JPS6120161A - Protection processing method of data set - Google Patents

Abstract

PURPOSE:To detect a failure and to restore easily an data block before updating by providing two control blocks and storing the data block before updating in one control block. CONSTITUTION:The 1st and 2nd control blocks CTL1, CTL2 for controlling data blocks A-C divided into respective groups at updating of these blocks are formed. A writing state flag indicating OFF state when the updated data block coincides with the 1st data block or ON state in case of inconsistency is set up in the 1st control block CTL1 and the unupdated data block is held in the 2nd control block CTL2. A trouble is detected by detecting the ON state of the writing state flag and the data block is restored by transferring the contents of the 2nd control block CTL2 to the 1st control block CTL1.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a data set protection processing method. In general, a combination of data that can be directly read from 7-'-1 blocks stored on a disk or written directly as data blocks is called a directly organized data set.

In this directly organized data set, if a failure occurs during data reading or writing, the data cannot be processed unless the failure is detected and all the original data on the side where the failure occurred is restored.

For example, data A, B, C, D are A/, B/lσ.

When performing processing to change data to y, if data is used as a data storage medium, data A on a certain tape.

Since B, C, and D must be changed to IZB/, σ, and D' and written to another tape, even if a failure occurs during processing, the original data A, B will remain intact.

C and D remain on the first tape.

However, in the case of disk, which is a directly organized data set, processing is performed on one medium rather than on another medium like tape.
processing within one medium.

Therefore, for example, if a failure occurs when A, B, C, and D are changed to A/ and B/, A and B before the change are already destroyed, making it difficult to process the data.

For this reason, in a directly organized data set, it is necessary to detect the failure and also restore the original data.

The present invention relates to a data set protection processing method that allows failure to be detected and data to be restored when the data set is destroyed in such a VCC direct organization data set.

[Conventional technology]

The conventional data set protection processing method will be explained below by dividing it into a failure detection method and an original data restoration method.

First, as a failure detection method, there is a method of notifying the program side of an error when a read/write command cannot be executed normally. Further, as a second method, contrary to the first method, there is a method in which the program itself checks for data inconsistency.

Next, as a method for restoring one data, there is a method that has a history of data writing and restores the history by going back to the time before the occurrence of a failure.

[Problem that the invention seeks to solve]

Among the above-mentioned conventional techniques, the first fault detection method has a problem in that when the entire system fails, the signal path is often abnormal, and the error notification is not transmitted to the program side button. Further, the second fault detection method is not checked unless the program side detects a data inconsistency.

For example, if there is no data in an area where data should originally exist, the program side will detect a failure as a data contradiction.

However, even if the content of the data that should have been originally changed due to some kind of failure, for example, when processing seat reservations at the green counter of Japanese National Railways, there may be some changes in the data, even though the seats should actually be full. Even if a vacant seat exists, the program does not recognize this phenomenon as a data inconsistency, so it is not checked out and is not considered to be a failure.

Next, the above-mentioned conventional data restoration method that has a history of data writing has the problem that all the data that has been read, changed, and written must be stored, resulting in a huge amount of history information. .

[Means for solving problems]

The present invention provides a data set protection processing method that solves the above-mentioned problems, and the method is a direct organization data set in which data blocks can be directly read and written. Contents corresponding to updated data blocks are sequentially written. A first management block for writing and a second management block for holding that holds the contents corresponding to the data block before update are provided, and the final pi HC write status flag of the first management block is set and the flag is set. indicates an off state if the updated data block and the first management block match, and an on state if they do not match, and if the above flag is on after updating the first data block, the second management block indicates an off state. This is done by a data set protection processing method characterized by restoring the first management block and data block to their states before writing and before updating, respectively, based on the contents of the management block.

[Effect]

In the above method, two management blocks are provided to manage the blocks when updating the data processors divided into each Darug, and the first management block has a data block that is divided into data processors. If they do not match, a write status flag is set indicating an off state and an on state, and since the data block before update is held in the second management block, the on state of the write status flag is detected. Faults can be found more easily than K, and in the event of a fault, the data block before update and the first management block before writing can be reliably restored by moving the contents of the second management block to the first management block.

〔Example〕

Hereinafter, the present invention will be explained by way of examples with reference to the accompanying drawings.

FIG. 1 is a diagram showing the configuration of an apparatus for implementing a data set protection processing method according to the present invention. The apparatus shown in FIG. 1 includes two management blocks CTLI provided in a direct organization data center, 7) DB, for example, a disk.

2 and the address 'Is' managed by these two management blocks! It consists of +6+"4r&6 data blocks.

Generally, one management block has the function of managing the location of a data block on the disk, that is, the address, and this management block determines what kind of data block is stored at which address on the disk. You can now see if there are any.

To explain with the embodiment of FIG. 1, the management block CTL 1
Bit 11, which is the first storage area, puts data block A at address lL1, bit 12 puts data block B at address al[, hit 13 puts data block C at address aBK, and bit 14 puts data block A at address lL1. Unused book) Empty book is at address a4,
Bit 15 indicates that empty block Empty 3 is stored at address a.

