JP4435142B2 - Cluster system and synchronization method - Google Patents

Description

  The present invention relates to a cluster system in which a standby computer takes over a service started by an active computer when a failure occurs in the active computer. The present invention relates to a cluster system and a synchronization method suitable for synchronizing data stored in the data storage means.

  2. Description of the Related Art Conventionally, a cluster system including a plurality of computers capable of providing the same service is known as a computer system that increases system availability. In this type of cluster system, when a failure occurs in a computer that currently provides a service, another normal computer starts the service. By taking over this service, the time during which the service becomes unavailable can be minimized.

  In order for another computer to take over the service from the computer in which the failure has occurred, it is necessary for each computer to share data necessary for continuing the service. As one of the data sharing methods, a distributed sharing method is known in which all computers (nodes) included in a cluster system individually hold the same data.

In a cluster system to which the distributed sharing method is applied, when data update occurs in a certain node, update information is transmitted from the node to other nodes in the cluster system via the network. With this update information, the same data update as the data update at a certain node is performed at other nodes in the cluster system. Thereby, matching (synchronization) of data held by all the nodes in the system is achieved. Such a mechanism for synchronizing the data held by all the nodes in the system is described in, for example, Patent Document 1.
JP 2000-276420 (paragraph 0026)

  In a cluster system, there is a node (standby node) that is waiting for a failure of a node (computer) that provides a service. When the standby node stops due to the occurrence of a failure, data updates that occur in other nodes during the stop period are not reflected on the standby node. For this reason, when the standby node recovers from the failure, it is necessary to perform a synchronization process for matching the data held by the standby node with the data of the other node by having the data transmitted from the other node. . Until this synchronization process is completed, even if a failure occurs in the node providing the service (active node), the standby node can use the data held by the standby node. Therefore, the service will not be taken over.

  The present invention has been made in consideration of the above circumstances, and its purpose is that the standby computer takes over the service from the active computer even before the synchronization processing performed at the time of recovery from the failure of the standby computer is completed. It is an object of the present invention to provide a cluster system and a synchronization processing method.

  According to one aspect of the present invention, one of the first and second computers functions as an active computer with respect to providing a predetermined service, and the rest of the first and second computers are standby computers. When a failure occurs in the active computer, a cluster system is provided in which the standby computer takes over the service started by the active computer. Each of the first and second computers of the cluster system includes data storage means for storing data, and information on the area of the data storage means in which data update has occurred while functioning as the active computer. Update reflection wait information storage means for storing as update reflection wait information representing an area not reflected in data update in the standby computer, and update request information for reflecting the generated data update in the standby computer Update request information transmitting means for transmitting to the standby computer, and information on an area in which data update generated in the active computer is not reflected in the standby computer in a state of functioning as the standby computer is an unupdated area Unupdated area information storage means for storing information, and transmission of the update request information of the active computer in a state of functioning as the standby computer The update request information transmitted from the stage is received, the data update generated in the active computer is reflected in the data storage means of the standby computer based on the update request information, and the data update is reflected The updated area information of the updated area is notified to the active computer, and the unupdated area information corresponding to the area in which the data update is reflected is stored in the unupdated area information storage unit of the standby computer When the update area information is notified from the update request information receiving means for deleting the unupdated area information to the active computer from the update request information receiving means of the standby computer with respect to the update request information, Update reflection waiting means for deleting update reflection waiting information corresponding to the area indicated by the update area information from the update reflection waiting information storage means of the active computer; and When a failure occurs when the computer is functioning as a computer, the standby computer temporarily stops operating, and then recovers from the failure, requesting the active computer to start synchronization processing The update reflection wait information stored in the update reflection wait information storage means of the active computer is acquired from the active computer, and the update reflection wait information is used as unupdated area information to determine whether the standby computer When the start of synchronization processing is requested from the built-in request unit stored in the update area information storage unit and the built-in request unit of the standby computer in a state of functioning as the active system computer, The update reflection waiting information stored in the update reflection waiting information storage means is transmitted to the incorporation request means, and the area represented by the update reflection waiting information Is read from the data storage unit of the active computer, and update request information necessary for synchronization processing for updating the data in the corresponding area of the data storage unit of the standby computer is updated with the data. Embedded request receiving means for transmitting to the update request information receiving means of the system computer.

  According to the present invention, at the start of the synchronization processing in the standby computer, the update reflection waiting area information (update reflection waiting information) acquired from the active computer is stored in the unupdated area information storage means as unupdated area information. Remembered. Therefore, the standby computer can check the unupdated area based on the non-updated area information stored in the non-updated area information storage unit even if a failure occurs in the active system before the synchronization processing is completed. Therefore, it is possible to quickly take over the service that has been started on the active computer.

Embodiments of the present invention will be described below with reference to the drawings.
[First Embodiment]
FIG. 1 is a block diagram showing a configuration of a cluster system according to the first embodiment of the present invention. This cluster system is composed of, for example, two computers (nodes) 10-1 and 10-2. Note that the cluster system may be composed of three or more computers (nodes). The nodes 10-1 and 10-2 are interconnected by a communication path 11 such as a network.

  In the state of FIG. 1, it is assumed that the node 10-1 functions as an active system and the node 10-2 functions as a standby system for providing a certain service. In this state, for another service, the node 10-2 may function as an active system, and the node 10-1 may function as a standby system. However, in the following description, for the sake of convenience, it is assumed that the node 10-1 is an active node and the node 10-2 is a standby node. In FIG. 1, a client terminal that receives service provision is omitted.

  The nodes 10-1 and 10-2 have cluster control units 100-1 and 100-2, respectively. The cluster control units 100-1 and 100-2 are connected via the communication path 11 so that one (one) of the nodes 10-1 and 10-2 is an active system and the other (remaining) is a standby system. Work with mutual agreement. The cluster control units 100-1 and 100-2 include failure detection units 101-1 and 101-2. The failure detection units 101-1 and 101-2 detect the failure of the nodes 10-2 and 10-1 by communicating with each other via the communication path 11, respectively.

