JPH11353292A - Cluster system and its fail over control method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a cluster system which executes a proper fail over operation in accordance with the operating state of the fail over destination. SOLUTION: In this cluster system, two server computers 10a and 10b which are connected to a public LAN 1 are loosely coupled to each other with an interconnect LAN 2. Then both computers 10a and 10b execute the communication to confirm their normal operations with each other using the LAN 2 at every prescribed interval. When one of both computers detects a failure of the other, the computer fails over the system resources operating at the other computer on its own computer. In this case, however, the control is carried out not to mechanically fail over the all system resources of the faulty computer but to change the priorities of system resources according to its own operating state and also to chance the priorities of the system resources which are originally operating at its own computer.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a cluster system including a plurality of loosely coupled computers and a failover control method for the cluster system.

[0002]

2. Description of the Related Art In recent years, computerization has been rapidly advanced due to the spread of computers, and computer systems have been constructed in various industries. In addition, the demand for improved fault tolerance of these computer systems is increasing year by year. A cluster system exists as a system for realizing this fault tolerance.

[0003] This cluster system is, for example, a system composed of a plurality of loosely coupled computers sharing a magnetic disk device or the like, and in addition to functioning as a distributed system for distributing the load of data processing,
When one of a plurality of computers fails, the system resources (such as application programs including utilities) running on that computer are taken over and run on another computer (this is called failover). Also functions as a fault-tolerant system.

[0004] In this cluster system, a computer having a relatively low performance is connected to a LAN (Local Area).
The system has recently attracted attention in terms of cost because high performance and high reliability can be obtained with a device configuration such as connection using a network.

In this cluster system, communication for confirming normal operation of each other is performed at predetermined intervals (this is called a heartbeat), and when the normal operation of the other party cannot be confirmed, that is, When the failure of the partner is detected, the system resources that have been operating on the partner computer are taken over and executed on the own computer, thereby realizing fault tolerance.

[0006]

As described above, a cluster system in which a plurality of computers are loosely coupled to each other to back up each other can construct a high-performance and highly reliable system at a relatively low cost.

In the conventional cluster system, it is only necessary to set a partner who executes heartbeat with each other and which system resource of the partner is to be failed over. Therefore, when the partner computer fails, all the set system resources are failed over regardless of the operation status of the own computer, and as a result, the high priority which originally operated on the own computer is used. In some cases, system resources were adversely affected. In addition, a low-priority system resource originally operating on its own computer continues to operate unconditionally, and in some cases, a high-priority system resource to be failed over cannot be started.

The present invention has been made in view of such circumstances, and an object of the present invention is to provide a cluster system that performs appropriate failover in accordance with the operating status of a failover destination and a failover control method for the system. .

[0009]

According to the present invention, in order to achieve the above-mentioned object, a plurality of computers are connected via a network, and when one of the plurality of computers fails, In a cluster system in which the system resources operating on that computer are taken over and operated on another computer,
A failover control means for controlling a change in priority including suspension of the system resources to be taken over in accordance with an operation state of the other computer.

In the cluster system according to the present invention,
When one of the computers performing the heartbeat fails, the system resources that were operating on the failed computer and that were set to be failed over are replaced with the system resources as in the past. In order to change the priority according to the operating status of the failover destination computer (in some cases, it does not start) instead of mechanically failing over all with the same priority, it originally operates on the failover destination computer There is no adverse effect on the prioritized system resources.

Also, in the cluster system according to the present invention, when the failover control means executes the takeover of the system resources, the priority change including suspension of the system resources originally operating on the other computer is performed. Is controlled in accordance with the operation status of the other computer.

In the cluster system according to the present invention,
To change the priority of the system resources originally operating on the failover destination computer when the system resources are failed over from the failed computer to the failover destination computer according to the operating status of the failover destination computer (It may be stopped in some cases), a situation in which a low-priority system resource that originally operated on the failover destination computer continues to operate unconditionally and a high-priority system resource to be failed over cannot be started. No cause.

