / O 25, and
Recognizes this and executes the corresponding processing.

The mouse 30 is used to designate an image with an arrow (pointer) and a button displayed on the CRT 28 and to input the image by pressing (clicking) the button. Like the keyboard, an operation signal is sent to the CPU 21 via the I / O 25, and the CPU 21 executes the processing. Network 3
1 is a communication line with the outside that receives information from the outside and
It notifies CPU 21 via C26. If the received information is a write command to the magnetic disks 1, 2, 3, and 4-32 (hereinafter, represented by the magnetic disk 1-32), C
The PU 21 instructs the transfer of the received information to the memory 22, and then instructs the magnetic disk 1-32 to write. If it is an external information image read command, the CPU 21 reads the designated information from the magnetic disk 1-32, and
7, and is sent to the network 31 via the NIC 26.

The magnetic disk 1-32 executes processing for writing and reading information. That is, in the writing process,
The magnetic disk 1-32 receives and writes information stored in the memory 22 via the SPC 27. At this time, the CPU 21 creates error correction information in another area on the memory 22 based on the information stored in the memory 22 and uses any one of the magnetic disks 32 (in this example, the magnetic disk 4-32). Write to. In response to the read command, the designated data is read from the magnetic disk 1-32 and written to the memory 22. The CPU 21 sends the information read out via the NIC 27 to the network 31 when the writing to the memory 22 is completed.

Next, a case where a faulty magnetic disk exists will be described. Now, assuming that the faulty magnetic disk is the magnetic disk 2-32, the CPU 21 uses the read information and the error correction information from the other than the faulty magnetic disk 2-32 written in the memory 22 to store other areas in the memory 22. Then, read information for the failed magnetic disk 2-32 is created, and this information is sent to the network 31 via the NIC 27 as read information for the failed magnetic disk 2-32. The SCSI bus 33 has a plurality of magnetic disks 1, 2,.
It is a communication line connecting 3, 4-32.

To specify any one of the plurality of magnetic disks, an ID number is set for each of the magnetic disks 1, 2, 3, and 4-32. The ID number is set by hardware using a switch or a jumper pin mounted on the magnetic disk 1-32. The CPU 21 passes through the SPC 27,
It communicates with each magnetic disk 1-32 based on the ID number and confirms the existence of each magnetic disk 1-32. The same ID number must not exist on one SCSI bus 33. if,
If there are two or more magnetic disks 1-32 having the same ID number, when the CPU 21 issues a command or inquiry using the ID numbers, the two magnetic disks 1-32 are used.
Since −32 simultaneously transmits different information, bit collision occurs and a parity error occurs, preventing normal communication.

RAID2, RAID3, RAID4, R
In AID5, a plurality of magnetic disks 1, 2, 3,
One of 4-32 is used for storing error correction information, and stores error correction information. With this error correction information, even if one magnetic disk 2-32 fails, the information on the failed magnetic disk 2-32 can be created by the error correction information, so that the magnetic disk device can be continuously operated. .

Switches 1, 2, 3, and 4-34 (hereinafter, referred to as switches 1-34) are switches for connecting and disconnecting the magnetic disk 1-32 and the SCSI bus 33. This switch is provided from the CPU 21 to the I / O 35
, And when a failed magnetic disk 2-32 occurs, disconnection and connection can be executed. In the example of FIG. 1, four switches 1, 2, 3, 4 for the magnetic disks 1-32 to 4-32 are provided.
Equipped with -34. I / O 35 is switch 1-34
Is an interface for controlling the CPU 21.

Next, with reference to FIG. 2, an operation flow of a repairing operation for the replaced magnetic disk 2-32 will be described. First, the repair work is started from step S10 shown in FIG. As described above, this repair work is performed during a normal access, so that a table indicating whether the repair work has been completed or not completed is required. The area of the memory allocated to manage such repair work is called a repair management table (to be described later with reference to FIG. 3), and is referred to as step S.
At 11, the repaired flag recorded in the repair management table is erased over the entire area of the table (all are 0 in the initial state).

Next, the process proceeds to step S12, where it is determined whether or not there is a normal access. If there is a normal access, the process proceeds to step S13 to perform a normal access process. In step S14, it is confirmed whether the repair work has been completed for the accessed area (whether the area has been repaired). If not repaired (restoration not completed), step S
Then, the process proceeds to step S16, where a restoration process is executed, a restoration completion flag is set for the area, and restoration is recorded in the restoration management table.

After the execution of the restoration process in step S16, the process proceeds to step S17, and the restoration target magnetic disk 2-32.
It is confirmed whether or not the restoration has been completed over the entire area. If the restoration has been completed, the process proceeds to step S18 to end the restoration work. Move on to the next area repair work. If it is determined in step S14 that the accessed area has been restored, the process proceeds to step S17 without doing anything, and as described above, it is determined whether the restoration has been completed for the entire area of the magnetic disk 2-32 to be restored. .

