JP2772068B2 - Data assurance processing method for inherited information - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Overview] By transferring the information set in the nonvolatile shared memory from the active process to the standby process, when the active process fails, the standby process replaces the active process.
A method for data assurance processing of handover information for fault-tolerant operation in a computer system that continues to operate in a business system, enabling efficient and reliable pre-construction of an operating environment for fault-tolerant operation For this purpose, the area of the non-volatile shared memory in which the handover information is set has a dual structure of the current version and the new version, and a base information area and a difference area are provided for each of the current version and the new version. The current process writes to the current version, the standby process reads from the new version, and the current process transfers the contents of the current version's basic information area to the new version's basic information area after the lapse of a predetermined time or the accumulation of a predetermined amount of information. Copy and switch between the current version and the new version,
It is configured to notify it.

[Industrial applications]

According to the present invention, the information set in the nonvolatile shared memory is
By taking over from the active process to the standby process,
The present invention relates to a data assurance processing method of takeover information for making a fault-tolerant computer system in which a standby process replaces an active process and continues to operate when a failure occurs in an active process.

In a computer system, a method of duplicating a control process is effective in realizing fault tolerance. One is a control process that operates for normal operation, and the other is a control process that is on standby to continue operations when the normal operation control process is stopped due to some failure. is there.

The former control process is called an active process, and the latter control process is called a standby process.

When switching the processing subject from the active process to the standby process due to a failure, it is necessary to take over various types of resource information. There is a need for a means for quickly and reliably transferring the resource information.

[Conventional technology]

 FIG. 5 shows an example of the prior art.

In FIG. 5, when the active process 10 is stopped due to a failure, the standby process 11 takes the place of the active process 10 to continue the business. Must be taken over by the standby process 11.

Information for inheriting the operating environment is transferred to the current process
If all 10 are taken over in the event of a crash, it takes a long time to set the operating environment based on the inherited information, and the performance of switching from the active process 10 to the standby process 11 deteriorates.

Therefore, a method of sequentially taking over resource information such as an operation environment when the active process 10 is operated has been considered as shown in FIG.

[Problems to be solved by the invention]

For example, when an operation environment for controlling network resources in advance in the standby process 11 for hot standby of a communication control program, etc.
The active process 10 writes the changed operating environment information to the non-volatile shared memory 12 and notifies the standby process 11 at a certain cycle or when the amount of changed information reaches a predetermined amount. Must meet the following requirements:

(1) There is no problem in performance (processing steps) for operation in the active process 10. That is, there is no effect on the processing of the active process 10 due to the preliminary construction of the operating environment that is performed intermittently in the standby process 11.

(2) Uniqueness of data viewed from the current process 10 and the standby process 11 in a time section is guaranteed.

However, in the conventional method, since the area of the nonvolatile shared memory 12 is simply used as one structure, the above two requirements cannot be satisfied at the same time. In other words, when the active process 10 is writing, and the standby process 11 simultaneously reads that location,
Data integrity may not be guaranteed. for that reason,
If exclusive control is performed by lock control or the like, the processing of the active process 10 may be delayed, and there is a problem that response and the like become slow even during normal operation.

SUMMARY OF THE INVENTION It is an object of the present invention to solve the above-mentioned problems and to efficiently and surely pre-construct an operation environment for fault-tolerant operation.

[Means for solving the problem]

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

In FIG. 1, reference numeral 10 denotes an active process that operates for normal operation, 11 denotes a standby process that stands by to continue operation when the active process 10 fails, 12 denotes a non-volatile shared memory, 13A, 13B is an area for storing takeover information called current version and new version, respectively, 14A and 14B are difference areas, 1
5A and 15B denote a basic information area, and 16 denotes a nonvolatile shared memory management process.

The nonvolatile shared memory 12 is a storage device whose storage contents are guaranteed even when the program of the active process 10 or the standby process 11 is stopped. For example, it is configured by a semiconductor memory device with a separate power supply.

The standby process when the active process 10 fails
The resource transfer information such as the operating environment of the active process 10 is sequentially set in the nonvolatile shared memory 12 so that the standby process 11 can perform the processing on behalf of the active process 10. The standby process 11 operates based on the information. Create an environment in advance.

In the present invention, the area of the nonvolatile shared memory 12 in which the handover information is set has a double structure of the current version 13A and the new version 13B.

The current version 13A and the new version 13B contain basic information areas that store information to be transmitted from the active process 10 to the standby process 11, respectively.
15A and 15B, and difference areas 14A and 14B in which information indicating the information of the change is stored.

When handover information such as the operating environment is generated, the active process 10 updates the contents of the basic information area 15A of the current version 13A by processing P1 and indicates which information in the basic information area 15A has been changed. Is set in the difference area 14A.

Then, when a certain time has elapsed or the amount of change has exceeded a certain amount, the contents of the basic information area 15A are copied to the basic information area 15B by the process P2. After that, the current version 13A becomes the new version and the new version
The non-volatile shared memory management process 16 is requested to switch the 13B to the current version.

Next, in process P3, an intermittent synchronization notification is sent to the standby process 11, which means a request to reflect the takeover information.