  The nodes 10-1 and 10-2 include data storage units 102-1 and 102-2, update reflection waiting information storage units 103-1 and 103-2, and unupdated area information storage units 104-1 and 104, respectively. -2. In the present embodiment, the data storage units 102-1 and 102-2 are each secured on a disk storage device such as a hard disk drive. On the other hand, the update reflection waiting information storage units 103-1 and 103-2 and the non-updated area information storage units 104-1 and 104-2 are secured in rewritable nonvolatile memories, respectively. The update reflection waiting information storage units 103-1 and 103-2 and the unupdated area information storage units 104-1 and 104-2 may be secured on the disk storage device, respectively.

  The data storage units 102-1 and 102-2 store data that is updated / referenced by a request from the service 105 provided by the active system. The storage areas of the data storage units 102-1 and 102-2 are managed by being divided into a plurality of areas (blocks, partial areas) of a certain size. Each area is assigned an address which is a continuous number. FIG. 1 shows an example in which the storage areas of the data storage units 102-1 and 102-2 are divided into nine areas and managed for convenience of drawing. Assume that addresses 9 to 9 are assigned to these nine areas, as shown in FIG.

  In the example of FIG. 1 in which the node 10-1 is an active system, the service 105 is started on the node 10-1. The service 105 issues update request information 106 when data change (update) is necessary, and issues reference request information 107 when data reference is necessary.

  FIG. 2 shows an example of the data structure of the update request information 106 and the reference request information 107. Here, the update request information 106 includes update area information and update data. The update area information specifies an area range to be updated in the data storage unit 102-1. In the present embodiment, the update area information is composed of the address (start address) of the start area and the address (end address) of the end area of the area range to be updated. Note that the update area information may be composed of a start address and the number of areas. On the other hand, the reference request information 107 includes reference area information. The reference area information specifies an area range to be referred to in the data storage unit 102-1. In the present embodiment, the reference area information includes a start address and an end address of an area range to be referred to. The reference area information may be composed of a start address and the number of areas.

  The update reflection wait information storage units 103-1 and 103-2 store update reflection wait information (update reflection wait information queue) having a queue structure. The update reflection wait information queue is a queue of update area information in the update request information. Here, in a state where the node 10-1 or 10-2 functions as an active system, the area of the data storage unit 102-1 or 102-2 based on the update request information in the node 10-1 or 10-2. When a data update occurs for the update update wait information storage unit 103-1 and the update area information (update area information) as update reflection wait information indicating an area not yet reflected in the data update in the standby node. It is connected (added) to the update reflection waiting information queue in 103-2. The unupdated area information storage units 104-1 and 104-2 store information indicating unupdated areas (unupdated area information).

  The node 10-1 has a data request monitoring unit 108. The data request monitoring unit 108 monitors the update request information 106 and the reference request information 107 issued by the service 105 when the node 10-1 is active. When the data request monitoring unit 108 detects the update request information 106, the data request monitoring unit 108 updates the data in the corresponding area in the data storage unit 102-1 according to the update request information 106, and updates the update request information 106, which will be described later. The information is sent to the information transmission unit 109-1. Although omitted in FIG. 1, a data request monitoring unit corresponding to the data request monitoring unit 108 is provided in the node 10-2. However, the data request monitoring unit of the node 10-2 does not operate when the node 10-2 is in a standby state.

  The nodes 10-1 and 10-2 respectively include update request information transmitting units 109-1 and 109-2, update request information receiving units 110-1 and 110-2, and update reflection waiting units 111-1 and 111-. 2, incorporation request units 112-1 and 112-2, and incorporation request reception units 113-1 and 113-2.

  When receiving the update request information 106 from the data request monitoring unit 108, the update request information transmitting unit 109-1 of the node 10-1 stores the update area information included in the update request information 106 in the update reflection waiting information storage. Added to the section 103-1. The update request information transmitting unit 109-1 transmits the update request information 106 to an update request information receiving unit other than the node 10-1 (here, the update request information receiving unit 110-2 of the node 10-2).

  The update request information transmission unit 109-1 receives the update reflection wait information stored in the update reflection wait information storage unit 103-1 from the incorporation request reception unit 113-1 and includes the integration request unit 112-2 of the node 10-2. In this case, the data (updated data) in the area indicated by the update reflection waiting information is acquired from the data storage unit 102-1. The update request information transmitting unit 109-1 generates update request information including update reflection waiting information (update area information) and the acquired data, and the update request information is received by the update request information receiving unit of the node 10-2 To 110-2.

  The update request information transmitting unit 109-2 of the node 10-2 performs the same operation as the above-described update request information transmitting unit 109-1 when the node 10-2 is an active node.

  The update request information receiving unit 110-2 of the node 10-2 receives the update request information when the update request information is transmitted from the update request information transmitting unit 109-1 of the node 10-1, and the data storage unit The data stored in 102-1 is updated according to the update request information. The update request information receiving unit 110-2, when the non-updated area information identical to the update area information included in the received update request information is stored in the unupdated area information storage unit 104-2, Unupdated area information is deleted from the unupdated area information storage unit 104-2. After completing the data update according to the update request information, the update request information receiving unit 110-2 notifies the update reflection waiting unit 111-1 of the node 10-1 of the update area information included in the update request information. The update request information receiving unit 110-1 of the node 10-1 also receives the update request information from the update request information transmitting unit 109-2 of the node 10-2 when the node 10-1 is a standby node. In this case, the same operation as the update request information receiving unit 110-2 described above is performed.

  The integration request unit 112-2 of the node 10-2 causes a failure in the node 10-2 (that is, the standby node) while the node 10-2 is operating as a standby system, and the node 10-2 Stops, then recovers from the failure and starts again to start the synchronization process. When the node 10-2 is a standby node that has recovered from a failure, the integration request unit 112-2 requests the integration request reception unit 113-1 of the active node 10-1 to start the synchronization process. . When the update request wait information is received as a response to the synchronization processing start request from the request to receive synchronization 113-1 of the node 10-1, the update request wait information 112-2 is not updated. This is reflected in the area information storage unit 104-2. The integration request unit 112-1 of the node 10-1 also has a failure in the node 10-1 when the node 10-1 is a standby node, and the node 10-1 is stopped. When the synchronization processing is started after the operation is resumed and the operation is started again, the same operation as the above-described incorporation request unit 112-2 is performed.