Further, returning to step S12, in the confirmation of the presence or absence of the normal access, if there is no normal access,
In step S15, the restoration work is performed by independently accessing the area where the restoration work is not completed, setting the restoration completed flag for the area, and recording the restoration completion in the restoration management table. Thereafter, the process proceeds to step S17. As described above, if the repair has been completed over the entire area of the magnetic disk 2-32 to be repaired, the process proceeds to step S18 to end the repair work. Check whether there is access processing for the next area, and move on to the repair work for the next area.

Next, an example of the repair management table will be described with reference to FIG. The repair management table is a memory for storing information for executing the repair work secured in a certain area of the memory, and is preferably a nonvolatile memory. Although not shown in FIG.
-32 or may be placed anywhere else. The unit of access (writing, reading) of the magnetic disk 1-32 is defined, and this is called a block size. The position information in units of the block size is called a logical address. Access is performed by designating write or read, a logical address, and the number of blocks. In the case of write, write information is sent thereafter.

In the case of reading, the designated information is read from the magnetic disk 1-32 and sent to the outside that issued the read request. The magnetic disk drive includes a plurality of magnetic disks 1
-32, the external access is treated as one magnetic disk, and inside the magnetic disk device, the write information is divided according to the RAID 2 to 5 system applied there, and error correction is performed based on this information. Information is created and stored on the plurality of magnetic disks 1-32. Inside the magnetic disk device, the CPU 21 manages information such as the usage status of the logical address of each magnetic disk 1-32 and the magnetic disk 4-32 for storing error correction information.

The management information is stored in each magnetic disk 1-
A storage area is secured at the beginning of the address 32 (for example,
Logical addresses 0 to 99), followed by a logical address 10
0 (example) is used for external information storage. The restoration management table shown in FIG. 3 is obtained by adding restoration completion information to each magnetic disk management table of the magnetic disk device.

In FIG. 3, reference numeral 40 denotes a flag indicating that the repair operation is being performed (repair operation of the replaced magnetic disk is being performed). If the bit being repaired is 1, it indicates that the repair operation is being performed. Reference numeral 51 denotes a logical address 0, which is an area used for managing a magnetic disk. Therefore, the use status of these logical addresses is stored to indicate that they are being used, but they are not to be repaired.

Reference numeral 42 denotes a logical address 100. The busy flag indicates bit 1, indicating that external information is stored. 43 is a logical address 37500,
The in-use bit is 0, indicating that the bit is not yet used for storing external information, and is not a target of the repair work.

Next, with reference to FIG. 4, the flow of the process of the repair operation in response to the read command will be described. In FIG. 4, it is assumed that the magnetic disk 2-32 is a magnetic disk to be repaired, and the error correction information is stored on the magnetic disk 4-32. The read command 80 is a read command received from the outside via the network 31, which is a communication means with the outside, and indicates that the magnetic disk device has received the command. Read command 81 to the magnetic disk is read command 8
0 is converted to a storage address (address of the magnetic disk 1-32) inside the magnetic disk device,
32 to output a read command. Since the magnetic disk 1-32 is to be restored, correct information has not been written, and thus no read command has been output.

The read time 82 depends on each magnetic disk 1-3.
2 is the time required for each reading. The information 83 read out and transferred to the memory is read after the elapse of the read time 82.
This indicates that information read from each magnetic disk 1-32 is transferred to the memory 22. Repair target magnetic disk 2
The information 85 created for −32 indicates that correct information of the magnetic disk 2-32 to be repaired is created from the information 83 including the error correction information transferred to the memory 22.

Normally, the error correction information is stored in each magnetic disk 1
The correct information for the magnetic disk 2-32 to be repaired is created by taking the exclusive OR of the block information including the error correction information. As shown in FIG. 4, the transfer information 84 to the network is read out from the network 31 in the order of the magnetic disk 1-32, the magnetic disk 2-32, and the magnetic disk 3-32. Indicates transfer.

On the other hand, the write command 86 indicates that a command to write correct information created for the repair target magnetic disk 2-32 to the repair target magnetic disk 2-32 is issued. The information 87 transferred to −32 restores the write information to the recovery target magnetic disk 2-3.
2 is transferred. The write time 88 is a time required for the write information to be written by the magnetic disk 2-32 to be restored. After the write information is transferred to and stored in the magnetic disk 2-32 to be repaired, the repaired bit of this address in the repair management table is set to 1.

The next read-startable position 89 indicates a position at which, when reading from all the magnetic disks 1-32 is completed, if there is a next reading request, the reading can be started. . In other words, the repair work requires a write time 88 for the repair target magnetic disk 2-32, but during this time, the next read work is possible, and the repair work can be performed concurrently with the read work. The effect of the repair operation on the read operation is very small.

In the repair work by voluntary access,
The operation is the same as the above except that there is no read command 80 from the network 31 and no transfer information 84 to the network 31. In short, the CPU 21 searches the repair management table for an unrepaired address, outputs a read command 81 to the unrepaired address of the repair target magnetic disk 2-32, and transfers the read command 81 to the network 31 without transferring the read command 81 to the network 31.
Perform the same repair work as described above.

