JP2002123406A - High reliability system - Google Patents

Abstract

(57) [Summary] The present invention appropriately addresses failure occurrence when employing a mirroring configuration in which a disk of an active node is used as an original and a disk of a standby node connected via a network is used as a shadow. The purpose is to be able to cope. When a failure occurs in a disk of an active node, a node operating as a standby system is brought into an operating state, and a node that has been operating as an active system is stopped so that a redundant system as a redundant system is obtained. Process to stop the operation. Further, when a failure occurs in the disk of the standby node, processing is performed so as to stop the operation as the redundant system by stopping the node operating as the standby system. Further, when a failure occurs in the network prepared for accessing the shadow disk, the node operating as the standby system is stopped to stop the operation as the redundant system.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention has a mirror set function in which a local disk of an active node is used as an original disk and a local disk of a standby node connected via a network is used as a shadow disk.
When a failure occurs in one of the disks, a highly reliable system that performs business by using the other disk that works normally, and in particular, a highly reliable system that can appropriately cope with the failure About.

[0002]

2. Description of the Related Art A highly reliable system composed of a server that performs business and a server that stands by in preparation for a failure of the server is widely used. The former is called a working node, and the latter is called a standby node.

If the active node goes down due to some kind of failure, the standby node takes over the business. In order to take over the business, it is necessary to transfer business data from the active node to the standby node.

Until now, in order to take over the business data, a SCSI node has been connected between the active node and the standby node.
In this configuration, a shared disk device connected by an interface or the like is prepared.

[0005] However, the shared disk device is extremely expensive. Therefore, a mirror set function (RAID1 function) is used in which the local disk of the active node is used as an original disk and the local disk of the standby node connected via a network is used as a shadow disk. When a failure occurs, a highly reliable system employing a configuration in which a normal disk is used to perform a task is being used.

In this highly reliable system, as shown in FIG. 5, the active node writes the business data to the shadow disk using a net disk driver when writing the business data to the original disk.
It adopts a configuration to execute mirroring.

When the original disk is normal, the active node performs the job by accessing the original disk, and when the original disk fails, the active node performs shadowing using the net disk driver. The configuration is such that the business is performed by accessing the disk.

[0008]

However, this prior art has the following problems.

That is, when a failure occurs in the original disk, the active node performs all input and output of business data to and from the shadow disk, thus deteriorating the data input performance. However, there is a problem that the performance is deteriorated.

Further, when a failure occurs in the shadow disk, this prior art does not employ a configuration in which the standby node is stopped.

[0011] From now on, if the active node goes down when the active node performs a task using the original disk after the failure of the shadow disk, the user will be asked to enter the standby node. However, there is a problem that if the work is supposed to be taken over but is not executed, there is anxiety about the reliability of the system.

In addition, in the event of a system failure of the active node, even if it is possible to use a shadow disk by avoiding a failure point, the business data of the shadow disk must not be updated. From
New active node (node with the shadow disk)
However, there is a problem that the latest business data cannot be taken over.

In addition, even when a failure occurs in a network (dedicated network in FIG. 5) prepared for the active node to access the shadow disk, the prior art has a configuration in which the standby node is stopped. Not taken.

From this, there is a problem similar to that when a failure occurs in the shadow disk.

The present invention has been made in view of such circumstances, and has a mirror set function in which a local disk of an active node is used as an original disk, and a local disk of a standby node connected via a network is a shadow disk. When a failure occurs in one of the disks and a configuration is adopted in which the business is performed by using the other normal disk, the failure can be appropriately dealt with. The purpose is to provide a new highly reliable system.

[0016]

In order to achieve this object, a high reliability system according to the present invention uses a local disk of an active node as an original disk and a local disk of a standby node connected via a network. When a configuration is adopted in which a mirror set function is used as a shadow disk and one of the disks fails, the business is performed by using the other normal disk. Detecting means for detecting whether a failure has occurred in the original disk, switching means for switching the standby node to the operating state when the failure detection of the original disk is detected by the detecting means, and node switching by the switching means. The redundant system by stopping the node that was operating as the active node until then. Configured to and a stop means for.

In the high-reliability system of the present invention configured as described above, when a failure occurs in the hard disk of the node operating as the active system, the node operating as the standby system is brought into the operating state, and By stopping the node that has been operating as the active system until then, processing is performed to stop the operation as the redundant system.

Thus, the node that newly executes the business process can execute the business using the hard disk of its own node, so that the data input performance does not deteriorate and the performance of the entire system is reduced. Can be prevented. Then, since the operation as the duplex system is stopped, no inconvenient operation is performed thereafter.