  The integration request reception unit 113-1 of the node 10-1 stores in the update reflection waiting information storage unit 103-1 when the synchronization request is started from the integration request unit 112-2 of the node 10-2. The updated reflection waiting information is transmitted to the incorporation request unit 112-2. At this time, the update request information transmission unit 109-1 sequentially extracts the update reflection wait information stored in the update reflection wait information storage unit 103-1 from the head, and data in the area indicated by the update reflection wait information (updated data) Is acquired from the data storage unit 102-1. The update request information transmitting unit 109-1 generates update request information including update reflection waiting information (update region information) and the acquired data, and transmits the update request information to the update request information receiving unit 110-2. To do. Even if the service 105 of the node 10-1 is stopped, the update request information transmitting unit 109-1 operates as long as information is stored in the update reflection waiting information storage unit 103-1. The incorporation request receiving unit 113-2 of the node 10-2 is also requested to start the synchronization process from the integration request unit 112-1 of the node 10-1 when the node 10-2 is an active node. In the case, the same operation as the above-described built-in request receiving unit 113-1 is performed. The same applies to the update request information transmission unit 109-2 of the node 10-2.

Next, the operation of the cluster system of FIG. 1 will be described with reference to the flowchart of FIG.
First, regarding execution of the service 105, it is assumed that the node 10-1 functions as an active node and the node 10-2 functions as a standby node. In other words, it is assumed that the service 105 is started in the node 10-1.

  The data request monitoring unit 108 of the node 10-1 monitors request information issued by the service 105 so that the service 105 updates or references the data stored in the data storage unit 102-1 (step S1). . When the data request monitoring unit 108 detects that the request information has been issued by the service 105, the data request monitoring unit 108 determines the type of the request information (step S2). That is, the data request monitoring unit 108 determines whether the detected request information is update request information 106 or reference request information 107.

  If it is the update request information 106, the data request monitoring unit 108 updates the data in the corresponding area in the data storage unit 102-1 according to the update request information 106 (step S3). Next, the data request monitoring unit 108 sends the update request information 106 used for data update to the update request information transmitting unit 109-1 (step S4). In this case, the data request monitoring unit 108 returns to step S1 and monitors request information issued by the service 105.

  When the update request information transmission unit 109-1 receives the update request information 106 sent from the data request monitoring unit 108 (step S11), the update request information transmission unit 109-1 stores the update area in the update request information 106 in the update reflection waiting information storage unit 103-1. Information is added as update reflection waiting information (step S12). Here, the update area information in the update request information 106 is linked to the tail of the update reflection waiting information queue stored in the update reflection waiting information storage unit 103-1. The update request information transmitting unit 109-1 transmits the received update request information 106 to the update request information receiving unit other than the node having the update request information transmitting unit 109-1 (that is, another node) via the communication path 11. (Step S13). Here, the update request information 106 is transmitted to the update request information receiving unit 110-2 of the node 10-2. When executing step S13, the update request information transmitting unit 109-1 waits for new update request information 106 from the data request monitoring unit 108.

  When receiving the update request information 106 from the update request information transmitting unit 109-1 (step S21), the update request information receiving unit 110-2 receives the data in the corresponding area in the data storage unit 102-2 according to the update request information 106. Is updated (step S22). Next, the update request information receiving unit 110-2 receives unupdated area information indicating the same area (address) as the update area information included in the update request information 106 used for data update. It is determined whether it is stored in 104-2 (step S23). That is, the update request information receiving unit 110-2 uses the unupdated area information stored in the non-updated area information storage unit 104-2 to store the data in the data storage unit 102-1 that has been updated in step S22. It is determined whether it is indicated as an update area.

  If the data updated area is indicated by the unupdated area information that the data has not been updated, the update request information receiving unit 110-2 stores the unupdated area information in the unupdated area information storage unit 104-. Delete from 2 (step S24). Then, the update request information receiving unit 110-2 notifies the update completion by returning the update area information to the update reflection waiting unit 111-1 of the node 10-1 via the communication path 11 (step S25). On the other hand, when the non-updated area information regarding the data-updated area is not stored in the non-updated area information storage unit 104-2 (step S23), the update request information receiving unit 110-2 skips step S24. Then, the update reflection waiting unit 111-1 is notified of the update completion (step S25). In this case, the update request information receiving unit 110-2 returns to step S21 and waits for new update request information 106 to be transmitted from the update request information transmitting unit 109-1 of the node 10-1.

  When the update reflection waiting unit 111-1 receives an update completion notification from the update request information receiving unit 110-2 of the node 10-2 (step S31), the data updated by the update request information receiving unit 110-2 The update area information designating the same area as is deleted from the update reflection waiting information storage unit 103-1 (step S32). Then, the update reflection waiting unit 111-1 returns to step S31 and waits for a new update completion notification from the update request information receiving unit 110-2.

  On the other hand, when the request information detected in step S1 is the reference request information 107 (step S2), the data request monitoring unit 108 indicates an unupdated area indicating an area having the same address as the reference area specified by the reference request information 107. It is determined whether the information is stored in the unupdated area information storage unit 104-1 (step S5). If the unupdated area information indicating the same area as the reference area is stored in the unupdated area information storage unit 104-1, the data request monitoring unit 108 determines that the reference area is an unupdated area. In this case, the data request monitoring unit 108 waits for the non-updated region information indicating the same region as the reference region to be deleted from the non-updated region information storage unit 104-1 (step S6). Eventually, it is assumed that the unupdated area information indicating the area having the same address as the reference area is deleted from the unupdated area information storage unit 104-1. Then, the data request monitoring unit 108 reads the data from the reference area, assuming that the reference area of the data storage unit 102-1 designated by the reference request information 107 is no longer an unupdated area, and sends the data to the service 105. Return (step S7). Then, the data request monitoring unit 108 returns to step S1 and monitors request information issued by the service 105.

  In this state, a failure occurs in the node 10-2 operating as a standby node for providing the service 105, the node 10-2 stops, and then the node 10-2 recovers from the failure. It is assumed that the operation has started again. In this case, synchronization processing is performed in the cluster system of FIG. This synchronization processing will be described with reference to the operation explanatory diagram of FIG.

