JP2001125878A - Cluster type computer system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a cluster type computer system in which even when any failure occurs in a host which should execute timer announcement, the other host can execute the timer announcement instead. SOLUTION: Each host is provided with a timer registering means for receiving information (timer announcement information) related with timer announcement being a request for the start of processing to the designated destination at a designated time, the other host monitoring means for monitoring whether or not the other host can normally operate, and for changing the value of an execution flag indicating whether or not the host can execute timer announcement, a timer announcement table for storing the execution flag and the announcement information, and a timer communicating means to be called from an on-line real time processor at the designated time for executing the timer announcement according to the execution flag and the timer announcement table.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a cluster-type computer system, and more particularly to a cluster-type computer system in which, even if a failure occurs in a host for which timer notification is to be performed, another host can execute the function on behalf of the host. It relates to a computer system.

[0002]

2. Description of the Related Art There is a cluster computer system in which a plurality of host computers exist in the system. As a conventional cluster-type computer system, for example, Japanese Patent Application Laid-Open No. 8-6910
And a computer system disclosed in Japanese Patent Application Laid-Open No. 8-314875.

[0003] The cluster type computer system described in Japanese Patent Application Laid-Open No. Hei 8-6910 has a working host and a standby host, which are interconnected. The standby host confirms the presence or absence of the operation of the active host based on a message (alive message) issued from the active host at regular intervals. That is, when a message cannot be received even after a certain period of time has elapsed, it is determined that a failure has occurred in the active host.

When a failure occurs in the active host, the standby host changes the operating state of the active host or receives processing information of the process from the active host. On behalf of Such processing is intended to prevent the entire system from going down due to a failure of some of the hosts.

The cooperative distributed processing apparatus described in Japanese Patent Application Laid-Open No. 8-314875 includes a plurality of hosts, each of which is connected to each other, and mutually monitors whether or not a failure has occurred in each host. While running the process.
When a failure occurs in a plurality of connected hosts, the other hosts take in the processes to be executed by the failed host into the processes to be executed by themselves, and reorganize and execute the processes.

For example, suppose that a certain host X executes a process A, and another host Y executes a process B and a process C. At this time, when a failure occurs in the host X, the host Y, which is another host, takes in the process A to be executed by the host X and reorganizes the process. The rearranged processes are, for example, process B, process A, and process C. And host Y
Executes the reorganized process, and the host X
Can perform the process to be performed.

When a failure occurs in a part of a plurality of connected hosts due to such processing, another host takes over a process to be executed by the failed host, thereby improving the reliability of the entire system. Is being improved.

[0008]

In the cluster computer system described in Japanese Patent Application Laid-Open No. Hei 8-6910, when a failure occurs in the active host, the standby host executes the process to be executed by the active host on behalf of the active host. However, there is a problem in that the process up to the point is complicated, and it takes time from the occurrence of a failure to the recovery. Further, in the cooperative distributed processing apparatus described in Japanese Patent Application Laid-Open No. 8-314875, there is a problem that it takes time to reorganize a process, and it takes time to recover after a failure occurs.

Accordingly, the present invention provides a cluster-type computer system in which even if a failure occurs in a connected host, another host can quickly perform a process to be performed by the host. Aim.

[0010]

According to a first aspect of the present invention, there is provided a cluster type computer system comprising a plurality of hosts,
In a cluster-based computer system capable of transmitting information between hosts, each host shares timer information, which is information on a timer notification that is a processing start request for a specified host at a specified time, so that a timer notification is transmitted. When a failure occurs in the host to be executed, another host substitutes for the timer notification.

In the cluster-type computer system according to the second aspect of the invention, each host sets timer information in the timer information storage means, and sets timer information in the timer information storage means when a timer notification request occurs. A configuration is provided that includes timer registration means for transmitting timer information to another host.

In the cluster type computer system according to the third aspect of the present invention, the timer information includes a notification time indicating a time at which the timer notification is executed, a notification destination indicating a destination at which the timer notification is executed, and the host executing the timer notification. This is an execution flag indicating whether or not to execute.

In the cluster type computer system according to the present invention, the timer registering means in the host which has received the timer notification request sets the execution flag to the information for the host to perform the timer notification, and sets the execution flag in the other host. The timer registration unit sets information indicating a host that performs timer notification in the execution flag.

In the cluster type computer system according to the present invention, each host monitors whether or not a failure has occurred in another host, and determines whether or not the host should execute a timer notification. Means.

