JP2001043105A - High-availability computer system and data backup method of the system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To actualize a system which is more tolerant of a fault by securing multiple backup servers and effectively copying data to the servers. SOLUTION: More than two servers connected by a network, e.g. servers S1 to S4 are prepared and the server S1 which functions as a master according to the priority communicates with the slave servers S2 to S4 to periodically search for a server having a fault and servers having no fault. When a client modifies data of a file that the master server S1 has, the server S1 takes copies 81 and 82 of the modified data to the faultless servers S3 and S4 found through the searching operation. The servers S3 and S4 periodically searches for the master in the decreasing order of the priority from the server S1 having the top priority and becomes a master newly when having the top priority itself among the faultless servers when no master is found.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-availability system in which a plurality of server computers cooperate to perform processing, and even if a failure occurs in any one of the server computers, another server computer can take over the processing. High availability computer system)
In particular, a plurality of server computers that provide services to other client computers are linked by a network, and even if a failure occurs in any one of the server computers, the other server computers take over the service and the entire system is taken over. The present invention relates to a high availability computer system having a data backup function for minimizing service interruption time as much as possible and a data backup method in the system.

[0002]

2. Description of the Related Art A conventional high availability computer system has two server computers, one of which provides a service,
It was common to apply a backup method to the other side.

In this type of system, data is copied from a server computer providing a service to a backup server computer, so that when a failure occurs in the server computer providing the service, the service is provided. Is transferred to the other server computer, and the continuation can be performed.

[0004]

In the conventional high-availability computer system configured using the two server computers described above, even if a failure occurs in the server computer providing the service, the remaining standby state is maintained. It is possible to take over the service by the server computer (backup server computer). However, when there are two server computers, there is a possibility that a failure may occur in both servers, so that it is not always sufficient in terms of fault tolerance depending on the application.

Therefore, in order to construct a system that is more resistant to failure, it is conceivable to operate three or more server computers in cooperation with each other. In this case, it is expected that the cooperation between the server computers becomes more complicated as the number of server computers increases, and that the load on the server computers in the operating state also increases. For this reason, a mechanism for effectively linking three or more server computers is required.

The present invention has been made in view of the above circumstances, and has as its object to provide a backup server computer (backup server computer) using three or more server computers.
Is to provide a highly available computer system that is more resistant to failure and a data backup method in the system by securing a plurality of backup servers and performing effective data copying to the plurality of backup server computers.

Another object of the present invention is to link a plurality of server computers via a high-speed network and a low-speed network, and to use a data backup method suitable for each network, so that effective data can be flexibly adapted to a network configuration. It is an object of the present invention to provide a high availability computer system capable of performing backup and a data backup method in the system.

[0008]

A high-availability computer system according to a first aspect of the present invention includes at least three server computers connected via a network, one of which is a master server computer. If a service is provided to the client computer and a failure occurs in the master server computer, predetermined priority information (here, the master priority is shown for all the computers in the system and the master is switched) A priority server in which each of the remaining server computers becomes a new master server computer and takes over the processing according to the priority information in which the priority is cyclically used for each server computer. The following means, ie, a master search for finding a master server computer when the own computer is not the master server computer Master search means for periodically executing the operation; and when the master server means cannot find the master server computer and has the highest priority among the server computers in which no failure has occurred, the master computer As a master server computer, and a server that periodically executes a server computer search operation for searching for a faulty server computer and a faultless server computer when the own computer is the master server computer Computer search means, and the own computer is a master server computer,
Further, when the data of the file held by the own computer is changed from the client computer, the changed data is individually copied to all the fault-free server computers found by the server computer searching means.
Copy means for performing data copy by the n-to-n communication method.

In such a configuration, when data of a file held in the master server computer is changed by the client computer, the data is copied by the master server computer to all other server computers having no trouble, and the data is copied to each server computer. File contents are matched,
Moreover, since there are a plurality of backup server computers (slave server computers) for the master server computer, if a failure occurs in the master server computer, a plurality of other backup server computers (slave server computers)
One of which becomes a new master server computer, can use the copied data to take over the service that has been provided by the failed server, and is more resistant to failure. Can be realized.

Here, the master searching means has a function of searching for the master server computer by performing communication in order from the server computer having the highest priority at that time in the direction of decreasing priority according to the priority information. With this, the master search can be performed efficiently.

Further, the server computer searching means communicates in order from the server computer which is one rank lower than the own computer to the server computer having the lower rank in accordance with the priority information, so that the faulty server computer can be identified with the faulty server computer. By providing a function of searching for a server computer having no server, it is possible to efficiently search for a server computer having a lower priority than the server computer itself.

In the case where the own computer is a master server computer, if there is provided data transmission destination setting means for setting a server computer having no fault found by the server computer search means as a data transmission destination, the copying means of the copying means is provided. Data copy by the 1: n communication method can be efficiently performed according to the setting of the data transmission destination setting means.

Here, if a new server computer having no failure is found by the server computer search means, that is, if a server computer recovered from the failure is detected, the server computer is set as a data transmission destination. If additional settings are made by the data transmission destination setting means and the data of all the files held by the master server computer are copied to the server computer by the copying means of the master server computer, the server computer (recovered server computer) is surely connected. In addition, it can quickly become one of the backup computers.