In order to achieve this object, a high reliability system according to the present invention uses a local disk of an active node as an original disk and a local disk of a standby node connected via a network as a shadow disk. If one of the disks has a failure and the other disk fails, the shadow disk fails And a stopping unit for stopping the redundant system by stopping the standby node when the detection of the failure of the shadow disk is detected by the detecting unit.

In the high reliability system of the present invention configured as described above, when a failure occurs in the hard disk of the node operating as the standby system, the node operating as the standby system is stopped to thereby provide a redundant system. Is processed to stop the operation as.

Thus, the node executing the business process can inform the user that the standby node is not prepared, so that the user is concerned about the reliability of the system. And the inconvenience of taking over the business with old business data does not occur.

To achieve this object, the high reliability system of the present invention uses a local disk of an active node as an original disk and a local disk of a standby node connected via a network as a shadow disk. When one of the disks has a failure and one of the disks fails, the working node is used to perform the job by using the other normal disk. Detection means for detecting whether or not a failure has occurred in the shadow network prepared for accessing the network, and duplication by stopping the standby node when the detection means detects the failure of the shadow network. And a stopping means for stopping the system.

In the high reliability system of the present invention configured as described above, when a failure occurs in the shadow network, the operation of the redundant system is stopped by stopping the node operating as the standby system. To process.

Thus, the node executing the business process can inform the user that the standby node is not prepared, so that the user is concerned about the reliability of the system. And the inconvenience of taking over the business with old business data does not occur.

Thus, according to the present invention, the local disk of the active node is set as the original disk,
It has a mirror set function that uses the local disk of the standby node connected via the network as a shadow disk, so that if one of the disks fails, the other normal disk can be used. When adopting a configuration of performing a task, it becomes possible to appropriately cope with the occurrence of a failure.

[0026]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail according to embodiments.

FIG. 1 shows a system configuration of a high reliability system to which the present invention is applied.

As shown in the figure, a high reliability system to which the present invention is applied includes an active node 1 for executing business processing, a standby node 2 prepared as a standby system of the active node 1, A dedicated network 3 for connecting the active node 1 and the standby node 2, a business terminal 4 operated by a user, and a connection between the active node 1 and the standby node 2 and a connection for the active node 1 / A business network 5 connecting the standby node 2 and the business terminal 4.

Here, the dedicated network 3 is used as a communication line when the active node 1 accesses the hard disk of the standby node 2, and is exchanged between the active node 1 and the standby node 2. It is used as a communication line for alive notification, and is preferably duplicated.

The active node 1 includes a hard disk 10 functioning as an original disk for mirroring.
An application program 11 for executing business processes using the hard disk 10, a net disk driver 12 for accessing the hard disk 20 of the standby node 2, a mirror driver 13 for executing mirroring for the hard disks 10, 20. And a cluster control unit 14 for executing the control processing.

On the other hand, the standby node 2 has a hard disk 2 functioning as a shadow disk for mirroring.
0 and application program 1 of active node 1
1, an application program 21 prepared in association with the active node 1, a net disk driver 22 for accessing the hard disk 10 of the active node 1, a mirror driver 23 for executing mirroring on the hard disks 20, 10, and an overall control process. And a cluster control unit 24 to execute.

FIG. 2 shows an embodiment of the present invention which functions when a failure occurs in the hard disk 10 of the active node 1.

In the figure, the active node 1 is “#node 1”, the standby node 2 is “#node 2”, the hard disk 10 of the active node 1 is “HD1”, and the hard disk 20 of the standby node 2 Is indicated by “HD2”.

Next, the processing executed by the present invention when a failure occurs in the hard disk 10 of the active node 1 will be described with reference to FIG.

(1) Hard disk 1 of active node 1
It is assumed that some trouble occurs in 0.

(2) When the application program 11 of the active node 1 issues a business data write request to perform business processing when this failure has occurred,
(3) The mirror driver 13 of the active node 1 writes business data requested to be written to the hard disk 10 of the own node in order to execute mirroring.
Using the net disk driver 12, the business data requested to be written is written to the hard disk 20 of the standby node 2. At this time, the mirror driver 13 of the active node 1 detects that a failure has occurred in the hard disk 10 of the own node because the mirror driver 13 cannot execute the writing to the hard disk 10 of the own node.

(4) Mirror driver 1 of active node 1
3 detects that a failure has occurred in the hard disk 10 of its own node, and notifies the cluster control unit 14 of its own node to that effect.