  First, the incorporation request unit 112-2 of the node 10-2 starts operation, and requests the incorporation request reception unit 113-1 of the node 10-1 to start the synchronization process (step S41). At this time, the data A is stored in all nine areas (the nine areas having addresses 1 to 9) to which the addresses 1 to 9 are assigned in the data storage unit 102-2 of the node 10-2. To do.

  On the other hand, among the nine area areas with addresses 1 to 9 in the data storage unit 102-1 of the node 10-1, the data in the areas with addresses 2, 3, 4, 5, and 8 is stopped by the node 10-2. It is assumed that they are updated to B, B, B, C, and D, respectively, during the period that was being performed. In other words, the update of the data in the areas of addresses 2, 3, 4, 5, and 8 in the data storage unit 102-1 is not reflected in the same address area (the same area) of the data storage unit 102-2. To do. In this case, update reflection waiting information indicating addresses 2, 3, 4, 5 and 8 is stored in the update reflection waiting information storage unit 103-1 of the node 10-1.

  When the integration request reception unit 113-1 of the node 10-1 is requested to start the synchronization process from the integration request unit 112-2 of the node 10-2 (step S41), the update reflection waiting information storage unit The update reflection waiting information stored in 103-1 is transmitted to the integration request unit 112-2 as a response to the synchronization processing start request (step S42). This update reflection waiting information is in response to the update request indicated by the update request information transmitted from the update request information transmitting unit 109-1 of the node 10-1 to the update request information receiving unit 110-2 of the node 10-2. The update request information receiving unit 110-2 indicates that update completion has not been notified. In other words, the update reflection waiting information is information on an area where the data update is not completed in the node 10-2, and information on an area where the data update in the node 10-1 needs to be reflected in the node 10-2. Show.

  When receiving the update reflection waiting information from the integration request receiving unit 113-1, the incorporation request unit 112-2 reflects the update reflection waiting information in the unupdated area information storage unit 104-2 (step S43). That is, the incorporation request unit 112-2 adds the update reflection waiting information received from the incorporation request reception unit 113-1 to the unupdated area information storage unit 104-2 as unupdated area information. Here, the update reflection waiting information indicating the addresses 2, 3, 4, 5, and 8 is stored in the unupdated area information storage unit 104-2 as unupdated area information.

  In this state, it is assumed that a failure occurs in the node 10-1 and the node 10-2 needs to take over the service 105 started in the node 10-1. In this case, conventionally, the takeover of the service 105 by the node 10-2 is waited until the synchronization processing is completed. The reason is that there is no means for confirming which area data is not updated until the synchronization processing is completed.

On the other hand, in this embodiment, the node 10-2 (embedding request unit 112-2) acquires the update reflection waiting information from the node 10-1 at the time of recovery from the failure, and the update reflection waiting information is not yet received. The updated area information storage unit 104-2 reflects it as unupdated area information. Therefore, as is apparent from the flowchart of FIG. 4, the node 10-2 has an unupdated area (in this case, addresses 2, 3 and 3) indicated by the unupdated area information stored in the unupdated area information storage unit 104-2. Although the process of referring to the areas (4, 5 and 8) cannot be performed, the process of referring to other areas and the process of updating the data in all areas can be performed. Therefore, the node 10-2 can take over the service 105. As shown in the flowchart of FIG. 3, the process of referring to the unupdated area is performed until the unupdated area information indicating the area is deleted from the unupdated area information storage unit 104-2, that is, the area is an unupdated area. It waits until it stops (step S6).
Now, the update request information transmitting unit 109-1 of the node 10-1 receives the update reflection waiting information stored in the update reflection waiting information storage unit 103-1 by the built-in request receiving unit 113-1 of the node 10-2. It also operates when it is transmitted to the incorporation request unit 112-2. In this case, the update request information transmission unit 109-1 sequentially extracts the update reflection wait information stored in the update reflection wait information storage unit 103-1 from the top, and the data in the area indicated by the update reflection wait information (updated) Data) is acquired from the data storage unit 102-1. The update request information transmitting unit 109-1 generates update request information including update reflection waiting information (update region information) and the acquired data, and transmits the update request information to the update request information receiving unit 110-2. To do. Even if the service 105 of the node 10-1 is stopped, the update request information transmitting unit 109-1 operates as long as information is stored in the update reflection waiting information storage unit 103-1.

  In the example of FIG. 4, first, update request information including area information (update area information) of the area of address 2 and data B of the area of address 2 is updated request information transmission section 109-1 of node 10-1. To the update request information receiving unit 110-2 of the node 10-2 (step S44). In this case, in the node 10-2, the data A in the area of the address 2 in the data storage unit 102-2 is updated to the data B (step S45). Then, the unupdated area information indicating the area of address 2 is deleted from the unupdated area information storage unit 104-2 of the node 10-2 (step S46). At this stage, the node 10-2 cannot perform the process of referring to the areas of the addresses 3, 4, 5, and 8, but can perform the process of referring to the other areas and the process of updating the data of all the areas. it can. The area that cannot be referred to is determined from the contents of the unupdated area information storage unit 104-2.

  Next, the update request information including the area information of the area of address 3 and the data B of the area of address 3 is received from the update request information transmitting unit 109-1 of the node 10-1 and the update request information of the node 10-2. Is transmitted to the unit 110-2 (step S47). In this case, in the node 10-2, the data A in the area of the address 3 in the data storage unit 102-2 is updated to the data B (step S48). Then, the unupdated area information indicating the area of address 3 is deleted from the unupdated area information storage unit 104-2 of the node 10-2 (step S49). At this stage, the node 10-2 cannot perform the process of referring to the areas of the addresses 4, 5, and 8, but can perform the process of referring to the other areas and the process of updating the data of all the areas.