A high-availability computer system according to a second aspect of the present invention provides each server computer in a network-connected system with the following means, that is, when its own computer is not a master server computer, according to priority information. A first server that periodically executes a first search operation for searching for one server computer having no failure by sequentially communicating in a direction of higher rank starting from a server one rank higher than the own computer When a faulty server computer is found before a faultless server computer is found by the computer search means and the first server computer search means, and the computer is a master server computer, the own computer is newly added to the master server computer. Master setting means to be set as a server computer, and starting from the server one rank lower than the own computer according to the priority information Ranking find one unobstructed server computer by communicating in the forward direction becomes lower second
A second server computer search means for periodically executing the search operation of the above, and when the own computer is the master server computer and the data of the file held by the own computer is changed from the client computer, the changed data The second
The first copying means for copying to the server computer without any trouble found by the server computer searching means of the first, and when data is copied from another server computer, the data is found by the second server computer searching means. And second copying means for copying to a server computer having no trouble.

In such a configuration, when data of a file held in the master server computer is changed by the client computer, the data is transferred to all non-failed server computers (slave server computers) except the master server computer. The data is copied from the first server computer to the last server computer via each server computer in the order of priority. In other words, the data is copied from the master server computer to the server computer in the next priority order,
The sequence of priorities is copied in order by the daisy chain method (relay type) up to the last server computer, such as copying from the server computer to the server computer of the next priority sequence. For this reason, the master server computer is assigned to each of the other server computers (slave server computers) by one.
Although the speed is slower than when data is individually copied by the n-to-n communication method, the load on the server computer is small, and a high-availability computer system that is resistant to failure and more resistant to load can be realized.

Here, if there is provided a data transmission destination setting means for setting a server computer having no fault found by the second server computer search means as a data transmission destination, the first and second copying means are provided. Data copy by the daisy chain method can be efficiently performed according to the setting of the data destination setting means.

If a second server computer having no fault is found by the second server computer searching means, that is, if a server computer recovered from the fault is detected, the server server transmits data to the server computer. If the data is changed and set by the data transmission destination setting means and the data of all the files held by the own computer is copied to the server computer by the first or second copying means, the server computer (the restored server computer) Can be reliably and promptly used as one of the backup computers.

A high-availability computer system according to a third aspect of the present invention includes a plurality of first server computers connected via a first network and a second network which is slower than the first network. Multiple connected second
Server computer, and a third server computer connected between the first network and the second network,
If one of the computers serves as a master server computer and provides services to the client computers, and a failure occurs in the master server computer, the priority order to become the master for all the computers in the system and the master A high-availability computer system in which one of a plurality of remaining server computers takes over as a new master server computer and takes over the processing in accordance with the priority information in which the priority is cyclically used every time the first server computer is switched. When the own computer is the master server computer and the data of the file held by the own computer is changed from the client computer, the changed data is given priority over the own computer connected to the first network. Copy to the highest-ranked server computer with lower rank and no fault When data is copied from the first copy unit and another server computer, the data is copied to the most preferred server computer among lower-priority and fault-free server computers connected to the first network connected to the first network. A second copy unit for copying to a server computer having a higher rank, wherein the second server computer has its own computer as a master server computer, and data of a file held by its own computer has been changed from a client computer. In this case, there is provided a third copying means for individually copying the changed data to all the server computers connected to the second network without any trouble, wherein the third server computer has the first server. When data is copied from one of the server computers, the data is individually copied to all non-failed second server computers on the second network. And over, the second
A fourth copy unit that copies the data to the first server computer having the highest priority among the first server computers having no failure on the first network when data is copied from the first server computer. It is characterized by having.

In such a configuration, on the first network which is a high-speed network, data backup by a daisy-chain method which requires a small load on a server computer is applied, and the second network which is a low-speed network is used.
In this network, the data backup by the 1: n communication method, in which the time required for matching the data of the server computers is short, is applied, and a system can be flexibly adapted to the network configuration.

The present invention relating to the above apparatus (high-availability computer system) is also realized as an invention relating to a method (data backup method in a high-availability computer system).

[0021]

Embodiments of the present invention will be described below with reference to the drawings.

[First Embodiment] (Schematic Configuration) FIG. 1 is a block diagram showing a configuration of a high availability computer system according to a first embodiment of the present invention. The system shown in FIG. 1 includes three or more server computers, for example, four server computers (hereinafter, simply referred to as servers) S1 and S
2, S3, S4, and a plurality (here, m) of client computers (hereinafter, simply referred to as clients) C1 to C
m, each of these servers S1 to S4 and client C1
To Cm.

The servers S1 to S4 are divided into one master server for providing a service and a plurality of slave servers for backing up the master server. In the state of FIG. 1, the server S1 is a master server, and the other servers S2 to S4
Is a slave server (backup server).

The clients C1 to Cm use the service provided by the master server (S1) through the network N and write the files of the master server (S1).

The master server (S1) transfers the data in its own (own computer) file in the slave server (S2).
To the master server (S4).
1) and the contents of the files of the slave servers (S2 to S4) are made equal. As a result, as shown in FIG. 2, when a failure occurs in the master server (S1), one of the slave servers (S2 to S4), for example, the slave server S2 becomes a new master server, and the copied data is transferred. The client then takes over the service provided by the master server (S1) and provides the continuation of the service to the clients C1 to Cm. The above is the schematic configuration of the high availability computer system in the present embodiment.

