JP3487732B2 - Apparatus and method for avoiding duplicate database data inconsistency - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a duplicate database inconsistency avoiding apparatus and a duplicate database inconsistency avoiding method in a duplicate database system.

[0002]

2. Description of the Related Art A conventional duplicate database system is a system that is operated so that a duplicate of a master database at one site is made at another site and the consistency between the master database and the duplicate is maintained.

In the conventional replicated database system, in order to maintain the consistency between the master database and the replica, for example, the update contents of the master database are periodically provided with a communication circuit from the site having the master database to the site having the replica. The copy is updated by the updated contents.

For example, in a company with a head office, the head office is selected as a site with a master database,
The branch is chosen as the site with replication. In this case, there is no need to access the master database of the head office via the communication circuit every time the branch office accesses the customer's database, which brings advantages such as shortened access time and reduced communication cost. So it is popular.

However, since the conventional duplicate database system has only one master database and is designed based on the idea that the contents of the duplicate database system should be reflected in the duplicates of other sites, a plurality of databases can be used as master databases.
If you proceed from the conventional basic configuration of classifying into databases and their replicas and try to treat multiple databases as equal, that is, if you want to update the data in any database, use the conventional replica. If the update method (replication database reflection method) is applied to both databases, if the same record in the same table in multiple databases is updated with different values at the same time, the respective reflection information will be stored in the copy destination database device. There was a possibility that they would be transferred and changed by each other's reflection information.

As a method of preventing such data inconsistency between sites, a time stamp column is provided in a record, and when data with an old time stamp is propagated from another site, a replica is created. There is a method of not reflecting the information and a method of only performing replication by updating at the operation level such as addition / subtraction.

[0007]

However, in the first method of preventing the occurrence of inconsistency described above, since a column dedicated to replication whose attribute is a time stamp is newly provided in the actual record, the data area of the data area is In addition to the increase, there is a problem that special processing is required to perform column maintenance (processing that embeds the execution time when inserting data). However, there is a problem that it is not possible to process non-countable data such as character strings and image data.

[0008] The present invention has minimal impact on the existing users use the system, there is not only a particular attribute data taken as the replication target, replication attribute <br/> of data of all the hand and an object thereof is to allow avoid data inconsistencies in as the target replica database.

[0009]

A replicated database inconsistency avoidance apparatus according to the present invention includes a log journal file for recording an update history in a replicated database inconsistency avoidance apparatus in a replicated database system in which a database is updated at each site. A log journal section to manage, a replication information section to manage a replication file that records a replication history, and the latest replication execution time from the log journal section when attempting replication, and attempts to perform replication. If the update time of the record is acquired from the replication information section and the update time of the record to be replicated is earlier than the latest replication execution time, the replication is executed and the replication is executed later. Lever, characterized in that it comprises a replication execution processing unit for the error notification without executing the replication.

According to the duplicate database inconsistency avoidance method of the present invention, the latest replication execution time is updated when attempting replication in the duplicate database inconsistency avoidance method in the duplicate database system in which the database is updated at each site. The procedure for acquiring from the log journal section that manages the log journal file that records the history, the procedure for acquiring the update time of the record that is going to be replicated from the replication information section that manages the replication file that records the replication history, and the latest If the update time of the record to be replicated is earlier than the replication execution time, the replication is executed, and if it is later, the error notification is issued without executing the replication. Characterized in that it has a.

A recording medium according to the present invention is characterized by recording a program for causing a computer to execute the above-mentioned duplicate database inconsistency avoiding method.

[Operation] In order to determine whether or not data inconsistency will occur due to the execution of replication, the determination is made based on the log journal information and replication information managed at the local site. Therefore, the determination process can be performed only at the own site, and the amount of propagation data on the network can be reduced.

Since no special column data is required in the actual record data, the maintenance process for the actual record data is not required, and it can be applied to the existing system.

Since it is irrelevant to the attribute / content of the column data, there is no problem that the determination of data inconsistency is effective only for the processing on a specific column, and the data inconsistency does not occur for all column attributes of the target DBMS. The judgment becomes valid.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the present embodiment will be described in detail with reference to the drawings.

FIG. 1 shows the structure of each site in this embodiment.

The files used are a log journal file 1 that actually stores history information, a database file 2 that stores data, and a replication file 7 that stores the execution history of replication. Log journal file 1 and database file 2
Is managed by the database management system 3, and the log journal file 1 is managed by the log journal unit 4, which is one of the internal components of the database management system 3. The replication file 7 is managed by the replication information unit 8, and the replication execution processing unit 5 acquires log journal information from the log journal unit 4 in the database management system 3 and replication information from the replication information unit 8 to execute replication. Take control.

Next, in bidirectional replication, the same
Replicate update results for each site for files
To avoid file inconsistency that occurs when
Will be described in detail .

The states after execution of the bidirectional replication for the sites A and B are 301 and 302. Therefore, in 301 and 302, site A,
The table of the replication target of the site B has the same data content.