[0013]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a configuration of a cluster system according to this embodiment. As shown in FIG. 1, the cluster system according to this embodiment includes a public L
Two server computers 10a to 10a connected to AN1
b are loosely coupled by an interconnect LAN 2 and the two server computers 10a
-B share data using a shared disk 20 connected by SCSI / FC (scogy / fiber channel) 3 together.

The cluster system according to this embodiment performs data processing requested by a client computer via the public LAN 1 on server computers 10a to 10b.
Is a distributed system that is executed in a distributed manner.
The resource 1 group operates on the side (here, this is node 1), while the resource 2 group operates on the server computer 10b (here, this is node 1) side.

The server computers 10a-b
Both of them execute communication (heartbeat) using the interconnect LAN 2 for confirming the normal operation of each other at predetermined intervals.

The execution of the heartbeat and the execution of a failover, which will be described later, are loaded from the shared disk 20 into the system memory mounted on the server computers 10a and 10b, and the execution is controlled by the CPU mounted on the server computers 10a and 10b. This is performed by an operating system or a program operating under the operating system.

The resource group 1 and the resource group 2 executed by the server computers 10a to 10b are also converted from the files on the shared disk 20 to the server computers 10a to
b is loaded into a system memory mounted on the server computer 10a, and is configured as a program (executed under an operating system) whose execution is controlled by a CPU mounted on the server computers 10a and 10b. Then, the failover of each of all the system resources belonging to either the resource group 1 or the resource group 2 is controlled by the resource table stored on the shared disk 20.

FIG. 2 is a diagram showing an example of the resource table according to this embodiment. As shown in FIG. 2, the resource table according to the present embodiment includes two items of “resource name” and “failover priority”, and is used when the failover of each system resource indicated in the “resource name” column is executed. Is shown in the column "Failover priority".

For example, "network name" and "I
The “P address” is set to “always execute” when the failover is performed. In this case, if the server computer on which the “network name” or “IP address” operates fails, the other server computer is activated. Regardless of the situation, the failover is always performed with the same priority.Also, “application 1” and “app 2” are set to “reduce the execution priority” when the failover is performed.
In this case, when the server computer on which “App 1” or “App 2” operates fails, if the operating status of the other server computer does not exceed a predetermined condition, the priority is kept as it is, and , The failover is executed after the priority is lowered. Further, the handling of the “application 3” at the time of executing the failover is set to “do not failover”. In this case, when the server computer on which the “application 3” operates fails, the operation status of the other server computer is changed. Regardless, failover is not always performed.

The predetermined condition used when the handling at the time of executing the failover is set to “lower the execution priority” is the shared disk 20.
It is managed by the condition table stored above.

FIG. 3 is a diagram showing an example of the condition table of this embodiment. As shown in FIG. 3, the condition table of this embodiment includes two items, “check item” and “set value”. For each of the items shown in the “check item” column, the failover destination server computer Are shown in the "set value" column.

Therefore, "application 1" and "application 2" have the same priority when the operation status of the server computer at the failover destination satisfies the items of the swap file size, the CPU usage rate, and the memory usage. Failover is performed, and if not satisfied, the priority is lowered and then failover is performed.

Here, the operation procedure of the cluster system of this embodiment will be described with reference to FIGS. FIG.
9 is a flowchart for explaining an operation procedure when a system resource operating on the failed computer is failed over.

If any one of the plurality of computers fails, the cluster system of this embodiment executes the following processing for the system resources operating on the failed computer.

First, in the cluster system of this embodiment, the priority of the system resource is acquired by referring to the resource table (step A1), and it is first determined whether or not the priority is set as "always executed" (step A).
2). Then, if it is set as "always executed" (YES in step A2), the failover of the system resource is executed (step A3).