(5) Cluster controller 1 of active node 1
4 receives this notification, (6) stops the application program 11 of its own node, and then transitions its own node from the operating state to the stopped state. (7) Subsequently, the mirror driver 13 of its own node is By stopping the operation, the operation as the redundant system is stopped.

(8) Subsequently, the cluster control unit 14 of the active node 1 that has stopped operating as the active system uses the business network 5 to operate the cluster control unit 24 of the standby node 2 in the operating state. Notify the switch to.

(9) Upon receiving this notification, the cluster control unit 24 of the standby node 2 changes its own node from the stopped state to the operating state, and (10) subsequently activates the mirror driver 23 of its own node. (11) The application is restarted by activating the application program 21 of the own node.

Here, when the mirror driver 23 is started, a process of changing the mirroring configuration information is performed so that only the hard disk 20 of the own node can be accessed and the failed hard disk 10 is not accessed. .

As described above, according to the present invention, when a failure occurs in the hard disk of the node operating as the active system, the node operating as the standby system is brought into the operating state, and the node operating as the active system has been operated until then. By stopping the existing node, processing is performed to stop the operation as a redundant system.

As a result, the node that newly executes the business process can execute the business using the hard disk of the own node, so that the data input performance does not deteriorate and the performance of the entire system is reduced. Can be prevented. Then, since the operation as the duplex system is stopped, no inconvenient operation is performed thereafter.

FIG. 3 shows an embodiment of the present invention which functions when a failure occurs in the hard disk 20 of the standby node 2.

In the figure, the active node 1 is “#node 1”, the standby node 2 is “#node 2”, the hard disk 10 of the active node 1 is “HD1”, and the hard disk 20 of the standby node 2 Is indicated by “HD2”.

Next, the processing executed by the present invention when a failure occurs in the hard disk 20 of the standby node 2 will be described with reference to FIG.

(1) Hard disk 2 of standby node 2
It is assumed that some trouble occurs in 0.

(2) When the application program 11 of the active node 1 issues a business data write request to perform business processing while this fault has occurred,
(3) The mirror driver 13 of the active node 1 writes business data requested to be written to the hard disk 10 of the own node in order to execute mirroring.
Using the net disk driver 12, the business data requested to be written is written to the hard disk 20 of the standby node 2. At this time, the mirror driver 13 of the active node 1 detects that a failure has occurred in the hard disk 20 of the standby node 2 because this writing cannot be performed on the hard disk 20 of the standby node 2.

(4) Mirror driver 1 of active node 1
3 detects that a failure has occurred in the hard disk 20 of the standby node 2 and notifies the cluster control unit 14 of the own node of the failure.

(5) Upon receiving this notification, the cluster control unit 14 of the active node 1 notifies the cluster control unit 24 of the standby node 2 of the truncation process using the business network 5, 6) In response to this, the cluster control unit 24 of the standby system node 2 stops the operation as the redundant system by making a transition from the standby state to the stop state.

As described above, according to the present invention, when a failure occurs in the hard disk of the node operating as the standby system, the operation of the redundant system is stopped by stopping the node operating as the standby system. Process as follows.

Thus, the node executing the business process can inform the user that the standby node is not prepared, so that the user is concerned about the reliability of the system. And the inconvenience of taking over the business with old business data does not occur.

FIG. 4 shows an embodiment of the present invention which functions when a failure occurs in the dedicated network 3.

In the figure, the active node 1 is “#node 1”, the standby node 2 is “#node 2”, the hard disk 10 of the active node 1 is “HD1”, and the hard disk 20 of the standby node 2 Is indicated by “HD2”.

Next, the processing executed by the present invention when a failure occurs in the dedicated network 3 will be described with reference to FIG.

(1) It is assumed that some failure occurs in the dedicated network 3.

(2) When the application program 11 of the active node 1 issues a business data write request to perform business processing while this fault has occurred,
(3) The mirror driver 13 of the active node 1 writes business data requested to be written to the hard disk 10 of the own node in order to execute mirroring.
Using the net disk driver 12, the business data requested to be written is written to the hard disk 20 of the standby node 2. At this time, the mirror driver 13 of the active node 1 detects that a failure has occurred in the dedicated network 3 because this write cannot be executed on the hard disk 20 of the standby node 2.

(4) Mirror driver 1 of active node 1
When detecting that a failure has occurred in the dedicated network 3, the node 3 notifies the cluster control unit 14 of its own node.