In the cluster computer system according to the present invention, when the other host monitoring means determines that the own host should execute the timer notification in place of the timer notification, the other host monitoring means transmits the execution flag setting information to the own host. Change to the information for timer notification.

[0016]

Embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a cluster computer system according to the present invention. In the cluster computer system shown in FIG. 1, a plurality of hosts 101 to 10n are connected, and information can be transmitted between the hosts. In each of the hosts 101 to 10n, the online real-time processing apparatuses 111 to 11n include timer registration units 131 to 13n, and other host monitoring units 141 to
14n, timer notification means 151 to 15n and timer notification request registration tables 121 to 12n. Each host 1
Timer registration means 131 to 13n in 01 to 10n,
The other host monitoring means 141 to 14n and the timer notification means 151 to 15n are mutually connected between the hosts.
Here, the host number of the host 101 is set to 1, and the host 1
The host number of 02 is 2, and the host number of the host 10n is n.

In the host 101, the timer registration means 131
Receives the timer notification request and stores the information in the timer notification request registration table 121 together with the execution flag.
The execution flag will be described later in detail. The timer notification request is a request to execute a timer notification. The timer notification is a processing start request to a designated destination (host) at a designated time. Therefore, the host that has received the timer notification starts a predetermined process.

The other host monitoring means 141 determines whether or not a failure has occurred in a host connected to another, and changes the value of an execution flag which is a flag indicating whether or not the host executes timer notification. Then, the online real-time processing device 111 calls the timer notification unit 151 at the time of executing the timer notification. Then, the timer notification unit 151 stores the timer notification request registration table 121
A timer notification is performed based on the information stored in. The other hosts have the same configuration as the host 101.

FIG. 2 shows the timer notification request registration table 12
It is explanatory drawing which shows an example of the information stored in 1-12n. FIG. 3 is a flowchart for explaining the operation of the timer registration units 131 to 13n. FIG. 4 is a flowchart for explaining the operation of the timer notification means 151 to 15n. FIG. 5 shows another host monitoring means 141 to 14.
4 is a flowchart for explaining the operation of the n.

The operation of the cluster computer system shown in FIG. 1 will be described below with reference to the flowcharts of FIGS. This cluster computer system is used, for example, in a bank host computer system.

It is assumed that a timer notification request has been issued to the host 101. That is, the host 101 becomes the reception host. The information of the timer notification request includes notification time information that can specify when the timer notification is to be performed and a notification destination that indicates to which host the timer notification is performed. In this example, the time specified from the notification time information is 10:00,
The notification destination is the host 104. The time when the host 101 receives the timer notification request is set to 9:00 on the same day.

The timer registration unit 131 determines whether the timer notification request is for distribution (step S30).
1). In this example, since the host 101 is the receiving host, the determination here is NO. Here, distribution means that when a certain host receives a timer notification request, the host notifies other hosts of the information of the timer notification request.

Next, the timer registration means 131 calculates the notification time from the timer notification request (step S30).
2). Then, it is determined whether the designation of the notification time is correct (step S303). The notification time must always be specified in the future. Therefore, for example, when a past time is specified, the determination in step S303 is NO.
And the subsequent processing ends. In this example, the time at which the timer notification request was received is 9:00, and the time specified from the notification time information is 10:00, so the determination in step S303 is YES.

Next, the timer registration means 131 determines whether or not the designation of the notification destination is correct (step S30).
4). The notification destination must specify the connected host. Therefore, for example, if a host that is not connected is specified, the determination in step 304 is NO, and the subsequent processing ends. In this example, host 1
04 is connected, the determination in step S304 is YES.

Next, the timer registration unit 131 stores the notification time and the notification destination in the timer notification request registration table 121 (step S305). Then, the value of the execution flag is changed to 0, and the timer notification request registration table 121 is changed.
(Step S306). That is, the information stored in the timer notification request registration table 121 (hereinafter referred to as notification information) is as shown in FIG. Here, the notification information includes a notification time, a notification destination, and an execution flag.

Next, the timer registration means 131 distributes the notification information stored in the timer notification request registration table 121 to another host (step S30).
7). Therefore, other hosts also hold the notification information. Therefore, when a failure occurs in the host 101, another host can substitute for a timer notification to be executed by the host 101.