On the other hand, in the standby process 11, the current process 10
While writing to A, the reading from the new version 13B is performed, and the operating environment is pre-generated by reflecting it in the standby process 11.

In addition, while the active process 10 is writing to the current version 13A, the standby process 11 completes the pre-generation of the operating environment from the new version 13B, so that the time interval value and The amount of information change is determined at the time of system design.

If there is an intermittent synchronization notification from the active process 10, new takeover information is notified. The standby process 11 queries the nonvolatile shared memory management process 16 to confirm that the current version 13A is the current new version. Judgment is made, and the area of the new version is set as a reading target of new takeover information.

[Action]

The area of the non-volatile shared memory 12 is changed as in the prior art.
As a simple structure, when intermittent information is transferred from the active process 10 to the standby process 11, while the standby process 11 is reading, the active process 10
Process is awaited, causing a problem in business.

In the present invention, the takeover area in the nonvolatile shared memory 12 is divided into the current version 13A and the new version 13B, and this dual structure area is cyclically switched for use. Disappears.

Further, if the current version and the new version are simply switched, there is a possibility that the standby process 11 may forcibly perform the switching when the reading from the new version 13B is not completed. In this case, the part that has not been read out is not reflected on the standby process 11 side. Although it is considered a rare case, if it happens, normal processing cannot be continued.

Therefore, in the present invention, in particular, the active process 10 includes a sequence number (1 is added for each notification) in the intermittent synchronization notification so that the standby process 11 can determine whether or not the intermittent synchronization process is performed in a normal cycle. Is added.

As a result, when the intermittent synchronization process is not sequential, the standby process 11 determines that accurate handover cannot be performed only with the handover (difference area) information of the changed portion, and executes the handover process from the base information area. When switching between the current version and the new version, the basic information area 15A should be used so that all inherited information is always retained.
Information in the base information area 15B of the new version 13B
It is copied to.

〔Example〕

FIG. 2 is an example of an application system of the present invention, FIG. 3 is a data structure of a nonvolatile shared memory area according to an embodiment of the present invention,
FIG. 4 shows an example of processing according to an embodiment of the present invention.

The present invention can be applied to, for example, a complex system as shown in FIG. In FIG. 2, those having the same reference numerals as those in FIG. 1 correspond to those shown in FIG. 1, and 20-1 and 20-2 are physical units which can exclusively use a CPU and a memory, respectively. Processor module (PM), 21 is PM
A bus (SS-BUS), which is a transfer path connecting between each other and between the PM and the nonvolatile shared memory 12, is shown.

In the system shown in FIG. 2, the following processing is performed to take over the resource information.

An external event such as data transmission / reception is input to the active process 10 from another process in the same PM or another process in another PM.

The active process 10 stores the input of the external event in an internal control table or the like, and writes the information in the current version 13A of the nonvolatile shared memory 12 in order to transfer the information to the standby process 11.

When a certain time elapses or the amount of writing exceeds a certain amount, necessary information is copied to the new version 13B, and the current version and the new version are switched.

Instruct the standby process 11 to read the takeover information by sending a message.

Since external events notified to the active process 10 occur continuously, the active process 10
The current version after the switch in the nonvolatile shared memory 12 (the original new version
Write the handover information to 13B).

In response to the notification, the standby process 11 reads the information written by the active process 10 in the processing and reflects the information in the standby process 11.

The processing information is the current version after the switch (the original new version 13
If a certain amount of data is written in B) or a certain time has elapsed, the current version and the new version are switched again. The relationship between the current version and the new version returns to the state before the processing switch.

Similarly to the above, after instructing the standby process 11 to read the takeover information by sending a message, the active process 10 continues to execute the non-volatile shared memory.
Write the takeover information to the current version 13A after the switch in 12.

The standby process 11 reads the information written by the active process 10 in the processing from the non-volatile shared memory 12 and reflects the information written in the process.

By repeating the above processing, the uniqueness of the data in the nonvolatile shared memory 12 is guaranteed.

FIG. 3 shows the data structure of the nonvolatile shared memory area in this embodiment.

In order to intermittently transmit the handover information from the current process to the standby process, the data structure of the non-volatile shared memory area has a double structure of the current version 13A and the new version 13B as shown in FIG. This information is taken over after a certain period of time or when a certain amount of information has been accumulated.
As information to be transmitted, not all information in the area of the non-volatile shared memory 12 is targeted, but only information of an updated portion is targeted.

The difference areas 14A and 14B are used to limit the information of the change in the standby process, and store the pointers into the basic information areas 15A and 15B. In the basic information areas 15A and 15B, actual handover information is stored, and here, all information transmitted from the active process to the standby process is stored. However, some information may become unnecessary depending on the work of the active process.
Such information is discarded halfway, so the basic information area 15
A and 15B do not overflow.

The communication control process will be described as an example as follows.

In the communication control process, for example, when a communication resource such as a line or a data link is activated, it is necessary to transmit, as information, the attribute of the activated communication resource.
Further, when deactivation is performed, it is necessary to transmit, as information, which resource, which has been notified, has been deactivated. After the inactivation is notified, the inheritance information such as the attribute of the resource becomes unnecessary, and it is not necessary to hold the information in the basic information area 15A or 15B.

