JP3397897B2 - File management method with consistency check function - Google Patents

Description

Detailed Description of the Invention

[0001]

The present invention relates, in the case related to the file management how to store by dividing the data file into a plurality of secondary storage devices, particularly secondary storage device has failed, failed two concerning the data in the file stored in the next storage device highly reliable file management how that can be restored definitely.

[0002]

2. Description of the Related Art In recent years, a central processing unit of a computer
(CPU) performance has improved significantly. However, in order to improve the performance as a computer system, C
It is not enough to improve the performance of the PU, and a technique for realizing high-speed file access is indispensable.

As a file system that enables high-speed file access, for example, Japanese Patent Laid-Open No. 5-250099
There is a system proposed in the issue. In this file system, a file is divided and stored in a plurality of disk devices, and at the time of reading, high-speed file access is realized by simultaneously activating a plurality of disk devices and reading them in parallel. FIG. 11 shows the configuration of this file system.

A conventional file system will be described with reference to FIG. The file is divided into a plurality of subfiles 2001 to 2004 by the file striping program 3100. Subfile 200
1 to 2004 are managed by the subfile management program 3300 using the subfile management tables 4201 to 4204 in which the block addresses on the disk devices storing the subfiles are described. The file is an integrated file management table 410 in which pointers to all sub-file management tables 4201 to 4204 in which the data of the divided files are managed are described.
It is managed by the integrated file management program 3200 using 0.

In the above file system, the user accesses the file by the following method.

First, the integrated file management program 320
0 is the integrated file management table 41 based on the file name passed from the application program 5000.
00, and the divided sub files 2001 to 20
Sub file management tables 4201 to 4 for managing 04
Get a pointer to 204. Next, the integrated file management program 3200 causes the sub-file management table 420.
Pointers 1 to 4204 are passed to the subfile management program 3300.

The subfile management program 3300 is
Based on the pointer passed from the integrated file management program 3200, the sub file management tables 4201 to 4
By referring to 204, the block address of the disk device in which the data is actually stored is obtained, and based on this, the data of the subfiles 2001 to 2004 is accessed.

As a result, the application program 5000 can access the target file without being aware that the file is divided and stored. In addition, the divided subfile 200
Since 1 to 2004 are accessed in parallel, high speed file access is possible.

In the file system disclosed in Japanese Patent Laid-Open No. 5-250099, the date and time of data update and the data length are managed for each subfile.

[0010]

SUMMARY OF THE INVENTION In order to improve the reliability of a file system in which files are divided and stored in a plurality of disk devices, one of the disk devices fails. However, the parity data of the divided file is stored in another disk device so that the data on the failed disk device can be restored from the data of the other disk device. A specific method is disclosed in, for example, Japanese Patent Application No. 5-326823.

However, in order to restore the data stored in the failed disk device, the data on which the parity data is generated and the parity data are correctly associated except for the data in the failed disk device. , It is necessary to calculate their exclusive OR.

For example, A, B, C and D are four data blocks stored in each disk device, and P is parity data created from data blocks A to D. Here, when the data block B cannot be read due to the failure of the disk device, in order to restore the data block B, the exclusive OR of the data blocks A, C, D and the parity data P is calculated. There must be. At this time, the correct data block B cannot be restored no matter which data block is missing or if the data is exclusive ORed with another data block. in this way,
In the conventional technique, even if the wrong data is restored, it may remain as it is.

That is, when the user replaces a failed disk device and tries to recover the data, if the configuration of the disk device that stores the data in a divided manner is set incorrectly, or a normal disk device should be replaced. If you accidentally replace the failed disk unit, you will have a serious situation. Wrong data will be restored as a result of the recovery attempt to restore the normal state.

Further, in the conventional file management method (Japanese Patent Laid-Open No. 5-250099) which manages the update date and time and the data length of each subfile, the data stored in the failed disk device has the latest update date and time. If
There was a problem that the update date and time could not be restored and the data length of the failed disk could not be restored correctly.

For example, it is assumed that the data blocks of A, B, and C each store valid data in a full block and the data length is 8 KB, and the data block of D has 8 data blocks.
It is assumed that only 1 KB of valid data is stored for a data block having a capacity of KB. At this time, the parity data block P stores valid data having a data length of 8 KB in order to obtain the exclusive OR of the data A to D. Here, if the D data block cannot be read due to a failure, the information of the effective data length in the original D data block is recorded only in the failed disk that stores the D data block. The information that the effective data length of the original D data block is 1 KB cannot be restored.

Furthermore, when the faulty disk device is divided into a plurality of partitions and the address space of a large capacity disk is divided and used, the partition information is recorded in each disk device. Therefore, even if the failed disk device is replaced with a new disk device, there is a problem that the original partition cannot be re-divided.