  Next, the update request information including the area information of the area of address 4 and the data B of the area of address 3 is received from the update request information transmitting unit 109-1 of the node 10-1 Is transmitted to the unit 110-2 (step S50). In this case, in the node 10-2, the data A in the area of the address 4 in the data storage unit 102-2 is updated to the data B (step S51). Then, the unupdated area information indicating the area of address 4 is deleted from the unupdated area information storage unit 104-2 of the node 10-2 (step S52). At this stage, the node 10-2 cannot perform the process of referring to the areas of the addresses 5 and 8, but can perform the process of referring to the other areas and the process of updating the data of all the areas.

  Next, the update request information including the area information of the area of the address 5 and the data C of the area of the address 5 is received from the update request information transmitting unit 109-1 of the node 10-1 Is transmitted to the unit 110-2 (step S53). In this case, in the node 10-2, the data A in the area of the address 5 in the data storage unit 102-2 is updated to the data C (step S54). Then, the unupdated area information indicating the area of address 5 is deleted from the unupdated area information storage unit 104-2 of the node 10-2 (step S55). At this stage, the node 10-2 cannot perform the process of referring to the area of the address 8, but can perform the process of referring to the other area and the process of updating the data of all areas.

  Next, the update request information including the area information of the area of address 8 and the data D of the area of address 8 is received from the update request information transmitting unit 109-1 of the node 10-1 and the update request information of the node 10-2. Is transmitted to the unit 110-2 (step S56). In this case, in the node 10-2, the data A in the area of the address 8 in the data storage unit 102-2 is updated to the data D (step S57). Then, the unupdated area information indicating the area of the address 8 is deleted from the unupdated area information storage unit 104-2 of the node 10-2 (Step S58). At this stage, the node 10-2 can perform a process of referring to and a process of updating data for all areas.

  As described above, in the present embodiment, the update reflection waiting area is transferred from the node 10-1 to the node 10-2 in response to the synchronization process start request given from the node 10-2 to the node 10-1 at the start of the synchronization process. Information (waiting for update reflection) is transmitted as unupdated area information, and the reference request for that area is blocked at the node 10-2, so that the synchronization processing at the node 10-2 has not been completed. Even if a failure occurs in the node 10-1, the node 10-2 can quickly take over the service 105. As described above, in the cluster system according to the present embodiment, the service can be taken over earlier than in the conventional technique, so that the availability is increased.

[Second Embodiment]
FIG. 5 is a block diagram showing the configuration of the cluster system according to the second embodiment of the present invention. In FIG. 5, the same elements as those in FIG. 5 is different from the nodes 10-1 and 10-2 in FIG. 1 in that the configuration of the nodes 10-1 and 10-2 includes data request units 201-1 and 201-2, respectively. The nodes 20-1 and 20-2 to which the data request receiving units 202-1 and 202-2 are added are used.

Next, the operation of the cluster system of FIG. 5 will be described with reference to the flowchart of FIG. 6, focusing on the differences from the cluster system of FIG.
First, it is assumed that the node 20-1 takes over the service 105 started by the node 20-2 and becomes an active system. Here, it is assumed that the node 20-1 has taken over the service 105 in a state where the synchronization process executed when the node 20-1 recovers from the failure is not completed. Further, when the synchronization processing of the node 20-1 is started, the node 20-1 sends a synchronization processing start request to the node 20-2, so that the update reflection waiting information is sent from the node 20-2 to the node 20-1. Is transmitted as unupdated area information, and a reference request for the area is blocked by the node 20-1. Thereby, even if a failure occurs in the node 20-2 in a state where the synchronization processing in the node 20-1 is not completed, the node 20-1 can take over the service 105 promptly. This is the same as the first embodiment except that the relationship between the node 20-1 and the node 20-2 (the relationship between the active system and the standby system) is the reverse of the first embodiment. It is the same.

  In the present embodiment in which the synchronization process of the node 20-1 is not completed and the node 20-1 takes over the service 105, the node 20-1 is not updated, contrary to the first embodiment. The area information storage unit 104-1 stores unupdated area information, and the update reflection wait information storage unit 103-2 of the node 20-2 stores update reflection wait information. In the node 20-1 where the service 105 has been started in this state (that is, the node 20-1 that has newly become active), as indicated by the arrow 51 in FIG. Shall be issued.

  When the data request monitoring unit 108 detects the reference request information 107 issued by the service 105 (steps S41 and S42), the data request monitoring unit 108 refers to the unupdated area information storage unit 104-1. If unupdated area information indicating an area having the same address as the reference area specified by the reference area information in the reference request information 107 is stored in the unupdated area information storage unit 104-1 (step S45), The data request monitoring unit 108 determines that the reference area is an unupdated area (that is, data stored in the reference area is unupdated data). The operations so far are the same as steps S1 to S5 of the first embodiment.

  If the data request monitoring unit 108 determines that the reference region is an unupdated region, the data request monitoring unit 108 passes the reference region information designating the reference region to the data requesting unit 201-1 as indicated by the arrow 52 in FIG. S46). Then, the data request monitoring unit 108 waits for the non-updated area information indicating the same area as the reference area to be deleted from the non-updated area information storage unit 104-1 (step S47).

  On the other hand, the data request unit 201-1 to which the reference area information is passed from the data request monitoring unit 108 indicates the reference area information as indicated by an arrow 53 in FIG. And request data (update data) of the area specified by the reference area information (step S50).

  When the data request receiving unit 202-2 receives the reference area information from the data requesting unit 201-1, the data request receiving unit 202-2 is indicated by an arrow 54 in FIG. 5 from the reference area of the data storage unit 102-2 specified by the reference area information. The data is read out (step S61). This data is update data corresponding to the reference area data (unupdated data) in the data storage unit 102-1 specified by the reference area information in the reference request information 107.

  The data request receiving unit 202-2 generates update request information including data read from the reference region of the data storage unit 102-2 designated by the reference region information and update region information that matches the reference region information. Then, the update request information is transmitted to the update request information receiving unit 110-1 of the node 20-1 as indicated by the arrow 55 in FIG. 5 (step S62). This data is data to be reflected in the data storage unit 102-1 of the node 20-1 in the synchronization process in the first embodiment.