(Internal Configuration of Server) Next, the internal configuration of the servers S1 to S4, which form the center of the system of FIG.
This will be described with reference to the block diagram of FIG.

First, the servers Si and Sj (i and j are 1 to 1)
4, where i ≠ j), three daemons (processing means operating in the background) of the state monitoring daemon 11, the data reception daemon 12, and the data transmission daemon 13 are configured to operate. Hereinafter, taking the server Si as an example, the status monitoring daemon 11, the data reception daemon 12,
The function of the data transmission daemon 13 will be described.

First, the status monitoring daemon 11 on the server Si
Communicates periodically with other state monitoring daemons 11 running on all servers other than the server Si on which the server itself resides, such as the state monitoring daemon 11 on the server Sj.
If there is a server that cannot communicate with this periodic communication, it can be determined that a failure has occurred in that server.

The status monitoring daemon 11 on the server Si
Which server in the system is in a failure state is stored as an internal state. Then, the status monitoring daemon 11 on the server Si sends the data receiving daemon 12 and the data transmission daemon 13 on its own server Si, based on the status of each server in the system, from which server to receive data, Indicate whether to send data to the server.

Data reception daemon 12 on server Si
Receives data transmitted from the data transmission daemon 13 of another server such as the server Sj and records the data in, for example, a disk storage device (not shown) which is a file storage unit of the server Si. The source server of the data to be received is specified by the status monitoring daemon 11 on the same server Si.

Data transmission daemon 13 on server Si
Monitors the data on the disk storage device of its own server Si, and transmits the changed data to the data reception daemon 12 of another server such as the server Sj. As described above, the server to which the data is to be transmitted (the destination server of the data) is specified by the status monitoring daemon 11 on the same server Si.

Next, the operation of the high-availability computer system including the servers S1 to S4 having the above-described configuration will be described by taking as an example a case where backup by the 1: n communication method is applied.

(Backup by 1: n communication method) First, backup by 1: n communication method will be described. Now, it is assumed that the server S1 is a master server and the servers S2 to S4 are slave servers, as shown in FIG. In this state, the clients C1 to C in FIG.
m writes data to a file held by the master server S1,
1 (by the data transmission daemon 13 of the own server) copies the data to the slave servers S2 to S4, that is, backs up the data. The master server S1 applies a one-to-n communication backup to the data backup from the master server S1 to the slave servers S2 to S4, and as shown by reference numerals 41 to 43 in FIG. , The same data is individually transmitted via the network N.

(Basic Operation of Master Server and Slave Server) Next, the basic operation of the master server and the slave server will be described. First, in the system of FIG. 1, the priorities of masters of all the servers S1 to S4 are determined in advance, and the information of the priorities is stored in the status monitoring daemon 11 of each of the servers S1 to S4. If the performance of each server is different, the priority may be set higher for a faster server. As shown in FIG. 5, for example, the priority order information indicates the order of circulation for each of the servers S1 to S4. In this example, the order is S1 → S2 → S3 → S4 → S1 → S
2 → ... In addition, (the identification information of) the server that is currently the master in the system also includes the servers S1 to S1.
4 is stored in the status monitoring daemon 11. At the time of the initial startup, the server of rank 1 (the server S1 in the example of FIG. 5) is stored as the master.

It should be noted that the priority information is S1, S2, S3,
S4 (only four), and the arrangement order itself is, for example, S1 → S2 → S3 → S4 to S2 → S3 → S4 → S1
May be dynamically circulated in the order of S3 → S4 → S1 → S2. In this case, the priority information itself holds the information of the server that is currently the master. Also, the priority information is set to S1, S2, S3, S4
(Four), the order of the masters may be fixed, and the pointer indicating the position of the master may be cyclically moved on the priority information.

The status monitoring daemon 11 of each of the servers S1 to S4 determines a partner server to receive and transmit data as follows. First, at the time of initial startup, the server of rank 1 determined by the priority information shown in FIG. 5, that is, the server S1, becomes the master server. On the other hand, (the status monitoring daemon 11 of) the servers S2 to S4 that are not the master servers
In the procedure shown in the flowchart of FIG. 5, the status monitoring daemon 1 of servers other than itself (starting from the server (server S1) having the highest priority at that time and moving in the direction of decreasing priority).
It communicates in order to 1) and searches for a server that returns a reply that it is the master (steps 61 to 6).
4). Here, (the status monitoring daemon 11 of) each of the servers S2 to S4 receives a reply from the server S1 that the server is the master.

The status monitoring daemon 11 of the servers S2 to S4
Finds the server (S1) that returns a reply that it is the master (step 62), stores the server (S1) as the master server, and notifies the data reception daemon 12 in its own server (step 65).

On the other hand, when the master server cannot be found even when all the servers are searched (steps 62 and 63), such as when the server S1 has a failure, the server S1 having no failure has been found. If the user has the highest priority (step 66), the server becomes the master server (step 67). Here, it is assumed that no failure has occurred in the server S1 having the highest priority at the time of the initial startup, and the server S1 becomes the master server as shown in FIG.

In this case, the status monitoring daemon 11 of the server (master server) S1 that has become the master performs the procedure shown in the flowchart of FIG. 7 by two steps from the server (S2) which is one rank lower than itself. It communicates with the server (S3) and the server three lower (S4) in order of decreasing order, and searches for all servers in which no failure has occurred (steps 71 to 75).