As described above, the first problem to be solved by the present invention is that erroneous data is prevented from being restored and the restored data is the same as that stored in the failed disk device. file management how to ensure is that the subjecting Hisage <br/>. The second challenge is to provide a file management how to be restored properly update date and data length of the sub-file as well as data. Further third object of the present invention, including the partition information of the failed disk device, is to provide a file management how to directly restore all the information about the failed disk device to a new disk device.

[0018] An object of the present invention is to solve the above problems, Oite the method of managing the file to be divided and stored in a data file into a plurality of secondary storage devices,
Even if the secondary storage device has failed to provide a fast reliable file management how it can restore the failed secondary storage information stored in the device correctly.

[0019]

A file management method according to the present invention divides data of a first file into a plurality of second data, and divides the divided second data into a plurality of secondary storage devices. Respectively, and manages the second data as a subfile for each of the secondary storage devices, and takes the exclusive OR of the data at the same displacement from the respective head positions of the plurality of subfiles, in the file management method for implementing a highly reliable file system by storing in a separate secondary storage device and this as parity data storing the second data secondary storage device, the
To each of the plurality of secondary storage devices, as the file system information is divided information to confirm the structure of these secondary storage device, and creates a file system processor
Server identification number, creation date and time, secondary storage device identification number, data
Indicates the logical order of each secondary storage device that stores the data separately.
Striping number, number of components, and secondary storage device
If any one of the plurality of secondary storage devices fails, (i) the file system information other than the failed secondary storage device is written.
Reading from normal respective secondary storage device, (ii) the respective secondary
Process of file system information read from storage device
Identification number, creation date and time, number of components, and secondary failure
Storage device striping numbers match, and the failure
After confirming that all striping numbers except for the secondary storage device are complete, restore the data of the failed secondary storage device, and when the restoration of the data is completed,
Failure of file system information in secondary storage device Secondary storage device
The striping number of the device is initialized .

[0020]

[0021]

Further, according to the present invention, for each subfile, the length of the data of the subfile, the date and time of the data update, and the storage order of the divided second data in the secondary storage device are stored in the file system management information In the embodiment described later, it is called as attribute information) in each secondary storage device storing each of the subfiles, and the data length of the first file and the parity data length of the parity file are recorded. Now, the storage order of the divided second data in the secondary storage device and the update date and time of the parity data are recorded in the secondary storage device storing the parity data as file system management information, and the plurality of the plurality of data are stored. When any one of the secondary storage devices fails, the above file system management information is read from the normal secondary storage device and written in it. It reads data and the parity data from subfiles normal secondary storage device according paid order, by taking their exclusive, to restore the data subfiles lost. Further, using the update date and time of the parity data as the update date and time of the restored subfile, a value obtained by subtracting the sum of the data lengths of the subfiles on the normal secondary storage device from the first data length. As the data length of the restored subfile, the file system management information of the restored subfile is generated and stored in a new normal secondary storage device.

The entire file length (data length of the first file) is used to accurately restore the subfile data length of the failed secondary storage device. For example, when 25 KB data is divided and stored in four secondary storage devices as 8 KB data blocks A to D, the A, B, and C data blocks are filled with valid data. However, even if only 1 KB of the 8 KB data block stores valid data in the D data block, the parity data block P contains blocks A to A.
As a result of taking the exclusive OR of the data of D, 8 KB of effective data is stored. Here, if the data block of D cannot be read due to a failure, the original D
Since the effective data length in the data block of No. 1 is recorded only in the faulty disk device storing the data block of D in the conventional system, the original information of the effective data length of 1 KB of D cannot be restored. Therefore, in the present invention, information that the entire length of the file is 25 KB is recorded in the secondary storage device storing the parity data block, and when this is used to restore the D data block, the other three blocks are recorded. Data length is 24 KB (8 KB x 3)
Therefore, the effective data length is 1 KB (25 KB-2
4KB) and can be restored.

Further, in the present invention, partitioning information of other secondary storage devices is stored in each secondary storage device, and when any one of the plurality of secondary storage devices fails, the failure occurs. After reading the above partitioning information from a normal secondary storage device in which the partitioning information of the secondary storage device is stored and partitioning a new secondary storage device according to this partitioning information, Make sure to restore the data.

The partition division information refers to format information regarding the secondary storage device. According to the present invention, each secondary storage device stores partitioning information of another secondary storage device. For example, the partition division information of an arbitrary secondary storage device is stored in the secondary storage device of the device number immediately before that. As a result, even if one of the secondary storage devices fails and the format information is no longer known, the new secondary storage device that has been replaced has failed by referring to the partition information from another normal secondary storage device. Can be initialized to the same format as.

