JP2680061B2 - Log data recording processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Outline] Regarding a log data recording processing method in transaction processing, storing log data records in blocks for the purpose of efficiently realizing log data recording. A circular storage medium, and a log data reading unit that processes the stored log data to read the log data in block units,
A log data purging unit that processes so as to purge unnecessary log data portions of the read block, and a new log data block by merging new log data with the purged log data block. And a log data merging means for processing so as to write the generated block to a block position on the circulating storage medium that forms a predetermined moving position from the read position of the log data reading means. In addition, the moving position is set so that the time required until the block of the written log data is read is the same for all the block positions on the circulating storage medium.

[Industrial applications]

The present invention relates to a log data recording processing method in transaction processing, and more particularly to a log data recording processing method that enables efficient recording of log data with a small memory capacity.

In order to realize recovery processing when the system is down, the data processing system samples and records the data information updated in the transaction processing as log data. Such log data recording processing needs to be efficiently realized by reducing the memory capacity.

[Conventional technology]

Conventionally, log data is recorded by sequentially storing the generated log data in a volume, and when the volume is full, recording is continued in another prepared volume. It was realized by the processing method. That is, in the prior art, not only the log data recording process for recording the execution history but also the log data recording process for the restoration process is performed so as to record in such an infinitely continuous data format. It was.

[Problems to be solved by the invention]

However, such a conventional technique has a problem in that at least two volumes must be prepared, and volume resources for other data processing are reduced.

Certainly, in the log data recording process for recording the execution history, it is necessary to record the log data in a data format that continues indefinitely, but the log data recording process for the restoration process In that case, if you proceed to the next transaction processing without any error in the system, the log data of the transaction processing that has been executed up to that point becomes unnecessary. It is not necessary to record log data. Therefore, it can be said that the conventional technology wastefully uses the volume resource.

The present invention has been made in view of such circumstances,
It is an object of the present invention to provide a log data recording processing method that enables efficient recording of log data.

[Means for solving the problem]

 FIG. 1 is a diagram showing the principle of the present invention.

In the figure, 10 is a log data collecting means for collecting log data generated in a transaction process, 20
Is a circular storage medium such as a rotating storage medium,
A unit for storing the records of the log data collected by the log data collecting unit 10 in block units, and a unit for reading the log data 21, which processes the log data written in the circulating storage medium 20 in units of blocks. ,twenty two
Is a log data purging means, and is a log data reading means
A process for purging an unnecessary log data portion of a block of log data read by 21. 23 is a log data merging unit, which is a log data purging unit.
The block of log data purged by 22 is processed to merge the new log data collected by the log data collection unit 10 to generate a new block of log data, and 24 is a log data writing unit. There is a unit for processing the block of the log data generated by the log data merging unit 23 so as to be written in the block position on the circulating storage medium 20 that forms a predetermined moving position with the reading position of the log data reading unit 21, 25 The moving position setting means sets the moving position required by the log data merging means 23.

[Action]

In the present invention, the movement position setting means 25 is arranged so that the block positions of the log data written by the log data writing means 24 are uniformly designated with respect to all the block positions on the circulating storage medium 20. Set the movement position. Therefore, the block of log data read by the log data reading means 21 has passed a predetermined fixed time after being written by the log data writing means 24.
From now on, if you set the movement position so that this fixed time is about the time when transaction processing can be completed,
The log data read by the log data reading means 21 is related to the log data of the transaction processing which becomes unnecessary when the processing ends.

From now on, the log data purging unit 22 discards the unnecessary log data, and the log data merging unit 23 reuses the free space to write the log data, so that the log necessary for the restoration process is obtained. Data can be recorded and held with a small memory capacity.

〔Example〕

 Hereinafter, the present invention will be described in detail with reference to examples.

FIG. 2 shows a system configuration diagram for realizing the log data recording processing method of the present invention. Here, an embodiment using a rotating storage medium as the circulating storage medium 20 is disclosed.

In the figure, 1 is a log data management unit that receives log data generated in a transaction process and records the received log data in a log file. 20a is a rotating storage medium such as DASD. Which is a temporary log file that stores records of blocked log data, and 30 is an archive log file, which corresponds to the history log file adopted in the prior art. It is a thing.

The log data management unit 1 of the present invention, when the format write is executed in the initialization process, causes the temporary log file
From the reading point of 20a (the place indicated by "α" in the figure),
A block of new log data is processed by sequentially reading blocks of log data, discarding unnecessary parts of the read block of log data, and merging newly received log data with this. Will be processed. Then, the blocks of the new log data thus generated are sequentially written to the writing point (the portion indicated by "β" in the figure) of the temporary log file 20a. When the log management information in the main memory indicates that there is no remaining log data, the log data management unit 1 does not perform the merging process but performs a process to write the blocked log data at the writing point. It will be.

As described above, the temporary log file 20a included in the present invention is completely different from the conventional temporary log file in character. That is, while the conventional temporary log file is provided to reduce the load of writing processing to the history log file, the temporary log file 20a of the present invention is replaced with unnecessary log data. It will be prepared for cyclic reuse such as storing new log data.

Next, the positional relationship between the read point and the write point, which is adopted by the present invention in order to realize such reuse, will be described.

In the present invention, the rotation angle formed by the read point and the write point is set so as to uniformly specify the block position of the log data on the temporary log file 20a. That is, when the block read at the read point is written to the write point, the time required until the written block is read next is calculated with respect to the block position on all temporary log files 20a. It will be set to be the same. Specifically, such a setting process is performed by the number "Y" of the physical block (actual block in the temporary log file 20a) corresponding to the theoretical "X" th block.
Is realized by determining according to the following formula.