  When the update request information receiving unit 110-1 receives the update request information from the data request receiving unit 202-2, the update request information receiving unit 110-1 stores the update request information in the data storage unit 102-1 as indicated by an arrow 56 in FIG. The data in the corresponding area is updated (step S71). As a result, the data (not yet stored) in the data storage unit 102-1 designated by the reference region information in the reference request information 107 from the service 105 (that is, the update region information in the update request information from the data request receiving unit 202-2). (Update data) is matched with data (update data) in the data storage unit 102-2 designated by the area information. Therefore, the update request information receiving unit 110-1 uses the non-updated region information storage unit 104 as shown by an arrow 57 in FIG. 5 to indicate the non-updated region information specifying the same region as the region where the data has been updated in step S71. -1 is deleted (step S72).

  The data request monitoring unit 108 receives from the unupdated region information storage unit 104-1 unupdated region information that specifies the same region as the region specified by the reference request information in the reference request information 107 during the waiting of step S47. When deleted, it is determined that the data in the area is matched between the node 20-1 (the data storage unit 102-1) and the node 20-2 (the data storage unit 102-2). In this case, the data request monitoring unit 108 reads the data from the area of the data storage unit 102-1 designated by the reference request information in the reference request information 107, and stores the data as indicated by an arrow 58 in FIG. It returns to the service 105 (step S48).

  On the other hand, the update request information receiving unit 110-1 indicates the update area information in the update request information as indicated by an arrow 59 in FIG. 5 as a notification of the update completion in accordance with the update request information from the data request receiving unit 202-2. To the update reflection waiting section 111-2 of the node 20-2 (step S73). When the update reflection waiting section 111-2 receives the update area information as an update completion notification from the update request information receiving section 110-1, the update reflection waiting information for designating the same area as the update area information is indicated by an arrow in FIG. As indicated by 60, it is deleted from the update reflection waiting information storage unit 103-2 (step S80).

  As described above, in this embodiment, when a reference request for an unupdated area occurs in the node 20-1 that has taken over the service 105 in a state where the synchronization processing is not completed, the data request unit 201 of the node 20-1 -1 requests the data request receiving unit 202-2 of the node 20-2 to update the data in the unupdated area. Then, the update request information receiving unit 110-1 of the node 20-1 receives the data (update data) of the corresponding area of the data storage unit 102-2 of the node 20-2 from the data request receiving unit 202-2, and Data in the unupdated area is updated with data. That is, in this embodiment, when a reference request for an unupdated area occurs, the data in the unupdated area is updated using the reference request as a trigger. As a result, it is possible to shorten the time for which the data request monitoring unit 108 waits until it can read the data in the area specified by the reference request after detecting the reference request.

[Third Embodiment]
FIG. 7 is a block diagram showing the configuration of the cluster system according to the third embodiment of the present invention. In FIG. 7, the same elements as those in FIG. The feature of the cluster system shown in FIG. 7 is that three nodes 30-1, 30-2 and 30-3 are used instead of the nodes 20-1 and 20-2 of FIG. The basic configuration of the nodes 30-1, 30-2 and 30-3 is the same as that of the nodes 20-1 and 20-2 in FIG. Therefore, the same operation as in the first and second embodiments is also performed in the cluster system of FIG.

  In the example of FIG. 7, the service 105 is started at the node 30-1. That is, in the example of FIG. 7, for the service 105, the node 30-1 functions as an active system, and the nodes 30-2 and 30-3 function as standby systems.

  The node 30-1 is different from the node 20-1 in that, in addition to the update reflection wait information storage unit 103-1, the node 30-1 has an update reflection wait information storage unit 203 similar to the update reflection wait information storage unit 103-1. It is in. The update reflection waiting information storage units 103-1 and 203 are secured in a rewritable nonvolatile memory. The update reflection waiting information storage units 103-1 and 203 are information on areas in which data updates in the data storage unit 102-1 of the node 30-1 are not reflected in the data storage units of the nodes 30-2 and 30-3, respectively. (Update area information) is stored as update reflection wait information. Also in the nodes 30-2 and 30-3, two update reflection waiting information storage units corresponding to the update reflection waiting information storage units 103-1 and 203, respectively, are secured in a rewritable nonvolatile memory. However, in FIG. 7, the update reflection waiting information storage unit other than the update reflection waiting information storage unit 103-2 in the node 30-2 is omitted. In the update reflection waiting information storage unit 103-1, the data update in the data storage unit 102-1 of the node 30-1 is not reflected in the data storage units of the nodes 30-2 and 30-3, respectively. An area for storing information as update reflection waiting information may be secured. In this case, the update reflection waiting information storage unit 203 is not necessary.

  In the cluster system of FIG. 7, when a failure occurs in the node 30-1, the failure detection unit 101-1 included in the cluster control unit 100-1 of the node 30-1 and the cluster control unit 100 of the node 30-2. -1 includes a failure detection unit 101-2 or a failure detection unit (not shown) included in the cluster control unit of the node 30-2, and the failure of the node 30-1 is detected. . In this case, the service 105 started at the node 30-1 is taken over by the node 30-2 or 30-3.

  In this embodiment, the cluster control unit 100-2 of the node 30-2 and the cluster control unit of the node 30-3 are the update reflection waiting information storage unit 103-1 and the update reflection of the node 30-1 where the service 105 has been started. By comparing the update reflection waiting information stored in the waiting information storage unit 203, it is determined which of the nodes 30-2 and 30-3 takes over the service 105. Specifically, the cluster control unit 100-2 of the node 30-2 and the cluster control unit of the node 30-3 are stored in the update reflection waiting information storage unit 103-1 and the update reflection waiting information storage unit 203, respectively. An agreement is made so that the node having the smaller total area of the update reflection wait indicated by the update reflection wait information takes over the service 105.

  As a result, even if the node that takes over the service 105 has not completed the synchronization process, the total amount of area waiting for update reflection is small, and therefore, until the synchronization process is completed and all areas can be referenced. Time is short. In addition, since the node with the smaller total area waiting for update reflection takes over the service 105, the probability of a reference request for an unupdated area can be reduced. Further, even if a reference request for an unupdated area occurs, the data in the area specified by the reference request can be read by making an update request for the data in the unupdated area, as in the second embodiment. You can also reduce the time to wait until

[Modification of Third Embodiment]
Next, a modification of the third embodiment of the present invention will be described. Since the configuration of the cluster system applied in this modification is the same as that of the third embodiment, FIG. 7 is used. The feature of this modification is that assuming that the synchronization process is started in the node 30-2 while the service 105 is started in the node 30-1, the built-in request receiving unit 113 of the node 30-1 is assumed. The update request information is not transmitted by the update request information transmitting unit 109-1 of the node 30-1 when the update reflection waiting information is transmitted to the node 30-2.