When the data of the first file is divided and stored in each secondary storage device, the second data is sequentially stored from the secondary storage device having the largest free space at the time of storage. Let go

[0027]

[Action] and C PU identification number that created the file system, creation date and time, the secondary storage device identification number, the data division rating
Stripe indicating the logical order of each secondary storage device
Number, configured number, and failed secondary storage device strike
File system information such as a ping number is stored in each secondary storage device, and when the data is restored, the C
The PU identification number, the creation date and time, the number of constituents , and the striping number of the failed secondary storage device are equal among the normal secondary storage devices, and all normal secondary storage devices except the failed secondary storage device. By confirming that all the striping numbers are correct, even if the configuration of the secondary storage device is mistakenly changed when the failed secondary storage device is replaced, it is possible to recover the data by mistake. It is possible to inform the user that correct data restoration cannot be performed.

Further, by storing the partition information in another secondary storage device, the partitioning information of the failed secondary storage device can be read from the normal secondary storage device, and the partition information of the failed secondary storage device can be read. A new secondary storage device can be formatted with the same conditions.

Further, by storing the entire length of the first file as file system management information (parity data attribute information) in the secondary storage device storing the parity data, it is possible to store the normal divided data. Even if the secondary storage device fails, by subtracting the capacity of the subfile on the normal secondary storage device from the total file length stored in this file system management information, the subfile stored in the failed secondary storage device The file data length can be accurately restored.

Also, when updating any of the data stored separately, the parity data is always updated, so the update date and time of the parity data is always the latest update date and time of the data. Therefore, by recording the update date and time in the attribute information of the parity data and setting the update date and time of the restored subfile to be the same as the update date and time of the parity data, the latest updated date and time is given to the restored data. It becomes possible.

As described above, the data in the failed secondary storage device can be accurately restored in the new secondary storage device. Further, even if the user mistakenly changes the configuration of the secondary storage device, it is possible to prevent the data from being restored in an incorrect form and notify the user of the error.

[0032]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 is a diagram showing the overall configuration of this embodiment. In this embodiment, the disk devices 1001-100
5, a memory 6100, a CPU 6200, an I / O bus interface 6300, a system bus 6400, an I / O bus 6500, a keyboard 6600, a CRT display 6700, and a network interface 6800.

In the memory 6100, a file striping program 3101, an integrated file management program 3201, a sub file management program 3301,
Parity data generation program 3401, parity file management program 3501, data recovery program 3
601, the integrated file management table 4100, the sub file management tables 4201 to 4204, and the parity file management table 4300 are loaded.

File Striping Program 31
01 is a file with a plurality of data 2001 to 20
This is a program that is divided into 04. The subfile management tables 4201 to 4204 are tables for managing the block addresses of the disk devices that store the data F1-1 to F1-4 (2001 to 2004) of the divided files for each data. The subfile management program 3301 stores and manages the data of files divided according to these tables written by the file striping program 3101 as subfiles.

Parity data generation program 3401
Calculates the exclusive OR for each data having the same displacement from the head position of each sub-file, and the parity data F1-
P (2010) is generated. The parity file management table 4300 is a table for managing the block address of the block storing the parity data 2010. Parity file management program 3501
Is the parity file management table 43 written by the file striping program 3101.
00, the parity data 2010 is stored and managed as a parity file.

The integrated file management table 4100 is a table for managing pointers to the sub file management tables 4201 to 4204 and the parity file management table 4300. Integrated file management program 3201
Manages each subfile and parity file using the integrated file management table 4100.

Further, each disk device 1001 to 1005
File system information (FSinfo) 2501 to 250 by the integrated file management program 3201.
5 is stored, and if necessary, data recovery is performed after checking the consistency of the file system information of each disk device by the data recovery program 3601 on the memory 6100.

FIG. 1 shows an example in which two files F1 and F2 are divided and stored in four sub-files. That is, the data of the first file F1 is F1.
-1 to F1-4 are divided and stored, and parity data F1-
P is stored in a disk device 1005 different from the data. The data of the second file F2 is F2-1 to F2-1.
It is divided into F2-4 and, together with the parity data F2-P,
It is stored in each of the disk devices 1001 to 1005.
These data include sub-file management tables 4201 to 4204, a parity file management table 4300, and an integrated file management table 4100 for each file.
Managed by.

The parity data is distributed by the file striping program 3101 to the respective disk devices 1001 to 1005 in file units and stored by the parity file management program 3501.
For example, the parity data F1-P of the file F1 is stored in the disk device 1005 and the parity data F1-P of the file F2 is stored.
2-P is stored in the disk device 1001, and so on.

In this embodiment, the storage system for dividing and managing files in this way is called a file system.