For example, in the host 102, the timer registration unit 132 stores the notification information distributed from the host 101 in the timer notification request registration table 122 (Step S308). Then, the value of the execution flag is changed to the reception host number, and stored in the timer notification request registration table 122 (step S309). In this example, host 101
Is the accepting host, the value of the execution flag is changed to 1. That is, the notification information stored in the timer notification request registration table 122 is as shown in FIG. The operations of the other hosts are the same as the operations of the host 102.

In the host 101, the online real-time processing device 111 calls the timer notification means 151 when the notification time comes. Then, the timer notification means 151
Determines whether the value of the execution flag is 0 (step S401). When the value of the execution flag is 0, the host executes the timer notification. That is, in this example, the determination in step S401 is YES, and a timer notification is executed (step S402). That is, in this example, at 10:00, the timer notification unit 151 is called from the online real-time processing device 111 and
04 is notified of a timer.

Next, the host 10 that should execute the timer notification
The case where a failure has occurred in No. 1 will be described with reference to the flowchart in FIG. The hosts 101 to 10n mutually monitor whether a failure has occurred in the host. For example, the host 102 is taken as an example. Other host monitoring means 142
Notifies the connected host of a message (alive message) for confirming whether or not a failure has occurred in the host (step S501).

Then, it is determined whether or not responses have been received from all the hosts (step S502). If there is a response from all the hosts (in the case of NO), since all the hosts are operating normally, the subsequent processing ends. In this example, a failure has occurred in the host 101, and there is no response from the host 101. Therefore, step S50
The determination at 2 is YES.

Next, it is determined whether the host that has not received a response is a host that should execute timer notification (step S503). If the host is not the one that should execute the timer notification (NO), the subsequent processing ends. In this example, the determination in step S503 is YES because a failure has occurred in the host 101 that should execute the timer notification.
Becomes

Next, it is determined whether or not the host number of its own host is the next number of the host number for which the timer notification is to be executed (step S504). Here, the next number is a number obtained by adding 1 to the host number of the host that should execute the timer notification, or the last number (n in this example, n is the host number of the host that should execute the timer notification). ) Is 1.

In this example, the host number of the host that should execute the timer notification is 1 and its own host number is 2, so the determination in step S504 is YES.
Then, the value of the execution flag is changed to 0 and stored in the timer notification request registration table 122 (step S505).
That is, the notification information stored in the timer notification request registration table 112 is set as shown in FIG.

Online real-time processor 112
Calls the timer notification means 152 when the notification time comes. The timer notification unit 152 determines that the value of the execution flag is 0.
Is determined (step S401). Since the value of the execution flag is 0, a timer notification is executed to the host 104. Therefore, the host 102
The timer notification that 1 should perform can be substituted.

The other host operations are performed in step S50.
Steps 1 to S503 are the same as those of the host 102. However, the determination in step S504 is NO.
Here, the host 103 is taken as an example of another host. In the host 103, in step S506, the other host monitoring unit 143 changes the value of the execution flag to the host number of the host that performs the timer notification. In this example,
Since the host 102 substitutes for the timer notification, the value of the execution flag is changed to 2. Then, the notification information stored in the timer notification request registration table 123 is as shown in FIG. Then, the timer notification unit 153 is called from the online real-time processing device 103 when the notification time comes. However, since the value of the execution flag is not 0, the timer notification unit 153 does not perform the timer notification to be executed by the host 101. The operations of the other hosts 104 to 105 are the same as those of the host 103.

Here, a case in which a failure has occurred in the host 102 acting as a substitute for the timer notification to be executed by the host 101 will be described. Take the host 103 as an example. Step S
The operations from 501 to S503 are the same as those of the host 102.

The timer notification means 153 determines whether or not its own host number is the next number of the host number of the host to execute the timer notification (step S504). In the previous step S506, the timer notification request registration table 12
The notification information stored in No. 3 is as shown in FIG. Since the host number of the host to be notified of the timer is 2 and its own host number is 3,
The determination in step S504 is YES.

Then, the timer notification means 153 changes the value of the execution flag to 0 and stores it in the timer notification request registration table 123. That is, the notification information stored in the timer notification request registration table 123 is changed as shown in FIG. When the notification time comes, the timer notification means 1
53 is called from the online real-time processing device 113 and executes timer notification. Therefore, the host 10
3 can substitute for a timer notification to be executed by the host 101.