(5) Upon receiving this notification, the cluster control unit 14 of the active node 1 notifies the cluster control unit 24 of the standby node 2 of the truncation process using the business network 5, 6) In response to this, the cluster control unit 24 of the standby node 2 makes a transition from the standby state to the stop state, and (7) sends the hard disk 20 of the own node from the mirror set to the mirror driver 23 of the own node. By instructing disconnection, the operation as a redundant system is stopped.

(8) Upon receiving this instruction, the mirror driver 23 updates the mirror set configuration information stored in the hard disk 20 of the own node to a form in which the hard disk 20 cannot be used, and thereby the mirror driver 23 The hard disk 20 is separated from the mirror set.

As described above, according to the present invention, when a failure occurs in the dedicated network 3, by stopping the node operating as the standby system, the processing as the redundant system is stopped.

As a result, the node executing the business process can inform the user that the standby node is not prepared, so that the user is concerned about the reliability of the system. And the inconvenience of taking over the business with old business data does not occur.

[0063]

As described above, in the high reliability system according to the present invention, when a failure occurs in the hard disk of the node operating as the active system, the node operating as the standby system is put into operation and Then, processing is performed so as to stop the operation as the redundant system by stopping the node that has been operating as the active system until then.

As a result, the node that newly executes the business process can execute the business using the hard disk of its own node, so that the data input performance does not deteriorate, and the performance of the entire system decreases. Can be prevented. Then, since the operation as the duplex system is stopped, no inconvenient operation is performed thereafter.

Further, in the high reliability system of the present invention, when a failure occurs in the hard disk of the node operating as the standby system, the node operating as the standby system is stopped to operate as a redundant system. Process to stop.

As a result, the node executing the business process can inform the user that the standby node is not prepared, so that the user is concerned about the reliability of the system. And the inconvenience of taking over the business with old business data does not occur.

Further, in the high reliability system of the present invention, when a failure occurs in the shadow network prepared for accessing the shadow disk, the node operating as the standby system is stopped, so that the redundant system can be used. Process to stop the operation.

As a result, the node executing the business process can inform the user that the standby node is not prepared, so that the user is concerned about the reliability of the system. And the inconvenience of taking over the business with old business data does not occur.

As described above, according to the present invention, the local disk of the active node is set as the original disk,
It has a mirror set function that uses the local disk of the standby node connected via the network as a shadow disk, so that if one of the disks fails, the other normal disk can be used. When adopting a configuration of performing a task, it becomes possible to appropriately cope with the occurrence of a failure.

[Brief description of the drawings]

FIG. 1 is a system configuration diagram of a highly reliable system to which the present invention is applied.

FIG. 2 is an embodiment of the present invention.

FIG. 3 is an embodiment of the present invention.

FIG. 4 is an embodiment of the present invention.

FIG. 5 is an explanatory diagram of a high reliability system to which the present invention is applied.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Working node 2 Standby node 3 Dedicated network 4 Business terminal 5 Business network 10,20 Hard disk 11,21 Application program 12,22 Net disk driver 13,23 Mirror driver 14,24 Cluster control unit

 ────────────────────────────────────────────────── ─── Continuing from the front page (72) Inventor Hidekazu Imai 98 Uno-ki-nu, Unoki-cho, Kawakita-gun, Ishikawa Pref.

Claims (3)

[Claims] 1. A mirror set function in which a local disk of an active node is used as an original disk and a local disk of a standby node connected via a network is used as a shadow disk. In the event of occurrence, in a highly reliable system that performs business by using the other normal disk, means for detecting whether or not a failure has occurred in the original disk; Means for switching the standby node to the operating state when detected, and means for stopping the node that has been operating as the active node and stopping the redundant system in response to the node switching. High reliability system characterized by that. 2. A mirror set function in which a local disk of an active node is used as an original disk and a local disk of a standby node connected via a network is a shadow disk. In the event of occurrence, in a highly reliable system that performs business by using the other normal disk, a means for detecting whether or not a failure has occurred in the shadow disk; When detected,
Means for stopping the redundant system by stopping the standby node. 3. A mirror set function in which a local disk of an active node is used as an original disk and a local disk of a standby node connected via a network is used as a shadow disk. If this occurs, in a highly reliable system that performs business by using the other normal disk, has a failure occurred in the shadow network provided for the active node to access the shadow disk? A high reliability system comprising: means for detecting whether a failure has occurred in the shadow network, and means for stopping a redundant system by stopping a standby node when the occurrence of a failure in the shadow network is detected.