FIG. 2 shows each disk device 1001-100.
File system information stored in 5 (FSinfo)
The contents of 2501-2505 are shown. As shown in the figure,
The file system information (FSinfo) stores the CPU identification number that created the file system, the creation date and time, the identification number of each disk device, the name of the file system, the striping number, the number of components, and the striping number of the failed disk device. It

Here, the striping number is a number which is logically given to the disk device for dividing and storing the data. That is, in the example shown in FIG. 1, the number of disk units is five, the striping numbers are sequentially assigned from the disk device 1001 as 1, 2, and 3, and the disk device 1005.
Has a striping number of 5. Therefore, the file system information 25 stored in the disk device 1001
The striping number of 01 is 1, the disk device 1002
The striping number of the file system information 2502 stored in the disk device 1002 is 2, ..., The striping number of the file system information 2505 stored in the disk device 1005 is 5, and each disk device has a striping number of You can know if you hit the device.

In the failed disk device number, the striping number of the disk device in which an error occurred when reading or writing data is written. In Figure 2,
Since the failed disk device number is 2, disk device 1
002 has a failure. In the normal state where there is no failed disk device, the failed disk device number is 0.
And

The creation date and time is recorded up to the hour, minute and second level as shown in the figure. The disc identification number is a number unique to each disc device. The file system name is an identifier managed by the system, and in one file system, the same file system name is stored in the file system information of each disk device.

As described above, by recording the CPU identification number of the file system and the creation date and time in each disk device, it is possible to guarantee that each disk device stores one file separately. . That is, since the system that has constructed the file system is specified by the CPU identification number, and when it was created by the creation date and time, if these pieces of information match, those disk devices are identified. It can be considered that the data of the file is partly stored and divided.

Further, by describing the number of constituents of the disk devices divided and stored and the striping number, it is possible to confirm whether or not all the disk devices divided and stored are prepared.

FIG. 3 shows the attribute values of each subfile and parity file of the divided and stored file. The sub file and the parity file are respectively stored in separate disk devices together with the attribute value data shown in FIG.

The attribute values of each subfile and parity file are the length of each file, the update date and time,
The striping start disk device number is stored (the storage position is not shown). In particular, the attribute value of the parity file also describes the total length of the file before division.

In the subfile, file data is divided and stored in units of a specific length (hereinafter referred to as a block length). Therefore, as shown in the figure, the length of each file may have different values. Similarly, the update date and time may be different for each file. This occurs when the update data is not divided and stored in all the disk devices because the update data is small.

The striping start disk device number is
The striping number for starting the divided storage of data is shown. Assuming that the subfile number (the number indicating the subfile order) and the striping number match,
To be more specific with the example in FIG. 3, since the striping start number is 3, data is stored in the order of subfile 3 to subfile 4, subfile 1, and subfile 2 in block length units.

FIG. 4 shows a storage state of the disk devices 1 to 5 thus divided and stored. In the figure, a square box indicates a state in which data is stored in block length units, and this data unit will be referred to as a block hereinafter. In this embodiment, the block length is 8 KB.

Disk device y (y = 1 to 1 in this embodiment)
The y in the description (integer of 5) indicates the striping number. As in the example shown in FIG. 3, FIG. 4 also shows an example in which the data is divided and stored in order from the disk device 3. In FIG. 4, Dx indicates the order of blocks of data. Hereinafter, x represents an arbitrary integer value. For example, D643 is 64 from the beginning of the file.
The third block is shown. When converted into a byte position, the block D643 is composed of 8,192B data stored from a position of 8KB (8,192B) × 642 = 5,259,264B from the beginning of the file.

Further, Rx represents the striping sequence number of each block. Here, the striping sequence is a sequence of horizontally arranged blocks as shown in the figure, and is a group of four blocks used for creating parity data.

Px represents the parity data corresponding to the striping sequence Rx. For example, P161 is R
Four blocks D641, D642, D643, D6 forming the 161 or 161st striping sequence
The parity data of the result of having calculated the exclusive OR of the data of 44 blocks is shown.

In the example of FIGS. 3 and 4, the total file length is 5,289,486B, so that a total of 5,289,486B / 8,192B = 645.689 blocks are required to store data. Is. That is, a block for packing 645 8 KB data and a block for storing less than 8 KB data are required. The block of D646 in FIG. 4 is a block for storing the data of less than 8 KB, and the effective data length is 5,289,486B− (8,192B × 645) = 5,6.
46B.

Therefore, the file length of the subfiles stored in the disk device 1 and the disk device 2 is 8,1.
92B × 161 = 1,318,912B, and the file length of the subfile stored in the disk device 3 is 8,
192B × 162 = 1,327,104B, and the file length of the subfile stored in the disk device 4 is
8,192B × 161 + 5,646B = 1,324,558
It becomes B.