If a search is made up to a rank (priority 4 in FIG. 5) that is one rank ahead of itself (priority 5 in the example of FIG. 5) on the priority order information, no failure has occurred. If the server cannot be found (steps 74 and 76), the status monitoring daemon 11 of the master server S1 notifies the data transmission daemon 13 of its own server to stop the operation (step 77). On the other hand, if a server having no failure is found (step 72), the status monitoring daemon 11 of the master server S1 notifies its own data transmission daemon 13 of the server as a data transmission destination (step 7).
3).

By doing so, the master server S1 detects all servers in which no failure has occurred, and sends the above-mentioned server 1 to the server in which the failure has not occurred. Data can be sent (copied) individually using the n-to-n communication method.

Here, as shown in FIG. 8, of the servers other than the master server S1,
If it is assumed that a failure has occurred in the server S2 among the servers S2, S3, and S4 (failed servers are marked with a cross),
As shown by reference numerals 81 and 82 in FIG. 8, the data is individually transmitted (copied) only to the slave servers S3 and S4 from the master server S1 (the data transmission daemon 13 thereof). The operation of the status monitoring daemon 11 on each server including the master server described above is performed periodically.

(Operation When Failure Occurs) Next, the operation when a failure occurs in the server in such a state will be described.
First, it is assumed that a failure has occurred in the master server S1, as indicated by reference numeral 90 in FIG. The occurrence of the failure of the master server S1 is detected by another normal server, that is, the slave servers S3 and S4, according to the algorithm shown in the flowchart of FIG. In this case, among the servers S3 and S4 in which no failure has occurred, the server S3 with the highest priority at that time (rank 3) next to the master server S1.
That is, the server S3 having the higher priority becomes a new master server in response to the failure detection of the master server S1 as shown in FIG. Then, as is clear from the priority order information of FIG. 5, the priority order of the servers S1 to S4 is S3 (rank 3) → S4 (rank 4) → S1 (rank 5) → S2 (rank 6) in descending order. Becomes

The status monitoring daemon 11 of the server S3 that has newly become the master is denoted by reference numerals 91, 92 and 93 in FIG.
As shown by, the server one rank lower than itself (S4)
From the next lower server (S1) and the third lower server (S1)
2) and communicate sequentially in the direction of decreasing order,
Find all servers that have not failed. In the example of FIG. 9, only the server S4 is detected as a server in which no failure has occurred, and data is copied from the new master server S3 to the server (slave server) S4.

On the other hand, in the state of FIG. 8, if a failure occurs in any of the slave servers S3 and S4, ie, the backup servers S3 and S4, the status monitoring daemon 11 of the master server S1 transmits the failure to its own server. The transmission from the data transmission daemon 13 to the failed server is stopped.

(Operation at Restoration) Next, the operation when the server that has been stopped due to the failure is restored will be described. now,
It is assumed that the server S1 that has been stopped due to a failure as shown in FIG. 9 has been restored as shown by reference numeral 100 in FIG.

Then, the status monitoring daemon 11 of the master server S3 detects that the server S1 has recovered by a periodic monitoring operation. In this case, the status monitoring daemon 11 of the master server S3 additionally specifies the server S1 as a data transmission destination to the data transmission daemon 13 of the own server.

On the other hand, the status monitoring daemon 11 of the recovered server S1 designates the current master server S3 as a data receiving destination (data transmission source) to the data receiving daemon 12 of the own server. Then, all the data of the master server S3, which is the destination of the server S1, is copied to the server S1, as indicated by reference numeral 101 in FIG.
Matching with data of other servers in the system is achieved.

Thereafter, any one of the clients C1 to Cm in FIG. 1 is assigned to the master server S3.
When data is written to a file held by a file, as shown by reference numerals 102 and 103 in FIG.
The data is sequentially copied to servers (slave servers) S4 and S1.

[Second Embodiment] In the first embodiment described above, the backup by the 1: n communication method is applied. However, the present invention is not limited to this. For example, the backup by the daisy chain method is used. It is also possible to apply.

A high-availability computer system according to a second embodiment of the present invention to which backup by the daisy-chain method is applied will be described with reference to the drawings. 1 and 3 are referred to for convenience of the system configuration.

(Backup by Daisy Chain Method) First, backup by the daisy chain method will be described. Now, as shown in FIG.
Is a master server, and servers S2 to S4 are slave servers. In this state, when any of the clients C1 to Cm in FIG. 1 writes data to the file of the master server S1, the operation of copying the data to the slave servers S2 to S4, that is, performing a backup operation is performed. Will be In the present embodiment, a daisy-chain backup is applied to this backup as follows.

Here, first, the master server S1 (the data transmission daemon 13 thereof) sends the slave server S2 to the server S2 among the slave servers S2 to S4, for example, by the reference numeral 111 in FIG.
The data is copied as indicated by. Next, from the slave server S2 (the data transmission daemon 13 thereof) to another slave server S3, for example, reference numeral 112 in FIG.
The data is copied as shown by. Then, from the slave server S3 (the data transmission daemon 13 thereof) to the remaining slave servers S4, reference numeral 11 in FIG.
The data is copied as shown at 3.

As described above, in the backup system using the daisy chain, data is copied from the master server to one of the slave servers, and then the data is copied from the slave server to another slave server. The data copy is repeated in a relay manner starting from the master server, and the data is copied to all the slave servers.