Next, with reference to FIG. 5, the flow of the processing of the repair work in response to the write command will be described. In FIG. 5, the magnetic disk device that has received the write command 90 from the network 31 outputs a write command 91 to each magnetic disk 1-32. Thereafter, the CPU 21 stores the write information 94 transferred from the network 31 in the R
The data is divided according to the division procedure determined by the AID, error correction information is created 95, and the divided information and error correction information are written to the respective magnetic disks 1-32. In the case of writing, the magnetic disk 2-32 to be restored
By writing in the same way as normal writing,
The restoration of the address is completed. After the repair is completed, the repaired bit at this address in the repair management table is set to 1.

[0045]

The present invention is constructed as described above,
In magnetic disk devices such as AID2, RAID3, RAID4, and RAID5, after replacing a failed magnetic disk,
Even if the repair work is started, the repair work can be performed while performing the normal access, so that the replaced magnetic disk is repaired without interrupting the normal access and the influence of the repair work on the normal access is small. Can be.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of the configuration of a magnetic disk drive that implements a data restoration method according to an embodiment of the present invention;

FIG. 2 is a flowchart showing a flow of a repair operation by the data repair method according to the embodiment of the present invention;

FIG. 3 is a view showing an example of a repair management table used in the data repair method shown in FIG. 2;

FIG. 4 is a diagram showing a timing of executing a repair operation in response to a read command in the data repair method according to the embodiment;

FIG. 5 is a diagram showing timing for executing a repair operation in response to a write command in the data repair method according to the embodiment;

FIG. 6 shows each configuration of RAID0 to RAID5,
(A) is a diagram showing RAID0, (B) is a diagram showing RAID1, (C) is a diagram showing RAID2 and RAID3,
(D) is a diagram showing RAID4, (E) is a diagram showing RAID5.

[Explanation of symbols]

1 RAID0 2 RAID1 3 RAID2, RAID3 4 RAID4 5 RAID5 6 Information written by cyclic storage S10 Repair start S11 Erase all areas of repaired flag S12 Confirmation of normal access S13 Normal access processing S14 Confirmation of area restoration S15 Independent Executing access restoration processing and setting a restoration flag for each area S16 Executing restoration processing and setting a restoration flag for each area S17 Confirming completion of restoration of all areas S18 Completion of restoration work 20 Magnetic disk device 21 CPU 22 Memory 23 Bus 24 VGA 25 I / O 26 NIC 27 SPC 28 CRT 29 Keyboard 30 Mouse 31 Network 1-32, 2-32, 3-32, 4-32 Magnetic disk 33 SCSI bus 1-34, 2-34, 3-34, 4-34 Switch 35 I / O 40 Repairing flag 41 Information of logical address 0 42 Information of logical address 100 Information of logical address 37500 80 Read command 81 Read command to magnetic disk 82 Read time 83 Information read and transferred to memory 84 Information transferred to the network 85 Information created for the magnetic disk 2-32 86 Write command 87 Information transferred to the magnetic disk 2-32 88 Write time 89 Next read startable position 90 Write command 91 Write to magnetic disk Command 92 Write information to be transferred to magnetic disk 93 Write execution time 94 Write information from network 95 Creation of error correction information

Claims (4)

[Claims] An error correction information is created from externally written information, the write information and the error correction information are divided and written on a plurality of magnetic disks, and an error is caused by a failure of the magnetic disk in response to an external read request. In a data recovery method for a magnetic disk in which information on a failed magnetic disk is created and read out from error correction information when an error occurs, after replacing the failed magnetic disk, write information or a write request for a normal external write or read request is made. Including a step of repairing the replaced magnetic disk using the read information, and automatically repairing the replaced magnetic disk using the write information or the read information when a normal external write or read request is made. A method of restoring data on a magnetic disk, characterized in that: 2. The method according to claim 1, wherein the step of using the read information includes a step of writing the read information for the exchange magnetic disk created from the read information to the exchange magnetic disk in parallel with a next read request. 2. The method according to claim 1, further comprising the steps of: 3. The method according to claim 1, wherein error correction information is created from externally written information, the write information and the error correction information are divided and written on a plurality of magnetic disks, and an error is caused by a failure of the magnetic disk in response to an external read request. In a data recovery method for a magnetic disk in which information on a failed magnetic disk is created and read out from error correction information when an error occurs, information that should be correctly stored on the failed magnetic disk is replaced after the failed magnetic disk is replaced. Performing a repair operation for writing to the failed magnetic disk when there is no normal external write and read request.When there is no normal external write and read request, the in-use address of the failed magnetic disk is The address of the corresponding replaced magnetic disk is automatically repaired. Over data repair method. 4. The method according to claim 1, wherein the step of automatically repairing the replaced magnetic disk is performed by referring to a repair management table.