In the conventional file system, the file length of the entire file is managed by the file length of each subfile. Therefore, in order to obtain the file length of the entire file, the sum of the file lengths of the subfiles is calculated, 318,912B + 1,318,912B + 1,327,
It is necessary to calculate as 104B + 1,324,558B = 5,289,486B. Therefore, if the disk device fails and the data length of the subfile cannot be read,
The correct data length of the whole file cannot be restored.

For example, in FIG. 4, when the disk device 4 fails, the capacity of the subfile stored in the disk device 4 is different from the file of any other disk and can be restored by the conventional file system. It is impossible.

On the other hand, in the file system of the present embodiment, the disk device for storing the parity file has the structure shown in FIG.
The attribute value of the parity file shown in is stored, and the entire length of the file is described in the attribute value of the parity file. Therefore, when any one of the disk devices fails, the file length of the subfile stored in the failed disk can be restored correctly. Further, if the disk device storing the parity file fails, the total length of the files before division can be calculated by obtaining the sum of the file lengths of the subfiles.

The file update date and time may be different for each sub-file, as shown in FIG. FIG. 3 shows an example in which data of less than the block length is added to the last part of the file, that is, subfile 4.

In the conventional file system, the file update date and time is managed for each subfile. Therefore, when data less than the block length is added to the subfile 4 which is the last part of the file, only the subfile 4 is added. Is updated and the other subfiles 1 to 3 are not updated, the file update dates and times of the subfiles 1 to 3 are left old, and only the file update date and time of the subfile 4 is updated. In such a state, if the disk device 4 storing the latest subfile fails, simply applying the update date / time of another subfile, the latest update date / time of the restored file is October 1993. It becomes 22 days, and a value older than December 3, 1993, which is the actual update date and time, is added.

In order to solve this problem, in the file system of this embodiment, the update date and time is also added to the parity file as an attribute value, and when the file is restored, the update date and time of the parity file is set as the update date and time of the restored file. To do. Further, when the disk device storing the parity file fails, the latest update date / time of the subfiles stored in another disk device may be set as the update date / time of the parity file. By doing so, it becomes possible to add the latest update date and time as an attribute value of the file without returning the file update date and time to the old direction even after the data is restored.

FIG. 5 shows an example in which the storage state of the files shown in FIG. 4 is expanded and a plurality of files are stored. In the figure, Ax, Bx, and Cx indicate blocks of data of different files. That is, this figure shows an example in which three separate files are distributed and stored. APx, BPx, CPx represent blocks storing the parity data of each file.

In the example of this figure, the file A stores the data by dividing the data sequentially from the disk device 1, but the files B and C store the data from the disk device 3 and the disk device 4, respectively. By doing this,
Data can be evenly stored in each disk device. That is, in the present embodiment, when a new file is stored, the disk device with the largest free space at present is set as the striping start disk device to improve the usage efficiency of the disk device.

As described above, when the striping start disk device is changed for each file, it is necessary to store from which disk device the data is stored. In this embodiment, as shown in FIG. 3, by storing the striping start disk device number as the attribute value of each subfile, no matter which disk device fails, the data structure, that is, the data block and the parity data are stored. The order relation of the blocks can be determined, and the data can be accurately restored.

Further, in this embodiment, as shown in FIG.
The logical volume of the disk device can be divided into a plurality of pieces for use. This logical volume is called a partition. In the example of FIG. 6, five disk devices are divided into three partitions, and seven files A to G are stored.

The sizes of these partitions are set when the disk device is initialized. Therefore, when the data of the failed disk device is restored, information on the partitioning of the disk device is required.

FIG. 7 shows partition information indicating the status of partition division in each disk device.
In this embodiment, the partition information shown in this figure is stored for each disk device in the disk device with the striping number immediately before it. That is, the partition information of the disk device 1 is stored in the disk device 5, the partition information of the disk device 2 is stored in the disk device 1, the partition information of the disk device 3 is stored in the disk device 2, and so on.

By doing so, even if a certain disk device fails, the partition information stored in the disk device with the striping number immediately before the failed disk is read and the partition state is accurately restored. be able to.

In the example of FIG. 7, the state of partition division is described in the number of divisions, and the offset and size of each divided partition in byte units. For example, the disk device 1 has three partition divisions, and the first partition is 134,217,728B from the 24,576Bth position from the beginning of the data of the disk device.
It shows that the minutes are secured.

Hereinafter, how data is input and output in this embodiment will be specifically described.
The specific processing procedure is the file striping program 3101, integrated file management program 3201, subfile management program 3301, parity data generation program 3401, parity file management program 3501, and data recovery program 3 in FIG.
601 is executed.

