JP5594668B2 - Node, clustering system, clustering system control method, and program - Google Patents

Description

  The present invention relates to a high-availability clustering system composed of a plurality of nodes, and more particularly to a technique for failing over a node IP address using a virtual IP address.

  Conventionally, a clustering system is known as a system architecture suitable for realizing load balancing, scalability, and high availability of a computer system. The clustering system is configured by connecting a plurality of nodes (computer systems) via a network, and services are provided to clients from individual nodes via the network.

  In a clustering system, as one measure for realizing high availability, when one node stops a service due to a failure or maintenance, the resources of the stopped node are taken over by other nodes and the resources are taken over. There is a case where a process called failover is performed in which the service of the stopped node is continuously provided.

  An IP address for a node to provide a service to a client is one of resources to be taken over by failover. A technique for taking over an IP address using a virtual IP address (also called an IP alias) is known (see, for example, Patent Document 1).

  Patent Document 1 is composed of one active node (hereinafter referred to as “active node”) and one standby node (hereinafter referred to as “standby node”). Each node is assigned a fixed IP address, and the active node serves a client. A clustering system in which a virtual IP address for providing a virtual IP address is provided is disclosed.

  This clustering system performs a health check between nodes through a fixed IP address, and when a failure of the operation node is detected, the virtual IP address is assigned to a standby node to perform failover of the IP address.

JP 2004-32452 A

  However, although Patent Document 1 shows a basic idea of failing over a node IP address using a virtual IP address, the disclosed specific example is based on one active node and one standby node. Only the operation mode (so-called ACT / SBY operation).

  There are various operation forms of the clustering system. Besides the operation form exemplified in Patent Document 1, for example, an operation form with one operation node and a plurality of standby nodes (an form in which redundancy of ACT / SBY operation is increased) ), Operation modes with a plurality of operation nodes (so-called ACT / ACT operation), and operation modes with a plurality of operation nodes and one or more standby nodes (ACT / ACT operation with redundancy added). A clustering system can be considered.

  In such a clustering system of various operation modes, when a node IP address is failed over using a virtual IP address, it is necessary to consider a difference in processing according to the operation mode. For example, in determining an alternative node that should take over the IP address of the operation node that has stopped the service, alternative node candidates will differ depending on the operation mode.

  The node function (control program) may be customized for each operation mode so that the IP address can be appropriately failed over according to the operation mode. However, this makes the node more versatile and adaptable to different operation modes. The problem of being damaged arises.

  The present invention has been made in view of the above problems, and is a node constituting a clustering system, in which failover of an IP address using a virtual IP address is unified in a clustering system of a plurality of different operation modes. To provide a node having excellent versatility and adaptability to different operation modes, a clustering system composed of such nodes, and a control method of the clustering system executed in such nodes With the goal.

In order to solve the above problems, the node of the present invention is one of a plurality of nodes constituting a clustering system that provides services to clients, and each of the plurality of nodes is managed. One fixed address for use and one or more virtual addresses for providing the service to the client are designated, and an operation state in which the service is provided through the virtual address and a state in which the service can be provided The clustering system can be in any one of a standby state waiting in the state and a stopped state in which the service cannot be provided . A first clustering system in which nodes can be in operation, and different virtual addresses for all nodes Any one of a second clustering system in which a node is designated and all nodes can be in an operating state, and a third clustering system in which a node reserved in a standby state is added to the second clustering system. When the stop node is detected, the stop node detecting unit that detects a node that is in the stop state from the operation state among the plurality of nodes as a stop node, and the stop node is detected. The virtual address specified for the own node is equal to the virtual address specified for the stop node and the virtual address specified for the stop node is equal to the virtual address specified for the own node Since the virtual address specified for the stop node is different, it is determined that the own node is an alternative candidate for the stop node. And when the local node is determined to be the alternative candidate, it actually replaces the stop node from one or more nodes in the same state as the local node among the plurality of nodes. An alternative node determination unit that determines an alternative node, and when the local node is determined as the alternative node, if the virtual address specified in the local node is equal to the virtual address specified in the stop node, it is specified as the local node If the status of the current node is changed to the operating status while maintaining the virtual address, and if the virtual address specified for the current node is different from the virtual address specified for the stop node, the status of the current node is changed to the operating status. By additionally specifying the virtual address of the stop node to the local node while maintaining the local node, the local node is passed through the virtual address specified for the stop node. And an alternative start unit that puts the service into an operational state.

  According to this configuration, the substitution determination unit determines whether or not the own node is a candidate for the stop node, whether the own node and the stop node have different virtual addresses, and the state of the own node is SBY or ACT. Judgment is made according to a characteristic condition defined by a combination of comparisons.

  For example, in a clustering system in which the same virtual IP address is specified for all nodes and at most one node can be in an operating state, different virtual IP addresses are specified for all nodes. In a clustering system that can be in an operational state, and in a clustering system in which a node that is reserved in a standby state is added to such a clustering system, it functions in common and effectively, so a virtual IP address is used. A node capable of performing IP address failover uniformly in a clustering system of a plurality of different operation modes is realized.