  Hereinafter, the synchronization processing of this modification will be described. First, it is assumed that the synchronization process is started in the node 30-2 while the service 105 is started in the node 30-1. In this case, the integration request unit 112-2 of the node 30-2 requests the integration request reception unit 113-1 of the node 30-1 to start the synchronization process. Then, the incorporation request receiving unit 113-1 transmits the update reflection waiting information stored in the update reflection waiting information storage unit 103-1 to the incorporation request unit 112-2. When receiving the update reflection waiting information from the integration request receiving unit 113-1, the incorporation request unit 112-2 adds the update reflection waiting information as unupdated region information to the unupdated region information storage unit 104-2. The operations so far are the same as those in the case of the synchronization processing in the first embodiment (steps S41 to S43).

  When the update request waiting information 112-2 is added to the unupdated area information storage unit 104-2 as update update area information, the incorporation request unit 112-2 stores the unupdated area stored in the unupdated area information storage unit 104-2. Data in the area indicated by the information is acquired from the data request receiving unit (not shown) of the node 30-3 by the data requesting unit 201-2. That is, the data request unit 201-2 transmits the non-updated area information stored in the non-updated area information storage unit 104-2 as reference area information to the data request receiving unit of the node 30-3, so that the reference area Requests data in the area specified by the information (unupdated area information).

  Then, in the second embodiment, the data request receiving unit 202-2 of the node 30-3 receives the reference area information from the data requesting unit 201-1 of the node 10-1 by the data request receiving unit 202-2 of the node 10-2. The same operation as when received is performed. That is, the data request receiving unit of the node 30-3 reads the data from the area of the data storage unit designated by the reference area information (unupdated area information) from the data requesting unit 201-1, and the data and the reference area information. Is generated, and the update request information is transmitted to the update request information receiving unit 110-2 of the node 30-2.

  When receiving the update request information from the data request receiving unit of the node 30-3, the update request information receiving unit 110-2 updates the data in the corresponding area in the data storage unit 102-2 in accordance with the update request information. As described above, in the present modification, the incorporation request acceptance is received in response to the synchronization processing start request from the incorporation request unit 112-2 of the node 30-2 to the incorporation request acceptance unit 113-1 of the node 30-1. The data of the area specified by the update reflection waiting information returned from the unit 113-1 is acquired from the data request receiving unit of the node 30-3 functioning as the standby system, and the data storage unit 102-2 is used as the data. The unupdated data in the corresponding area can be updated. Here, it should be noted that update request information is not generated and transmitted by the update request information transmitting unit 109-1 of the node 30-1. That is, in this modification, the synchronization process can be performed without imposing a burden on the node 30-1 where the service 105 is started.

  Now, the update request information receiving unit 110-2 of the node 30-2 receives the data (unupdated data) in the corresponding area in the data storage unit 102-2 according to the update request information transmitted from the data request receiving unit of the node 30-3. ) Is updated, the same processing as steps S23 to S25 in the first embodiment is performed, and the update completion information is notified by returning the update area information to the update reflection waiting unit 111-1 of the node 30-1. As a result, the update reflection waiting information specifying the same area as the update area information is deleted from the update reflection waiting information storage unit 103-1 by the update reflection waiting unit 111-1.

  Note that the present invention is not limited to the above-described embodiments as they are, and can be embodied by modifying constituent elements without departing from the scope of the invention in the implementation stage. Various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the above embodiments. For example, some components may be deleted from all the components shown in each embodiment.

1 is a block diagram showing a configuration of a cluster system according to a first embodiment of the present invention. The figure which shows the data structure example of the update request information and reference request information which are shown by FIG. 6 is a flowchart showing a processing procedure of each unit of the cluster system in the first embodiment. The figure for demonstrating the operation | movement of the synchronization process in the said 1st Embodiment. The block diagram which shows the structure of the cluster system which concerns on the 2nd Embodiment of this invention. 6 is a flowchart showing a processing procedure of each unit of the cluster system in the first embodiment. The block diagram which shows the structure of the cluster system which concerns on the 3rd Embodiment of this invention.

Explanation of symbols

  10-1, 10-2, 20-1, 20-2, 30-1, 30-2, 30-3 ... nodes (computers), 100-1, 100-2 ... cluster control units, 102-1, 102 -2 ... Data storage unit, 103-1, 103-2, 203 ... Update reflection waiting information storage unit, 104-1, 104-2 ... Unupdated area information storage unit, 105 ... Service, 106 ... Update request information, 107 Reference request information 108 Data request monitoring unit 109-1, 109-2 Update request information transmission unit 110-1, 110-2 Update request information reception unit 111-1, 111-2 Update reflection Waiting unit, 112-1, 112-2 ... Incorporation request unit, 113-1, 113-2 ... Incorporation request receiving unit, 201-1, 201-2 ... Data requesting unit, 202-1, 202-2 ... Data request acceptance department.