First, the integrated file management program 320
In step 1, each disk device is initialized. In this initialization processing, the partition information of each disk device is stored in the disk device having the previous striping number. Further, each disk stores file system information (FIG. 2) in which the CPU identification number and the creation date and time are recorded.

At this time, the striping number is appropriately set according to the number of disk devices constituting the file system. In the present embodiment, since the file system is composed of five disk devices as described above, striping numbers 1 to 5 are set for each disk device. The failed disk number is initialized to 0. The striped number of the failed disk will be written in the failed disk number when an error occurs later when reading or writing data and a failure is detected.

The disk initialization process described above may be performed once before the use of the disk device is started.

FIG. 8 is a PAD diagram showing the procedure of a file write operation. First, the file striping program 3101 determines a striping start disk device. Here, the free space of each disk device is measured, and the disk device with the largest free space is set as the striping start disk device.

Next, the data in the file is divided into blocks. Then, the subfile management program 33
01 sequentially stores the divided blocks from the striping start disk device according to the striping number.

At this time, after writing data in a block which is one less than the number of constituent disks, that is, after writing data for one column, parity data is written in the next last disk device. Specifically, the parity data generation program 3401 calculates the exclusive OR of the data written up to that point, that is, the parity data for one column, and writes this to the disk device by the parity file management program 3501. .

This write operation will be described in detail with reference to the file C in FIG. In the state before writing the file C in FIG. 6, files A and B are stored in the partition 1.
Is stored in partition 1 of each disk unit.
Comparing the number of blocks written in the disk drive 4, the disk device 4 has the smallest number of three blocks. Therefore, since the free space is the largest, the disk device 4 is set as the striping start disk device.

Since the number of disks is 5, the parity data is calculated for one row of data blocks when up to 4 blocks are stored, and this is stored in the next disk device. That is, after writing the data in the blocks C1 to C4, the exclusive OR of the data in these blocks is calculated, and this is used as parity data in the CP.
Write to 1.

This series of operations is performed for all data blocks of the file. After writing the last data string, in the example of FIG. 6, the data of blocks C5 and C6, the exclusive OR of the data of the two written blocks is calculated,
This is written in the block CP2 as parity data.

Finally, as the attribute value of each subfile,
The striping start disk number, the update date and time, and the file length of the sub file are written (not shown). In the parity file that stores the parity data, the entire length of the file is recorded in addition to the above attribute values.

FIG. 9 shows the procedure of file data read processing.

First, the subfile management program 330
1 opens each striped subfile and reads the attribute information and file system information of each subfile. Then, the striping start disk device number and the failed disk device number are obtained.

When the failed disk device number is other than 0, that is, when there is a failed disk device, the data recovery program 3601 reads the data of the normal disk device that has not failed, including the parity data, and reads the failed disk device. Restore the data. Data restoration of the failed disk device is performed by calculating the exclusive OR of the data for one column.

For example, in the example of FIG. 6, it is assumed that the disk device 4 has failed when the data of the block C1 of the file C is desired to be read. In this case, C2, C3, C4 and C
The exclusive OR of the block of P1 is calculated to restore the data of the block of C1. The data of the block of C5 is
The exclusive OR of C6 and CP2 is calculated and restored.

When the failed disk device number is 0, that is, when there is no failed disk device, the subfile management program 3301 reads the data of a normal subfile other than the subfile storing the parity data.

Finally, the integrated file management program 32
At 01, the data stored separately in the respective disk devices are integrated, and the process ends. That is, the file C in FIG.
In the above example, the data of C1 to C6 are combined into one in order, so that the data can be returned to the same single file format as the data stored in the disk device.

Next, with reference to FIG. 10, a procedure of a state recovery process for replacing a failed disk device and returning it to a normal state will be described. This processing is executed when the failed disk device is replaced or when a spare disk device for system recovery is used.

First, the data recovery program 3601
Read the file system information from each disk device,
Check the configuration of the disk device. This configuration check first confirms that the CPU identification number of each file system information, the creation date and time, the file system name, the number of disk device configurations, and the failed disk device number are the same, and that all striping numbers except the failed disk device are the same. This is done by confirming that all the items are available.

By this configuration check, after confirming that all the disk devices for storing the divided data except for the failed disk device are prepared, the data is stored in the disk device of the striping number immediately before the failed disk device. Initialize the new disk unit that has been replaced according to the partition information. In this initialization processing, the file system information is copied from another disk device to a new disk device, and the striping number is set as the failed disk device number.

Next, data is restored for all subfiles stored in the normal disk device. This restoration process will be specifically described with reference to FIG. 6 by taking the case where the disk device 4 fails as an example.