(Basic Operation of Master Server and Slave Server) Next, the basic operation of the master server and the slave server will be described. First, the master priority order for all the servers S1 to S4 is determined in advance, and the information on the priority order is stored in the state monitoring daemon 11 of each server S1 to S4.
Is stored in the status monitoring daemon 11 of each of the servers S1 to S4 in the same manner as in the above embodiment.

The status monitoring daemon 11 of each of the servers S1 to S4 determines a partner server to receive and transmit data as follows. First, at the time of initial startup, the server of rank 1 determined by the priority information shown in FIG. 5, that is, the server S1, becomes the master server. On the other hand, the servers S2 to S4 that are not master servers (the status monitoring daemon 11 thereof)
In the procedure shown in the flowchart of FIG. 2, the server starts with the server one rank higher than the server itself, the server with two ranks higher, the server with three ranks higher, and the master server S1 in order of higher rank. Communication is performed to search for one server in which no failure has occurred (steps 121 to 124).

The status monitoring daemon 1 of the servers S2 to S4 (
If 1) finds one server in which no failure has occurred (step 122), it terminates the above communication and informs the server to its own data reception daemon 12 (step 125).

If the status monitoring daemon 11 of the servers S2 to S4 detects that a failure has occurred in the master server (S1) before finding a server in which no failure has occurred (step 122). , 123), and the server that has detected the failure immediately becomes the master server (step 126).

After the above communication, the master server (S1),
Both the slave server (S2 to S4) and the slave server (S2 to S4) start from the server which is one rank lower than itself and follow the procedure shown in the flowchart of FIG. The server communicates in order from the lowest to the lowest server at the current time in the direction of lowering the order of the server to the server of three lower ranks, and searches for one server in which no failure has occurred (steps 131 to 134). .
Here, on the priority order information of FIG. 5, the server one rank lower than the server S4 of rank 4, that is, the server of rank 5, is the server S1. However, since this server S1 is currently the master server of rank 1, rank 5 is the rank of the master server S1 itself after one round, and the server S4 of rank immediately before that (rank 4) is the current server. Indicates that the server is the lowest server. Therefore, the server S4 does not perform communication.

Each of the servers (S1 to S3) searches for the server (S4) immediately preceding the current master server (S1) on the priority information starting from the server one rank lower, that is, the current server. Even if a search is made to a position (rank 4) immediately before the master server itself after one round (rank 5), if no server with no failure is found (step 133,
15), and notifies the data transmission daemon 13 of the server to stop the operation (step 136).

On the other hand, if each of the servers (S1 to S3) can find one server in which no failure has occurred (step 132), it terminates the communication and stores the server in its own data. The transmission daemon 13 is notified (step 137).

By doing so, as shown in FIG. 14, if a failure has occurred in the server S2 (the failed server is marked with an X), the data transmission daemon of the master server S1 13 is set to S3, S1 is set to the data reception daemon 12 of the server S3, S4 is set to the data transmission daemon 13, and S3 is set to the data reception daemon 12 of the server S4.

Thereafter, when data is changed in the master server S1 in the state of FIG. 14, first, as indicated by reference numeral 141, the data transmission daemon 1 of the master server S1 (
3) (Data reception daemon 1) of slave server S3
The data is sent to 2) and copied. Then, the data is sent from (the data transmission daemon 13 of) the slave server S3 to (the data reception daemon 12) of the slave server S4 and copied as indicated by reference numeral 142 in FIG.

As described above, in the present embodiment, data is transmitted from the master server to all slave servers in which no failure has occurred in a daisy chain system (relay system) across the slave servers in the order of priority of becoming the master. Sent.

(Operation when Failure Occurs) Next, an operation when a failure occurs in the server in such a state will be described.
First, it is assumed that a failure has occurred in the master server S1, as indicated by reference numeral 150 in FIG. The failure of the master server S1 is determined according to the algorithm shown in the flowchart of FIG.
Among the servers S4, the server S3 having the highest priority (order 3) next to the master server S1 at that time, that is, the server S3 having the higher priority is first detected. In this case, the server S3 becomes a new master server as shown in FIG.
The process of the server S1 that has been the master is taken over. In the new master server S3 (the data reception daemon 12 thereof), the server S1 is set as the data reception destination (data transmission source), but the setting is released.
Here, the copy of the data is denoted by reference numeral 151 in FIG.
Is performed only from the server S3 to the server S4.

On the other hand, when a failure occurs in the slave server (backup server), the following occurs. First, among the servers having higher priorities than the failed slave server, the data transmission destination of the server with the lowest priority (A) is the server among the servers having lower priorities than the slave server having the failure. The server is changed to a server with a higher priority (B), and conversely, the data reception destination (data transmission source) of B is changed to A. If B is the master server, A's data transmission daemon 13 stops.

Therefore, in the state of FIG.
As indicated by reference numeral 160 in FIG.
If a failure occurs in the slave server S4, the data transmission destination of the master server S1 is changed from S3 to S4.
Is changed from S3 to S1. In this case, data is copied only from the master server S1 to the slave server S4, as indicated by reference numeral 161 in FIG.

(Operation at Restoration) Next, the operation when the server that has been stopped due to the failure is restored will be described. In this case, among the servers having higher priorities than the restored server, the data transmission destination of the server having the lowest priority is changed to the restored server. If the server with the highest priority among the servers with lower priorities than the restored server is not the master server, the data receiving destination (transmission source) of this server is changed to the restored server.