Y = [Q × X] mod (P) where P is an appropriate prime number that does not exceed the total number N of blocks,
Q is a natural number that does not have a common divisor with P.

FIG. 3 shows a specific example in which N = 8, P = 7, Q = 3. To explain in detail, 1
The log data is written to the logical block # 2, the log data is read from the logical block # 2, the log data is written to the logical block # 2, the log data is read from the logical block # 4, and the logical # 5 is written. When the log data is sequentially read out and written sequentially such as writing the log data to the block, the log data is written to the third physical block as the physical block, and then the log data is written from the sixth physical block. Read out, 2
The log data is written in the physical block No. 1, the log data is read from the physical block No. 5, and the log data is written in the physical block No. 1.

As will be apparent if this operation description is continued, the physical block targeted for the read processing becomes the target for the write processing in the next cycle, and the physical block targeted for the write processing becomes , Will be subject to read processing in the next cycle. Therefore, if N and P are properly selected, the rotation angle can be set appropriately, so that the processing of the transaction that is the source of the log data ends until the written log data is read next time. Can be configured as

As described above, in the present invention, the time required until the block written at the writing point is read next is set to be the same for all the block positions on the temporary log file 20a. From now on, it is possible to keep the number of generated log data and the number of discarded log data in equilibrium while the system is operating normally.
You will be able to record the necessary log data only with a. As is clear from the above description, even if the log data recording processing method of the present invention is adopted, the implementation of the conventional log data recording processing method is not hindered. From now on, if the number of generated log data increases and the equilibrium state cannot be maintained, log data to be written to the archive log file 30 is left, and after that, archive log The log file 30 will be processed.

Next, the criteria for determining whether the log data may be invalidated will be described.

The simplest method for determining whether to invalidate log data is a method in which the log data to be stored is provided with a physical time or a logical time, and the log data is determined to be invalid according to the physical time or logical time. However, such a determination method is inefficient because each determination must be made. Therefore, the oldest transaction number among active transactions is managed, and the pre-update data belonging to older transactions is unconditionally discarded.
The oldest unwritten transaction number (which is older than the oldest active one) among unwritten transactions is also managed, and the updated data before this is expelled to the archive log file 30. If the batch invalidation determination method of performing or rejecting is adopted, the invalidation determination can be made efficient, and unnecessary log data need not be left in the temporary log file 20a for one rotation or more.

Then, as a countermeasure against the mixture of long transactions, the completion information regarding the log data of the long transactions that occur after these transaction numbers is retained, and the one with the completion information is determined to be invalid. It will be perfect if it also has an invalidation determination method.

Next, a method of using the log data recorded in the temporary log file 20a will be described.

When the system goes down, the latest block position is detected by searching in sequence according to the ordering modulo P and Q. Then, by repeatedly reading from the latest block position in the reverse order a required number of times, it is possible to acquire all pre-update data. Similarly, committed updated data can be acquired by searching in a normal order. That is, in the present invention, although log data is recorded formally in discrete physical blocks, it is managed in a continuous format as logical blocks, so that log data required for restoration processing is maintained as in the past. You can get it. Using the obtained pre-update data and post-update data, the process of reproducing the update status of the transaction and restoring the database or the like to the latest state can be realized by the traditional means.
At this time, the auxiliary archive log file 30
It is also possible to use the log data of. Backout processing for restoring the state at the time of starting a transaction is also possible by referring to the pre-update data obtained as described above. This reference starting point is preferably recorded in the transaction management list as a log block number. Also, in order to know the positional relationship between the pre-updated data,
It is preferable to have a back pointer. When the database fails, it just reads the updated data in the normal order.

When the log data is frequently generated, it is possible to use two or more temporary log files 20a to cope with the situation. For example, if W-way distribution is performed by logically dividing even and odd block numbers,
This can be realized by determining the “Y” -th physical block corresponding to the “X” -th logical block by Y = [[Q × X] mod (P)] mod (W).

Although the illustrated embodiment has been described, the present invention is not limited to this. For example, the storage medium for the log data is not limited to the rotating storage medium such as DASD, but may be any circulating storage medium having a substantially endless configuration. Further, this circulating storage medium is not limited to one physically configured by one constituent element.

〔The invention's effect〕

As described above, according to the present invention, since it becomes possible to store log data while reusing a circular storage medium, it becomes possible to record and hold log data required for restoration processing with a small memory capacity. is there.

[Brief description of the drawings]

FIG. 1 is a block diagram of the principle of the present invention, FIG. 2 is a block diagram of an embodiment of the present invention, and FIG. 3 is an explanatory diagram of setting of a rotation angle. In the figure, 1 is a log data management unit, 10 is a log data collecting unit, 20 is a circular storage medium, 20a is a temporary log file, 21 is a log data reading unit, 22 is a log data purging unit, and 23 is a log data merging unit. , 24 is a log data writing means, 25 is a moving position setting means, and 30 is an archive log file.

Claims (1)

(57) [Claims] 1. A method for recording log data in a transaction process, comprising a circular storage medium (20) for storing records of log data in blocks, and a circular storage medium (20) written in the circular storage medium (20). Log data,
A log data reading means (21) for processing to read in block units, and a log data purging means (21) for processing so as to purge an unnecessary log data portion of a block read by the log data reading means (21) ( 22) and a log data merging means (23) for merging new log data with the block of log data purged by the log data purging means (22) to generate a new block of log data. The block of log data generated by the log data merging means (23) is located at a predetermined moving position from the reading position of the log data reading means (21), and the circulating storage medium (2
0) and a log data writing means (24) for processing to write to the upper block position, and the time required until the block of the written log data is read is on the circular storage medium (20). A recording processing method of log data, wherein the moving position is set so as to be the same for all block positions.