Claims (5)

One of the first and second computers functions as an active computer with respect to provision of a predetermined service, and the rest of the first and second computers function as standby computers, causing a failure in the active computer. In the cluster system in which the standby computer takes over the service started by the active computer,
Each of the first and second computers is
Data storage means for storing data;
Update reflection storing information of the area of the data storage means in which data update has occurred while functioning as the active computer as update reflection wait information representing an area not yet reflected in the data update in the standby computer Waiting information storage means;
Update request information transmitting means for transmitting update request information for reflecting the generated data update to the standby computer to the standby computer;
Unupdated area information storage means for storing, as unupdated area information, information of an area in which data update generated in the active system computer in a state of functioning as the standby system computer is not reflected in the standby system computer;
Based on the update request information, the update request information transmitted from the update request information transmitting means of the active computer in a state of functioning as the standby computer is received, and the data update generated in the active computer is based on the update request information. To update the data storage means of the standby computer and notify the active computer of the update region information of the region in which the data update is reflected, and not update corresponding to the region in which the data update is reflected An update request information receiving means for deleting the unupdated area information when the area information is stored in the unupdated area information storage means of the standby computer;
If the update request information update area information to the operating system computer from the update request information receiving means of the standby computer for transmission to said standby computer is notified, corresponding to the region indicated by the update area information Update reflection waiting means for deleting the update reflection waiting information to be deleted from the update reflection waiting information storage means of the active computer;
When a failure occurs in the state of functioning as the standby computer and the standby computer temporarily stops operating and then recovers from the failure, it requests the active computer to start synchronization processing. The update reflection wait information stored in the update reflection wait information storage means of the active computer is acquired from the active computer, and the update reflection wait information is set as unupdated area information in the standby computer. Built-in request means for storing in the unupdated area information storage means,
When the start of synchronization processing is requested from the incorporation request unit of the standby computer while functioning as the active computer, the update stored in the update reflection waiting information storage unit of the active computer The reflection waiting information is transmitted to the incorporation request means of the standby computer, and the data of the area represented by the update reflection waiting information is read from the data storage means of the active computer, and the standby computer is read by the data. Embedded request receiving means for transmitting update request information necessary for synchronization processing for updating data in the corresponding area of the data storage means to the update request information receiving means of the standby computer ,
When the update request information receiving unit receives the update request information necessary for the synchronization processing transmitted from the built-in request receiving unit of the active computer while functioning as the standby computer, Based on the update request information, the data update generated in the active computer is reflected in the data storage means of the standby computer, and the update area information of the area in which the data update is reflected is notified to the active computer. In addition, when the unupdated area information corresponding to the area in which the data update is reflected is stored in the unupdated area information storage unit of the standby computer, the unupdated area information is deleted. Cluster system. Each of the first and second computers is
When the service is started on the computer, a reference request issued by the service is detected, and whether the area designated by the detected reference request is an unupdated area is not updated on the computer. Based on the non-updated area information stored in the area information storage means, and if it is an unupdated area, the reference request until the corresponding non-updated area information is deleted from the non-updated area information storage means 2. The cluster system computer according to claim 1, further comprising a data request monitoring unit for waiting for execution of the cluster system. Each of the first and second computers is
Specification of a reference request detected by the data request monitoring unit of the computer when the service is taken over in a state where unupdated region information indicating an unupdated region is stored in the unupdated region of the computer If the area to be updated is an unupdated area, reference area information representing the area is used to request the computer in which the service has been started and the service has been stopped until the service is taken over. A data requesting means for transmitting the request to the computer where the service is stopped,
When the reference area information is transmitted from the data request unit of the computer in which the service is started in the state where the service is stopped, the data of the area specified by the reference area information is read from the data storage unit, A data request for transmitting update request information for updating the data in the corresponding area of the data storage means of the computer on which the service has been started to the update request information receiving means of the computer on which the service is started. The cluster system according to claim 2, further comprising: a receiving unit. And further comprising at least one third computer having the same configuration as the first and second computers,
Each of the first and second computers and the at least one third computer operate in agreement with each other such that any one computer becomes an active system and the remaining computers become standby systems. Cluster control means for
The update reflection waiting information storage means included in each of the first and second computers and the at least one third computer is such that the computer functions as the active computer and each remaining computer is the standby computer. If the function is functioning as an update, the information of the area of the data storage means in which the data update has occurred in the active system computer is updated for each standby system computer to indicate the area that has not been reflected in the data update in the standby system computer. Used to store as reflection waiting information,
The cluster control means of the first and second computers and the at least one third computer are stored in the update reflection waiting information storage means of the active computer when a failure occurs in the active computer. Comparing the total area waiting for update reflection indicated by the update reflection waiting information for each standby computer, and making an agreement so that the computer having the smallest total area waiting for update reflection takes over the service. The cluster system according to claim 1. When a failure occurs in the active computer, the data storage means of each of the active computer and the standby computer applied to the cluster system in which the standby computer takes over the service started by the active computer In the synchronization method for synchronizing data,
When data update to the data storage unit of the active system computer occurs in the active system computer, update information indicating an area that has not been reflected in the data update in the standby system computer, information on the area in which the data update has occurred Storing in the update reflection waiting information storage means of the active computer as waiting information;
Sending an update request information for reflecting the generated data update to the standby computer to the standby computer from the operating system computer,
Based on the transmitted update request information from the operating system computer to the standby computer, the steps of the data update generated in the active system computer said standby computer is reflected in the data storage means of the standby computer,
A step of deleting the non-updated area information by the standby computer when the non-updated area information corresponding to the area where the data update is reflected is stored in the non-updated area information storage means of the standby computer; ,
Notifying the active computer from the standby computer of update region information of the region in which the data update is reflected;
Deleting the update reflection wait information corresponding to the area indicated by the update area information notified to the active computer from the update reflection wait information storage unit of the active computer;
A step of requesting the active computer to start synchronization processing from the standby computer when a failure occurs in the standby computer and the standby computer temporarily stops operating and then recovers from the failure. When,
When the start of the synchronization process is requested, the update reflection waiting information stored in the update reflection waiting information storage means of the active computer is transmitted from the active computer to the standby computer;
The standby computer stores the transmitted update reflection waiting information as unupdated region information in the unupdated region information storage unit of the standby computer;
The active computer reads the data in the area represented by the update reflection waiting information transmitted from the active computer to the standby computer from the data storage means of the active computer, and the data of the standby computer is determined by the data. Transmitting update request information necessary for the synchronization process for updating data in the corresponding area of the data storage means from the active computer to the standby computer ;
Based on the update request information necessary for the synchronization process transmitted from the active computer to the standby computer, the standby computer stores the data update generated by the active computer in the data storage of the standby computer. Steps reflected in the means;
If the non-updated area information corresponding to the area where the data update is reflected based on the update request information necessary for the synchronization processing is stored in the non-updated area information storage unit of the standby computer, the standby A synchronization method comprising: deleting a non-updated area information from the system computer .

Images