Therefore, if the server S1 that has been stopped due to a failure as shown in FIG. 15 has been restored (as a slave server) as indicated by reference numeral 170 in FIG. 17, the priority is higher than that of the restored server S1. The restored server S1 is newly set as the data transmission destination of the server S4 having the lowest priority among the servers having the higher priority. Here, since there is no normal server having a lower priority than the restored server S1, there is no server whose data reception destination (transmission source) is changed. Also, the restored server S
The first data receiving destination is set in the server S4. And
As shown by reference numeral 171 in FIG. 17, all data of the server S4, which is the destination of the server S1, is
And is matched with the data of the other servers in the system.

Thereafter, the procedure of data backup when data is changed in the master server S3 in the state of FIG. 17 is as follows. First, in FIG.
As shown at 72, the changed data is copied from the master server S3 to the slave server S4. Next, as shown by reference numeral 173 in FIG.
The data is copied from the slave server S4 to the restored slave server S1. In the above, the operation at the time of restoration of a server stopped due to a failure has been described as an example in the case where there is no normal server having a lower priority than the restored server.

Next, a specific example of the operation at the time of server restoration when a normal server having a lower priority than the restored server exists will be described.

Now, it is assumed that the server S2 that has been stopped due to a failure in the state shown in FIG. 14 has been restored (as a slave server) as indicated by reference numeral 180 in FIG.
In this case, since the server with the lowest priority among the servers with higher priorities than the restored server S2 is the master server S1, the data transmission destination of the server S1 is changed to the slave server S2 restored from the slave server S3. Be changed. Further, since the server S3 having the highest priority among the servers having lower priorities than the restored server S2 is not the master server, the server S2 whose data receiving destination (transmission source) of the server S3 has been restored from the master server S1. Is changed to The data receiving destination of the restored server S2 is set to the server S1, and the data transmitting destination is set to the server S3. Then, all data of the server S1 (here, the master server S1) that is the destination of the server S2 is
As indicated by reference numeral 181 in FIG. 8, the data is copied to the server S2 and matched with data of another server in the system.

Thereafter, the procedure of data backup when data is changed in the master server S1 in the state of FIG. 18 is as follows. First, in FIG.
As shown at 82, the changed data is copied from the master server S1 to the restored slave server S2.
Next, the data is copied from the slave server S2 to the slave server S3 as indicated by reference numeral 183 in FIG. Then, the data is copied from the slave server S3 to the slave server S4 as indicated by reference numeral 184 in FIG.

Third Embodiment Next, a high availability computer system according to a third embodiment of the present invention will be described with reference to FIG.
This will be described with reference to the block diagram of FIG. In the figure,
A plurality of servers S11 to S14 are connected to a LAN (local area network) 21 which is a high-speed network. In addition, WAN which is a low-speed network
(World Area Network) 22 has server S14
To S17 are connected. In the present embodiment, the server S14 is provided for interconnecting the LAN 21 and the WAN 22.
From the side to the WAN 22 side, from the WAN 22 side to the LAN 21 side
It is responsible for copying (backing up) data to the side.

The feature of the system shown in FIG.
The daisy chain method is applied to backup between the servers S11 to S14 connected by
A point-to-point communication method is applied to the backup between the servers S14 to S17 connected by the above-described method, and the server S14 forms an interface that enables a mixture of both methods. Here, at least the servers S14 to S14 on the LAN 21
In S13, it is assumed that a predetermined application program operates to perform a unique process even when the server itself is a slave server.

The advantages of the daisy chain method are as follows.
In comparison with the n-to-n communication method, the load on the master server can be reduced. Conversely, an advantage of the one-to-n communication method is that the information of each slave server (backup server) matches quickly. Therefore, high-distant LAN2
The servers (servers S11 to S13) connected to the server 1 and running the application are operated without interrupting the operation of the application by applying the daisy chain method to the backup.

However, WA as a low-speed network
N22 (servers S15 to S15)
In 17), if the daisy-chain method is used for backup, the time required for copying data becomes longer, the matching of data held by each server is delayed, and the degree of data mismatch between the servers increases.

Therefore, a dedicated server S 14 having a function of copying data from the WAN 22 to the LAN 21 is placed between the LAN 21 and the WAN 22 from the LAN 21 to the WAN 22, and the server (from the LAN 21 to the WAN 22) is provided. When it is necessary to copy the data to S15 to S17), the data is received by the server S14, and the data is copied from the server S14 to each server (S15 to S17) on the WAN 22 by the 1: n communication method. Also, the servers on the LAN 21 side from the WAN 22 side (S11 to S11)
If the data needs to be copied in S13), the data is received by the server S14, and the data is copied from the server S14 to the server with the highest priority (S15 to S17) on the LAN 21. Then, the data is sequentially copied from the server to another server on the LAN 21 in a daisy chain system according to the priority order.

Next, a specific example of data copy (data backup) in the system of FIG. 19 will be described. Here, the priority order is S11 → S12 → S13 → S14 → S
The order is 15 → S16 → S17, and it is assumed that the server S11 is the master server.