As described above, each of the hosts 101 to 10n
Stores the notification information in the timer notification request registration tables 121-121.
12n to be shared. The other host monitoring means 141 to 14n determine whether or not a failure has occurred in another host, thereby determining whether or not the host should substitute for the timer notification. change. Then, the timer notification means 151 to 15n called by the online real-time processing devices 111 to 11n execute the timer notification based on the notification information stored in the timer notification request registration tables 121 to 12n. Therefore, even if a failure occurs in the host to which the timer notification should be made, another host can substitute the timer notification by referring to the notification information. Further, even when a failure occurs in a plurality of hosts, another host can substitute for the timer notification.

In addition, even if a failure occurs in the host that should execute the timer notification, another host can immediately substitute for the timer notification. Can solve the problem of the technology.

[0041]

According to the first aspect of the present invention, in a cluster-type computer system including a plurality of hosts and capable of transmitting information between hosts, a processing start request to a designated host at a designated time is issued. By sharing the timer information, which is information related to a certain timer notification, with each host, when a failure occurs in the host that should execute the timer notification, another host can perform the timer notification on behalf of the host.

According to the second aspect of the present invention, each host has a timer information storage means for storing timer information, and sets a timer information in the timer information storage means when a request for timer notification is generated, and transmits the timer information to another host. On the other hand, since the configuration includes the timer registration means for transmitting the timer information, all the hosts can share the timer information.

According to the third aspect of the present invention, the timer information includes a notification time indicating a time at which the timer notification is performed, a notification destination indicating a destination at which the timer notification is performed, and whether or not the host performs the timer notification. Therefore, each host can execute the timer notification to the notification destination at the notification time according to the execution flag.

According to the fourth aspect of the present invention, the timer registering means in the host that has received the timer notification request sets the execution flag to the information on which the own host performs the timer notification. Since information indicating the host that performs the timer notification is set in the execution flag, another host has the effect of being able to recognize the host that performs the timer notification.

According to the fifth aspect of the present invention, each host is provided with another host monitoring means for monitoring whether or not a failure has occurred in another host, and determining whether or not the host should execute timer notification. Therefore, each host can determine whether or not a failure has occurred in the other connected host, and has an effect that it can determine whether or not the host executes the timer notification.

According to the sixth aspect of the present invention, when the other host monitoring means determines that the own host should execute the timer notification in place of the timer notification, the own host performs the timer notification of the setting information of the execution flag. Since the information is changed, even if a failure occurs in the host to which the timer notification is to be made, another host can perform the timer notification by changing the value of its own execution flag.

[Brief description of the drawings]

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

FIG. 2 shows timer notification request registration tables 121 to 12n.
FIG. 5 is an explanatory diagram showing an example of notification information stored in a storage device.

FIG. 3 is a flowchart for explaining operations of timer registration units 131 to 13n.

FIG. 4 is a flowchart for explaining the operation of timer notification means 151 to 15n.

FIG. 5 is a flowchart for explaining the operation of other host monitoring means 141 to 14n.

[Explanation of symbols]

 101 to 10n Host 111 to 11n Online real-time processing device 121 to 12n Timer notification request registration table 131 to 13n Timer registration means 141 to 14n Other host monitoring means 151 to 15n Timer notification means

Claims (6)

[Claims] 1. A cluster-type computer system comprising a plurality of hosts and capable of transmitting information between hosts, wherein timer information, which is information relating to a timer notification which is a processing start request to a specified host at a specified time, is transmitted. A cluster-type computer system characterized in that, when a host that should execute timer notification fails due to sharing by each host, another host substitutes for timer notification. 2. A host, comprising: a timer information storage unit for storing timer information; and a timer for setting the timer information in the timer information storage unit when a timer notification request occurs, and transmitting the timer information to another host. 2. The cluster-type computer system according to claim 1, further comprising registration means. 3. The timer information is a notification time indicating a time at which the timer notification is executed, a notification destination indicating a destination at which the timer notification is executed, and an execution flag indicating whether the host executes the timer notification. The cluster-type computer system according to claim 1 or 2. 4. The timer registering means of the host which has received the request for timer notification sets information for the host to perform timer notification in the execution flag, and the timer registration means of the other host transmits the timer to the execution flag. 4. The cluster-type computer system according to claim 3, wherein information indicating the following is set. 5. Each host further comprises another host monitoring means for monitoring whether or not a failure has occurred in another host and determining whether or not the host should execute timer notification.
The cluster-type computer system as described. 6. The other host monitoring means changes the setting information of the execution flag to information on which the own host issues a timer notification when the own host determines that the timer notification should be executed instead. Cluster computer system.