On the other hand, if it is not set as "always executed" (NO in step A2), then it is determined whether or not "lower execution priority" is set (step A4). If "lower execution priority" is set (YES in step A4), the operation status of the server computer at the failover destination is compared with the condition table (step A5), and the condition is not exceeded. In this case (NO in step A6), the failover is executed with the same priority, and when the condition is exceeded (YES in step A6), the priority is lowered and the failover is executed (step A7). ).

If it is not set as "lower execution priority" (N in step A4)
O), it is determined that "fail-over" is not set, and no failover is performed.

FIG. 5 is a flowchart for explaining an operation procedure for managing system resources originally operating on the failover destination computer. When a failover occurs, the cluster system of this embodiment executes the following processing for the system resources originally operating on the failover destination computer.

First, in the cluster system of this embodiment, the priority of the system resource is acquired by referring to the resource table (step B1), and it is first determined whether or not the priority is set as "always executed" (step B).
2). If it is set as "always executed" (YES in step A2), the execution of the system resource is continued without performing any processing.

On the other hand, if it is not set to "always execute" (NO in step B2), the operation status of the failover destination server computer is compared with the condition table (step B3), and the condition is exceeded. If not (NO in step B4), execution is continued without performing any processing for the system resource. If the condition is exceeded (YES in step B4), it is further determined whether or not the setting is made as "lower the execution priority" (step B5), and the setting is made as "lower the execution priority". In this case (YES in step B5), execution is continued after lowering the priority (step B6).
If it is not set as "lower execution priority" (NO in step B5), it is determined that "failover is not set" and the execution of the system resource is ended (step B7).

As described above, according to the cluster system of this embodiment, according to the operation status of the computer at the failover destination, the system resources that have been operating on the failed computer are set to execute the failover. Both the system resources that have failed and the system resources that originally operate on the computer at the failover destination are appropriately controlled.

[0032]

As described above in detail, according to the present invention, when one of a plurality of computers that execute heartbeat fails, the system resources operating on the failed computer Instead of mechanically failing over all system resources set to be failed over with the same priority as before, the priority is changed according to the operating status of the failover destination computer Because of this (it is not started in some cases), the high-priority system resources originally operating on the failover destination computer are not adversely affected.

Further, when system resources are failed over from a failed computer to the failover destination computer, priority is given to the system resources originally operating on the failover destination computer according to the operating status of the failover destination computer. System resources that were originally running on the failover destination computer continue to run unconditionally, causing higher-priority system resources to be failed over. It does not cause a situation that it cannot be started.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of a cluster system according to an embodiment of the present invention.

FIG. 2 is an exemplary view showing an example of a resource table according to the embodiment.

FIG. 3 is an exemplary view showing an example of a condition table according to the embodiment;

FIG. 4 is an exemplary flowchart for explaining an operation procedure at the time of failing over system resources operating on the failed computer of the embodiment;

FIG. 5 is an exemplary flowchart for explaining an operation procedure when managing system resources originally operating on the failover destination computer according to the embodiment;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Public LAN 2 ... Interconnect LAN 3 ... SCSI / FC 10a-b ... Server computer 20 ... Shared disk

Claims (4)

[Claims] 1. A plurality of computers are connected via a network, and when one of the plurality of computers fails, the system resources running on that computer are taken over by another computer. A cluster system, comprising: a failover control unit that controls a change of a priority including a stop of the system resource to be taken over according to an operation state of the other computer. 2. The failover control means according to claim 1, wherein, when executing the takeover of the system resources, the change of the priority including the suspension of the system resources originally operating on the other computer is activated by the operation of the other computer. 2. The cluster system according to claim 1, further comprising means for controlling according to a situation. 3. A plurality of computers are connected via a network, and when one of the plurality of computers fails, the system resources running on that computer are taken over by another computer. A failover control method for a cluster system, wherein the change of priority including the stop of the system resource to be taken over is controlled according to the operating status of the other computer. 4. The failover control method according to claim 3, wherein when the system resources are taken over, a change of a priority including a stop of a system resource originally operating on the other computer is controlled.