First, the block A4 of the file A is restored. This can be done by calculating the exclusive OR of the blocks A1, A2, A3 and AP1. In addition, the disk device 5
The size of the subfile stored in the disk device 4 can be calculated by reading out the attribute value of the parity file stored in the file and referring to the entire length of the file. In the example of FIG. Since it can be seen that the size of the subfile is 0, the block next to the block of A4 need not be restored in the subfile of the disk device 4. Similarly, B2, B6, C1, C5
Each subfile is restored while creating the data of the block.

At this time, the update date and time of the parity file of each file is set as the update date and time of the attribute value of each created sub-file. The data length of the sub file is calculated by the entire file length described as the attribute value of the parity file. That is, the size of the subfile to be restored can be obtained by subtracting the total size of the subfiles stored in the normal disk device from the entire file length. When the parity file is restored, the update date and time is set to the latest value among the subfiles, the total file length of each subfile is calculated to obtain the total file length, and this is stored as the attribute value.

Referring to FIG. 10 again, when the above-mentioned restoration processing is normally completed for all files, the failed disk device number of the file system information of each disk device is set to 0, and it is returned to the normal state. Mark and end the process. If the restoration process ends abnormally or if the first disk device configuration check fails to obtain all the disk devices required for recovery, the user is informed that the recovery process cannot be performed, and the process ends. In particular, when an abnormality is detected in the disk device configuration check, the user is warned that the configuration of the disk device is incorrect.

The data input / output processing and the status recovery processing have been described above. According to the present embodiment, after the disk device has failed, the disk device configuration is first checked to see if the data can be recovered correctly. By confirming, even if the disk device is mistakenly replaced, it is possible to warn the user of that fact and avoid erroneous recovery processing, and to provide a more reliable file system. Become.

Further, as an extension of this embodiment, it may be possible to confirm that the file system name is the same in all normal disks when the configuration is checked by the file system information.

Further, in the present embodiment, the configuration check of the disk device is performed only in the state recovery processing, but it is also possible to perform the configuration check when reading or writing the data. This makes it possible to realize a more reliable file management method that guarantees that the data is always correct when inputting and outputting the data.

[0099]

According to the present invention, in a file system in which a file is divided and stored in a plurality of secondary storage devices, the secondary storage device fails, and a secondary storage device that fails for system restoration is detected. Even if you accidentally change the configuration of the secondary storage device when replacing it with a normal secondary storage device, you can inform the user that correct data restoration cannot be done, and you can restore the data by mistake. Can be prevented. Therefore, it is guaranteed that the restored data is the same as that stored in the failed disk device.

Even when the file length and the update date and time are managed for each subfile, the subfile can be restored with the correct file length and the latest update date and time.

Furthermore, since the partition information is stored in each secondary storage device, a new secondary storage device can be formatted in the same state as the failed secondary storage device.

Further, when the file data is divided and stored, the secondary storage device having the largest free space at the time of storage is sequentially stored, so that the use efficiency of the secondary storage device is improved. Can be increased.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing the contents of file system information.

FIG. 3 is an explanatory diagram showing contents of attribute information of a sub file and a parity file.

FIG. 4 is a schematic diagram showing an arrangement of blocks for storing data on a disk device.

FIG. 5 is a schematic diagram showing an arrangement on a disk device when storing data of a plurality of files.

FIG. 6 is a schematic diagram showing an arrangement on a disk device when storing files in a plurality of partitions.

FIG. 7 is an explanatory diagram showing the contents of partition information.

FIG. 8 is a PAD diagram showing a file writing procedure.

FIG. 9 is a PAD diagram showing a file reading procedure.

FIG. 10 is a PAD diagram showing a procedure of a recovery process when a failed disk is replaced and data is restored.

FIG. 11 is a diagram showing a configuration of a conventional example.

[Explanation of symbols]

6100 ... Memory, 6200 ... CPU, 6300 ... I /
O bus interface, 1001 to 1005 ... Magnetic disk device.

─────────────────────────────────────────────────── ─── Continued front page (72) Inventor Hiroshi Yamashita 1099, Ozenji, Aso-ku, Kawasaki, Kanagawa Hitachi, Ltd. System Development Laboratory (72) Inventor Hideo Takahashi 1099, Ozenji, Aso-ku, Kawasaki, Kanagawa Hitachi Systems Development Laboratory (72) Inventor Kanji Kato 1099 Ozenji, Aso-ku, Kawasaki-shi, Kanagawa Hitachi System Development Laboratory (72) Inventor Akira Kitou 5030 Totsuka-cho, Totsuka-ku, Yokohama-shi, Kanagawa Company Hitachi, Ltd. Software Development Division (72) Inventor Hidenori Yamada 1 Horiyamashita, Hadano City, Kanagawa Prefecture Hitachi Computa Engineering Co., Ltd. (72) Toshiyuki Maki 1 Horiyamashita, Hadano City, Kanagawa Hitachi Computer Engineering Co., Ltd. Meeting (72) Inventor Hiroshi Asami 1 Horiyamashita, Hadano City, Kanagawa Prefecture Hitachi Computa Engineering Co., Ltd. (72) Inventor Akiko Takara 1 Horiyamashita, Hadano City, Kanagawa Hitachi Computer Engineering Co., Ltd. (56) References JP-A-6-161672 (JP, A) JP-A-6-110766 (JP, A) JP-A-5-108443 (JP, A) JP-A-3-182944 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G06F 3/06-3/08 G06F 12/00 G06F 12/16