The current version 13A is an area where the active process writes the takeover information, and the new version 13B is an area where the standby process reads the takeover information. Therefore, the pointer format from the difference areas 14A, 14B to the basic information areas 15A, 15B is different between the current version and the new version.

The process flow of the active process and the standby process will be described with reference to FIG. 4 by taking as an example the event of activation and deactivation of a line that is a communication resource. In order to simplify the explanation, the number of entries in the difference areas 14A and 14B is 2, and the information amount in the basic information areas 15A and 15B is 4.

When the activation request of the line L1 is instructed by a command or the like, the active process acquires an unused area from the base information area 15A of the current version, sets attribute information to be transmitted on the line L1, and sets a difference area. In 14A, an activation request code and a pointer to the information area of the line L1 in the basic information area 15A are set.

When there is an activation request for the line L2, the takeover information is set as in the processing. As a result, the difference area 14A and the base information area 15A of the current version are in a state as shown in FIG.

The active process, because the difference area 14A is full,
Start the intermittent synchronization process.

First, as shown in Fig. 4 (b), all necessary information in the basic information area 15A in the current version is
Copy to 15B.

Next, as shown in FIG. 4C, the current version and the new version are switched.

After that, an intermittent synchronization notification is sent to the standby process.

In the standby process, based on the intermittent synchronization notification, the updated information, that is, the activation of the line L1 and the line L2, is known from the information of the new difference area 14A, and is reflected in the standby process. The information in the difference area 14A after the reflection is cleared or invalidated.

In the active process, in parallel with the processing in the standby process, the setting process of the takeover information is performed for the new current version difference area 14B and the base information area 15B. For example, line
When an L2 deactivation request is instructed by a command or the like, a deactivation request code is set in the difference area 14B,
A pointer to the basic information area 15B in which information on the line L2 is stored is set.

Further, when there is an activation request for the line L3,
As in the processing, the setting of the handover information is performed. As a result, the difference area 14A and the basic information area 15B of the current version are in a state as shown in FIG. 4 (d).

Since the difference area 14B is full, the active process
Start the intermittent synchronization process.

First, all necessary information in the basic information area 15B of the current version is copied to the new basic information area 15A.

At this time, since the area in the basic information area 15A used by the line L2 becomes unnecessary by deactivation, it is released as an unused area. As a result, a state as shown in FIG.

Next, as shown in FIG. 4 (f), the current version and the new version are switched.

Then, an intermittent synchronization notification is sent to the standby process.

In the standby process, based on the intermittent synchronization notification, the deactivation of the line L2 and the activation of the line L3 are known from the information of the new version of the difference area 14B, and reflected in the standby process.

In the active process, in parallel with the processing in the standby process, the setting process of the handover information is performed for the new current version difference area 14A and base information area 15A. Hereinafter, processing similar to the above-described processing is cyclically performed.

Note that deactivation means that the resource information is eventually destroyed.

〔The invention's effect〕

As described above, according to the present invention, the process of the active process does not have to wait due to the pre-construction of the operating environment that is performed intermittently in the standby process, and the data to be inherited due to the delay of the process in the standby process. Can be prevented from being lost. Therefore, it is possible to reliably realize the fault-tolerant communication control program and the like.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of the present invention, FIG. 2 is an example of a system to which the present invention is applied, FIG. 3 is a data structure of a nonvolatile shared memory area according to an embodiment of the present invention, and FIG. FIG. 5 shows an example of processing according to the prior art. In the figure, 10 is the active process, 11 is the standby process, 12 is the non-volatile shared memory, 13A is the current version, 13B is the new version, 14A and 14B are the difference areas, 1
5A and 15B are basic information areas, 16 is a nonvolatile shared memory management process, and P1 to P4 are processing steps.

Continued on the front page (58) Fields surveyed (Int.Cl. ⁶ , DB name) G06F 11/16-11 / 20,12 / 16,15 / 16 G05B 9/03 H04M 3/22 H04L 11/20

Claims (1)

(57) [Claims] 1. An active process (10) and a standby process (11)
By passing the information set in the non-volatile shared memory (12) from the active process (10) to the standby process (11), the standby process (11) is activated when the active process (10) fails. In the data assurance processing method of the takeover information in the computer system that continues the operation instead of the current process (10), the area of the nonvolatile shared memory (12) where the takeover information is set is the same as the current version (13A). The new version (13B) has a dual structure. The basic information areas (15A, 15A, 13B) store information to be transmitted from the active process (10) to the standby process (11) in the current version (13A) and the new version (13B) 15B) and a difference area (14A, 14B) for storing information indicating the changed information, the active process (10) writes to the current version (13A) (P1), and the standby process (11 ) Reads from the new version (13B) (P4) The working process (10) copies the contents of the basic information area (15A) of the current version (13A) into the basic information area (15B) of the new version (13B) after a predetermined time or a predetermined amount of information has been accumulated. At the same time, the current version is switched to the new version, the new version is switched to the current version (P2), and the switching between the current version and the new version is notified from the active process (10) to the standby process (11) (P3). Data assurance processing method.