First, data is copied from the master server S11 to the server (slave server) S12 via the LAN 21, as indicated by reference numeral 191 in FIG. Next, as shown by reference numeral 192 in FIG. 19, the data is transmitted from the server S12 to the server S13 via the LAN2.
1 is copied. Next, the data is shown in FIG.
As indicated by reference numeral 193, the data is copied from the server S13 to the server S14 via the LAN 21. The server S14 transmits the data from the server S13 to the other servers S15 and S1 on the WAN 22 as shown by reference numerals 194, 195 and 196 in FIG.
6 and S17 are sequentially copied. If a server having a lower priority than the server S14 exists on the LAN 21, the server S14 (as a server on the LAN 21 to which the daisy-chain backup is applied) transmits data to the server having the highest priority. Copy

The same applies to the copying of data from the WAN 22 to the LAN 21 except that the data direction is reversed. Hereinafter, from the WAN 22 side to the LAN 21
As shown in FIG. 20, a specific example of data copy to the
The priority is S15 → S16 → S17 → S11 → S12 →
The order is S13 → S14, and a case where the server S15 is a master server will be described as an example.

First, the master server S15 on the WAN 22
To another server (slave server) on the WAN 22
In S16, S17, and S14, reference numeral 20 in FIG.
As shown by 1, 202 and 203, the same data is sequentially copied via the WAN 22 by the 1: n communication method.
Next, the data is copied from the server S14 to the server S11 having the highest priority (excluding the server S14) on the LAN 21 as indicated by reference numeral 204 in FIG. Next, the data is copied from the server S11 to the server S12 (having the next priority) via the LAN 21, as indicated by reference numeral 205 in FIG.
Next, as shown by reference numeral 206 in FIG. 20, the data is transmitted from the server S12 to the server S (of the next priority).
13 via the LAN 21.

After that, the above data may be copied from the server S13 to the server S14 (having the next priority). However, in this embodiment, since the server on the WAN 22 side is the master, the data is sent to the server S14. No data copy is performed. The reason is that when the server on the WAN 22 side is the master, the data is transmitted from the server S14 to the LAN2.
1 has been copied to the server on
14 already exists. The server S14
Is the master, the server S14 has the highest priority except for the server on the LAN 21 (here, the server S14).
11), and the data is sequentially copied to the other servers S15 to S17 on the WAN 22.

As described above, in this embodiment, the high-speed LAN
LAN and WA in a system in which servers connected by LAN and servers connected by low-speed WAN coexist
N and a dedicated server that controls data copying between N and N.
By applying the 1: n communication method on the AN, it is possible to construct a system flexibly corresponding to the network configuration.

[0085]

As described in detail above, according to the present invention, 3
By securing a plurality of backup server computers using more than one server computer and effectively copying data to the plurality of backup server computers, a highly available computer system that is more resistant to failures and a highly available computer that is more resistant to loads A system can be built.

According to the present invention, a plurality of server computers are linked by a high-speed network and a low-speed network, and a data backup method suitable for each network is used together, so that an effective data backup flexibly corresponding to a network configuration is achieved. And a more efficient high-availability computer system can be constructed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a high availability computer system according to a first embodiment of the present invention.

FIG. 2 is an exemplary view for explaining service takeover when a failure occurs in a master server in the embodiment.

FIG. 3 is an exemplary block diagram showing the internal configuration of each server in the embodiment.

FIG. 4 is an exemplary view for explaining data backup by the 1: n communication method applied in the embodiment;

FIG. 5 is an exemplary view for explaining a master priority order applied in the embodiment;

FIG. 6 is an exemplary flowchart for explaining the operation procedure of (the status monitoring daemon 11 of) the slave server in the embodiment.

FIG. 7 is an exemplary flowchart for explaining the operation procedure of (the status monitoring daemon 11 of) the master server in the embodiment.

FIG. 8 is a diagram for explaining a basic operation of the master server S1 in the system of FIG. 1 by taking a case where a failure occurs in the server S2 as an example.

9 is a diagram for explaining an operation when a failure occurs in the master server S1 in the state of FIG. 8 and the server S3 newly becomes a master server.

FIG. 10 is a diagram for explaining an operation when the server S1 is restored in the state of FIG. 9;

FIG. 11 is an exemplary view for explaining data backup by a daisy chain method applied in the high availability computer system according to the second embodiment of the present invention.

FIG. 12 is a flowchart for explaining an operation procedure of (the status monitoring daemon 11 of) the slave server in the second embodiment.

FIG. 13 is a flowchart for explaining an operation procedure of (the status monitoring daemon 11 of) the master server and the slave server in the second embodiment.

FIG. 14 is an exemplary view for explaining a basic operation of each server in the second embodiment, taking a case where a failure has occurred in a server S2 as an example.

FIG. 15 is a diagram for explaining an operation when a failure occurs in the master server S1 in the state of FIG. 14 and the server S3 becomes a new master server.

FIG. 16 is a view for explaining an operation when a failure occurs in the slave server S3 in the state of FIG. 14;

FIG. 17 is a view for explaining an operation when the server S1 is restored in the state of FIG. 15;

FIG. 18 is a view for explaining an operation when the server S2 is restored in the state of FIG. 14;

FIG. 19 is a diagram illustrating a data backup using a one-to-n communication method and a daisy-chain method used in the high availability computer system according to the third embodiment of the present invention;
FIG. 7 is a diagram for describing an example in which the upper server S11 is a master.

FIG. 20 is a view for explaining data backup when a server S15 on a WAN 22 is a master in the third embodiment.