Also, the last bit 16 contains a write status flag fg
is erected. The write status flag fg is turned on (1) at the start of updating each data block A, B, C, empty 1, empty, and turned off (0) at the end of the update.

Each data block A, B, C, sky I, and 9 mushrooms are usually updated collectively for each group. For example, if you want to store the wages of employees belonging to a certain department of a company and their male and female employees, the contents of block A are each section, and this k
In contrast to K, the contents of block B are the names and wages of male employees, and the contents of block C are the names and wages of female employees, as shown by the double dashed arrow.

Furthermore, management blocks CTL 1 and 2 are exactly the same in FIG. 1 except for the presence or absence of the flag fg, but this is because each data block A, B, C, Empty 1, and Empty 2 are in the state before updating. be. As will be described later, when updating of a data block is started, the flag fg is set to 1, and the first
The contents of the management block CTL 1 are also rewritten, and fg becomes O when the update of the data block is completed and the contents of CTL 1 are all correspondingly rewritten. However, if the update is not completely completed, for example, if a failure occurs during the rewriting of CTL 1, the two will remain inconsistent and fg will remain as it was at 1. On the other hand, the second management block CTL 2 retains the data block before the update from the start to the end of the update, and if a failure is detected by the first management block CTL IK, the contents of this CTL 2 are transferred to the contents of CTL 1. This attempts to restore the data before the update by replacing it with the .

Bit 21 of CTL 2 indicates that data block A is at address aH, bit 22 is that B is at address a, and bit 23 is that C is at address a.
317, bit 24 indicates that empty l is stored in &4, and bit 25 indicates that empty 2 is stored in a6.

Hereinafter, the operation of the apparatus having the above configuration will be explained based on FIG. 2.

In Figure 2, ■ is before update, ■ is just before update, ■ is in the middle of update, and ■
indicates completion of update, [F] indicates destruction of the data server (DB) due to failure, and [F] indicates data restoration.

In the embodiment shown in FIG. 2, a case will be described in which data blocks A and 6 are updated.

Before the update, [stock], data block A is stored at address a1, B is stored at a2, C is stored at a3, empty 1 is stored at &4, empty 2 is stored at lL5, and the data block A is stored at address a1. The flag of 1 management block CTL 1 is off (0).

At the start of updating data blocks A and B (■), the first
The contents of management block CTL 1 are transferred to the second management block CT
Outputs L2K. At the same time, the first management block CTL
1 write flag is turned on (1).

When updating (■), first update the updated data blocks A and B.
are stored in addresses &6 and &4, which were empty blocks, respectively. At the same time, the addresses al and al are empty and trapped. Further, as for address a3, since the address a3 is not updated, the state is the same as before the update.

Along with updating this data, the first management block CTL 1 is also updated sequentially from the first bits 11 to 15 (FIG. 1). In reality, this update of CTL 1 is performed in block CTLI/, where the flag on the program side remains O.

Therefore, as shown in the figure, the contents of CTLI' are empty starting from the upper bit! , empty 2, C, B, and A, and the update ends with these written to disk 0CTL 1 (■). At this time, the write status flag of CTL 1 is O (■).

However, the contents of C1"L 1' are empty 1 l empty!+C-
After writing up to (indicated by the broken line α), a failure occurred in the system (@). In this case, the first management block C
The contents of TL 1 are as follows: The status of the upper 3 bits is
, empty z, and C, which matches the write state of the data block. However, as shown by the broken line β,
Although the upper two bits indicate that addresses a4 and a are both empty blocks, in reality, updated data blocks B and A have been written to those addresses. That is, at this point, there is a contradiction between the data block and the management block CTL 1,
Data Sera) DB destruction has occurred. This destruction is detected by the write status flag remaining in the ON state (1) because the update of the first management block CTL 1 has not been completed even though the update of the data block has been completed, as shown by the broken line γ. .

When this write status flag 1 is detected, the second management block CTL 2 retains the state before update.
The original data block is restored by K ([F]
).

〔Effect of the invention〕

In the above method, two management blocks are provided to manage the data blocks divided into each group when they are updated, and the first management block is turned off if the updated data block and the first block match. If they do not match, a write status flag indicating an on state is set, and since the data block before update is held in the second management block, it is easy to detect the on state of the above write status flag. If a failure can be found in the first management block and the contents of the second management block are transferred to the first management block when the failure occurs, it is possible to reliably restore the data block before update and the first management program before writing.

[Brief explanation of drawings]

Figure 1 is a configuration diagram of an apparatus for carrying out the method of the present invention;
The figure is an explanatory diagram of the operation of the system of the present invention. CTL 1...first management block, CTL2...second management block, A, B, C...data block, empty. Empty 3...Empty block s 'I H'S P'S +
"4 + '5...Address, fg...Write status flag.

Claims (1)

[Claims] A first management block for writing in which the contents corresponding to the updated data blocks are sequentially written in a directly organized data set that can directly read and write data blocks, and a retention block that holds the contents corresponding to the pre-updated data blocks. A write status flag is set in the last bit of the first management block, and if the updated data block and the first management block match, the flag is set to off state. If the above flag is on after updating the data block, the first management block and the data block are set to the state before writing and before updating, respectively, based on the contents of the second management block. A data set protection processing method characterized by restoring data.