Claims (4)

(57) [Claims] 1. Data of a first file is divided into a plurality of second data, and the divided second data is stored in a plurality of secondary storage devices, respectively, and each of the secondary storage devices is stored. The second data is managed as a subfile, the exclusive OR of the data at the same displacement is taken from the respective head positions of the plurality of subfiles, and the second data is stored as parity data. in the file management method for implementing a highly reliable file system by storing in a separate secondary storage device and the secondary storage device, each of the plurality of secondary storage devices, the structure of their secondary storage device As the file system information that is the division information for checking , the file system
Identification number of the essa, creation date and time, secondary storage device identification number,
Set the logical order of each secondary storage device that stores data
Striping number to be shown, number of constituents, and failure secondary storage
The step of writing the striping number of the device, and when any one of the plurality of secondary storage devices fails, (i) the file system information is stored in the failed secondary storage device.
Reading from normal respective secondary storage device outside the ring is set to other than, (ii) file system read from the respective secondary storage device
Identification number of the processor of the system information, creation date and time, number of components,
And the striping numbers of the failed secondary storage devices match,
And the striping pin excluding the defective secondary storage device
After confirming that all the storage numbers are complete, restore the data in the failed secondary storage device, and when the data restoration is complete,
File system information failure Secondary storage striping
And a step of initializing a group number . 2. The data of the first file is divided into a plurality of second data, and the divided second data is stored in each of a plurality of secondary storage devices, and each of the secondary storage devices is stored. The second data is managed as a subfile, the exclusive OR of the data at the same displacement is taken from the respective head positions of the plurality of subfiles to create parity data, and the parity data is used as the parity file. In a file management method for realizing a highly reliable file system by storing the data in a secondary storage device other than the secondary storage device storing the data, for each subfile, the length of the subfile data,
Recording the data update date and time and the storage order of the divided second data in the secondary storage device as file system management information in the respective secondary storage devices storing the respective subfiles; The data length of one file, the parity data length of the parity file, the storage order of the divided second data in the secondary storage device, and the update date and time of the parity data are used as the file system management information. The step of recording the parity data in the secondary storage device, and when any one of the plurality of secondary storage devices fails, the file system management information is read from a normal secondary storage device and Data and parity data are read from the subfile of the normal secondary storage device according to the described storage order, and their exclusive The step of restoring the data of the lost subfile by taking the logical sum, and the update date and time of the parity data as the update date and time of the restored subfile, and the normal secondary storage from the length of the first data The value obtained by subtracting the sum of the data lengths of the subfiles on the device is used as the data length of the restored subfiles above to generate the file system management information of the restored subfiles, and create a new normal secondary storage device. A file management method comprising the step of storing. 3. The file management method according to claim 1 or 2.
By law, each secondary storage device has a partition on the other secondary storage device.
The step of storing the partition information, and when any one of the plurality of secondary storage devices fails
Is the partitioning information of the failed secondary storage device.
The par Tisho but from the stored normal secondary storage device
Read the partition information and use this partition
Therefore, partition the new secondary storage
And then restore the data on the failed secondary storage device.
File management characterized by further equipped with step
Method. 4. The data of the first file is divided into a plurality of second data, and the divided second data is stored in each of a plurality of secondary storage devices, and each of the secondary storage devices is stored. The second data is managed as a subfile, the exclusive OR of the data at the same displacement is taken from the respective head positions of the plurality of subfiles, and the second data is stored as parity data. In a file management method for realizing a highly reliable file system by storing in a secondary storage device different from the secondary storage device, a logical order is assigned to the plurality of secondary storage devices in advance.
In each of the secondary storage devices, the step of storing the partitioning information of the immediately preceding secondary storage device in that order, and one of the plurality of secondary storage devices are stored. when the platform has failed, it reads the partitioning information from the next to become the secondary storage device of the failed secondary storage device in the order,
After the new secondary storage device is partitioned according to this partitioning information, a step of restoring the data of the failed secondary storage device is provided.