Next, the update process for the record A of the site A is executed (303), and the update information 110 of the record A is generated in the log journal unit 102. Update processing for the same record A at site B is executed (30
4), the update information 210 of the record A is generated in the log journal unit 202.

After that, when the information is propagated by the replication (305), the update information 1 after the previous replication for the target record at the site A is obtained.
Since 10 exists, an error occurs. Similarly, site B
However, an error occurs because the update information 210 after the previous replication exists.

If an error message is issued when an error occurs, the user can know that a replication error due to data inconsistency has occurred based on the message contents, and individual measures can be taken.

Next, the outline of the determination process executed at each site will be described with reference to FIG.

As the pre-processing of the replication execution, the latest replication execution time is acquired from the replication information section 8 (step 401).

Subsequently, the following processes (402 to 406) are executed while the replica information exists.

When all the records are updated with the replica information, the process ends normally (step 402).

The record number of the update processing target (identifier capable of uniquely recognizing the record in the database) is acquired (step 403), and the update time in the log journal of that record number is acquired (step 404).

The latest replication execution time is compared with the update time (step 405). If the update time is later, an error may occur because data inconsistency may occur (step 406). The error processing 406 returns an error code as a return value or outputs an error message indicating a replication execution error.

If there is no update information after the latest replication execution time, the replica information is reflected (step 407).

In this way, it is determined whether replication is feasible, and the replication information is finally updated (step 408).

[0031]

As described above, according to the present invention, the multiple
Implement replication in a database system
If doing so causes data inconsistency
User as an error without executing the replication
Inform them of this and perform replication.
If the data inconsistency does not occur due to
Has the effect that the application can be executed
It

[0032] Also, the information before update is recorded in the log file of each site.
It is acquired from the
There is no increase in network load. By this
This reduces the impact on the network situation.
Wear. Further, since the problem of reconstructing actual data does not occur, maintenance is easy. This makes it possible to reduce the influence on the existing system.

Furthermore, since not the data itself but the history information is used as a material for judging whether or not the replication can be executed, the non-countable data item can be processed. Thus, Ji mismatches all attribute data
Check is possible.

Further, since the history information is used as the material for determining whether or not replication is possible, not the data itself, it is easy to find the cause of the replication error. This facilitates the subsequent planning of a coping method. If a replication error occurred because the record that was trying to reflect the replica information was updated at the local site after the last replication, the user sees this error and manually sets the value of that record. You can know what you have to decide in. Error notification is useful because humans often perform processing when such conflict occurs.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a database system including a duplicate database data mismatch avoidance device according to the present invention.

FIG. 2 is a process outline diagram showing an outline of an execution process according to the present invention.

FIG. 3 is a flowchart showing an operation and a avoiding method of the duplicate database data mismatch avoidance apparatus according to the present invention.

[Explanation of symbols]

1 log journal file 2 database files 3 database management system 4 log journal section 5 Replication execution processing unit 7 replication files 8 Replication information section

─────────────────────────────────────────────────── ─── Continuation of the front page (56) Reference JP-A-8-185346 (JP, A) “Asynchronous updating of replicated files is the core technology of distributed database” NIKKEI ELECTRONICS, June 6, 1994, no. 609, p. 101-110 (58) Fields investigated (Int.Cl. ⁷ , DB name) G06F 12/00

Claims (3)

(57) [Claims] 1. In a duplicate database inconsistency avoiding apparatus in a duplicate database system in which a database is updated at each site , each site manages a log journal file that records the update history of the record of the database of its own site. The journal section, the replication information section that manages the replication file that records the replication history for the records in the database at the local site, and the other information when executing this replication.
Records to be replicated this time
Information is propagated to the information of the propagated record
Update the records in the corresponding database on your site
The time is acquired from the log journal section and the
Application execution time from the replication information section
Acquired, if the previous times than replication execution time before the update time of the record that you attempt to now times of replication, run the now-time replication, the previous rounds of replication execution time now times of replication than A replication database inconsistency avoiding apparatus, comprising: a replication execution processing unit that gives an error notification without performing replication this time if the update time of the record to be executed is later. 2. In a replication database inconsistency avoiding method in a replication database system in which a database is updated at each site, when each site attempts replication this time, it becomes a target of this replication from another site. Record
The information of the propagated record is received.
Of the records in the database on your site that correspond to the
The update time of the record of the database of the own site
Obtained from the log journal section that records the update history,
The replication execution time of the
Record replication history for base records
From the replication information section,
This replication is performed after the application execution time.
If the record to be updated has the latest update time, this record
Application and last replication run
A record that tries to replicate this time rather than the time
If the update time of the node is later, this replication
A method for avoiding inconsistency in a duplicate database, which is characterized by notifying and executing error notification . 3. A recording medium on which a program for causing a computer to execute the duplicate database inconsistency avoiding method according to claim 2 is recorded.