[Explanation of symbols]

S1 to S4: Server (server computer) S11 to S13: Server (first server computer) S14: Server (third server computer) S15 to S17: Server (second server computer) N: Network 11: Status monitoring Daemon (master searching means, master setting means, server computer searching means, first server computer searching means, second server computer searching means, data transmission destination setting means) 12 data reception daemon 13 data transmission daemon (copying means) , A first copy unit, a second copy unit, a third copy unit, a fourth copy unit) 21 LAN (local area network, first network) 22 WAN (wide area network, second network) )

Claims (5)

[Claims] At least three server computers connected via a network, one of which serves as a master server computer and provides services to client computers, and a failure occurs in the master server computer In this case, one of the remaining server computers is newly added to the master server in accordance with the priority information indicating the priority order to be the master for all the computers in the system and the priority being used every time the master is switched. A high availability computer system that takes over processing as a computer, wherein each of the server computers periodically executes a master search operation for searching for a master server computer when its own computer is not a master server computer, The master search means cannot find the master server computer and a fault occurs Master setting means for newly setting the own computer as the master server computer when the own computer has the highest priority among the server computers which have not performed the operation, and a faulty server computer when the own computer is the master server computer Server computer search means for periodically executing a server computer search operation for searching for a server computer having no fault and a server computer in which the own computer is a master server computer and data of a file held in the own computer has been changed from a client computer A high-availability computer system, further comprising: a copy unit for individually copying the changed data to all server computers having no fault found by the server computer search unit. 2. When the own computer is a master server computer, the computer further includes a data transmission destination setting unit that sets a server computer without a fault found by the server computer search unit as a data transmission destination. 2. The high-availability computer system according to claim 1, wherein the copying of the data is sequentially executed for all the server computers of the data transmission destination set by the data transmission destination setting means. 3. At least three server computers connected via a network, one of which serves as a master server computer and provides services to client computers, and a failure occurs in the master server computer. In this case, one of the remaining server computers is newly added to the master server in accordance with the priority information indicating the priority order to be the master for all the computers in the system and the priority being used every time the master is switched. A high-availability computer system that takes over processing as a computer, wherein each server computer starts from a server one rank higher than its own computer according to the priority information when its own computer is not a master server computer. Searching for one server computer without trouble by communicating in order of increasing order A first server computer searching means for periodically executing the first searching operation, and a faulty server computer is found before a faultless server computer is found by the first server computer searching means, and the computer Is a master server computer, a master setting means for newly setting the own computer as a master server computer; and, in accordance with the priority information, starting from a server one rank lower than the own computer and sequentially decreasing in order. Search for one server computer without failure by performing communication
A second server computer search means for periodically executing the search operation of the above, and when the own computer is the master server computer and the data of the file held by the own computer is changed from the client computer, the changed data First copying means for copying the data to a server server having no fault found by the second server computer searching means; and when data is copied from another server computer, the data is copied to the second server computer searching means. A high-availability computer system comprising: a second copying unit for copying to a server computer having no fault found by the unit. 4. A plurality of first server computers connected via a first network, and a plurality of second server computers connected via a second network lower in speed than the first network. A third server computer connected between the first network and the second network, one of which serves as a master server computer to provide services to client computers, When a failure occurs in a computer, one of the remaining server computers is indicated in accordance with the priority information indicating the priority order to be the master for all the computers in the system and the priority being cyclically used every time the master is switched. Is a high-availability computer system that takes over the processing as a new master server computer, wherein the first server computer has its own When the computer is the master server computer and the data of the file held by the own computer is changed from the client computer, the changed data is given priority over the own computer connected to the first network. A first copy unit for copying to the highest-ranked server computer among the low-level and fault-free server computers, and when data is copied from another server computer, the data is connected to the first network. Second copying means for copying to a server computer having a lower priority than the own computer and having the highest priority among server computers having no failure, wherein the second server computer has a master server computer When the data of the file held in the own computer is changed from the client computer, the changed data A third copy unit for individually copying the data to all the server computers without failure connected to the second network, wherein the third server computer receives data from the first server computer. Is copied, the data is individually copied to all the non-failed second server computers on the second network,
When data is copied from the second server computer, the data is copied to the first server computer having the highest priority among the first server computers having no failure on the first network. A high availability computer system comprising a fourth copy unit. 5. At least three server computers connected via a network, one of which serves as a master server computer and provides services to client computers, and a failure occurs in the master server computer. In this case, one of the remaining server computers is newly added to the master server in accordance with the priority information indicating the priority order to be the master for all the computers in the system and the priority being used every time the master is switched. A data backup method in a high-availability computer system that takes over processing as a computer, wherein, if the own computer is not a master server computer, the order starts from the server computer with the highest priority at that time according to the priority information. Master server calculation by communicating in descending order If the master server operation is not found in the master search operation and the priority of the own computer is the highest among the server computers in which no failure has occurred, the master search operation is performed. Becomes a master server computer. When the own computer is the master server computer, communication is performed in order from the server computer one rank lower than the own computer to the lower rank in accordance with the priority information, thereby causing a failure. A server computer search operation for searching for a server computer having a fault and a server computer having no failure is periodically executed, and when data of a file held in the own computer is changed from a client computer, the changed data is transferred to the server. High availability characterized by individual copying to all fault-free server computers found in the computer search operation Data backup method in sex computer system.