  In addition, a return detection unit that detects that the stop node is in a standby state, and when the standby state is detected, a virtual address specified for the own node and a virtual address specified for the stop node, If the two nodes are different from each other, an alternative end unit may be provided that deletes the designation of the virtual address of the stop node from the own node while maintaining the state of the own node in the operating state.

  According to this configuration, the processing performed by the substitution start unit and the substitution end unit functions in a common and effective manner in, for example, the clustering system of the various operation modes described above. Therefore, an IP address using a virtual IP address Nodes that can perform the same failover in a clustering system in a plurality of different operation modes are realized.

  Further, the alternative node determination unit transmits an approval request that the own node becomes the alternative node to the one or more nodes in the same state as the own node among the plurality of nodes, and the approval node When an acceptance response to the approval request is received from all the nodes that transmitted the request, the node may be determined as the alternative node and the approval request may be committed.

  According to this configuration, it is possible to determine an alternative node without contradiction between a plurality of nodes that are alternative candidates based on the concept of two-phase commit.

  The alternative node determination unit may set a longer time as the load level of the own node is higher, and transmit the approval request after the set time has elapsed.

  This configuration intentionally delays the approval request transmission timing for nodes with a higher load level, so when multiple nodes become alternative candidates, the approval request is transmitted earlier with a lower load level. Will do. As a result, it is possible to obtain an effect that a node having a sufficient processing capacity is more likely to become an alternative node.

  The alternative node determination unit determines the alternative node as the alternative node only when there is no node having a lower load level than the own node among the nodes that transmitted the approval request, and the approval request is determined. You may commit.

  According to this configuration, only when there is no node having a lower load level than the own node, the own node is determined as an alternative node. Therefore, among the plurality of nodes that are alternative candidates, the node with the lowest load level is determined. A reasonable decision is made to be determined as an alternative node.

  The present invention can be realized not only as a node as described above but also as a clustering system including a plurality of such nodes. Further, it can be realized as a control method of a clustering system having a process executed by characteristic means included in the node as a step, or can be realized as a computer program.

According to the node of the present invention, one node among a plurality of nodes constituting a clustering system that provides services to clients, each of the plurality of nodes being one fixed for management. An address and one or more virtual addresses for providing the service to the client are specified, and an operation state in which the service is provided through the virtual address and a state in which the service can be provided are waited for. The clustering system can be in any of a standby state and a stopped state in which the service cannot be provided. In the clustering system, the same virtual address is assigned to all nodes and at most one node is in an operating state. The first clustering system that can be, and all the nodes are all assigned different virtual addresses Node second clustering system which can be the operation state, and is any one of the third clustering system plus nodes reserve remains in the standby state to the second clustering system, the 1 The own node, which is one node, includes a stop node detection unit that detects a node that has been stopped from the operation state as a stop node among the plurality of nodes, and when the stop node is detected, the own node enters a standby state. And the virtual address specified for the own node and the virtual address specified for the stop node are equal, or the own node is in an operating state and the virtual address specified for the own node and the stop node specified An alternative determination unit that determines that the own node is an alternative candidate for the stop node because the virtual address differs When it is determined that the own node is the alternative candidate, an alternative is to determine an alternative node that actually replaces the stop node from one or more nodes in the same state as the own node among the plurality of nodes. When the node determination unit and the local node are determined as the alternative node, the virtual address specified in the local node is maintained if the virtual address specified in the local node is equal to the virtual address specified in the stop node. If the virtual address specified for the local node is different from the virtual address specified for the stop node, the local node remains in the operational state and the local node remains in the operational state. By additionally specifying the virtual address of the stop node, the local node provides the service through the virtual address specified by the stop node. And an alternative start unit for setting the operation state.

  According to this configuration, the substitution determination unit determines whether or not the own node is a candidate for the stop node, whether the own node and the stop node have different virtual addresses, and the state of the own node is SBY or ACT. Judgment is made according to a characteristic condition defined by a combination of comparisons.

  For example, in a clustering system in which the same virtual IP address is specified for all nodes and at most one node can be in an operating state, different virtual IP addresses are specified for all nodes. In a clustering system that can be in an operational state, and in a clustering system in which a node that is reserved in a standby state is added to such a clustering system, it functions in common and effectively, so a virtual IP address is used. A node capable of performing IP address failover uniformly in a clustering system of a plurality of different operation modes is realized.

Schematic configuration diagram showing an example of a clustering system according to an embodiment of the present invention State transition diagram explaining the possible states of the node Block diagram showing an example of the functional configuration of a node The figure which shows an example of a state table Flow chart showing an example of virtual address control processing A flowchart showing an example of an alternative node determination process Flowchart showing an example of alternative start processing A flowchart showing an example of the substitution end process Sequence chart showing an example of operation of the system in the first embodiment Sequence chart showing an example of the operation of the system in the second embodiment

  A clustering system and nodes according to an embodiment of the present invention will be described in detail with reference to the drawings.

(Clustering system configuration)
FIG. 1 is a schematic configuration diagram illustrating an example of a clustering system 10 according to an embodiment of the present invention.

  The clustering system 10 is a system that provides services to clients 91 to 94 via a network, and is configured by connecting nodes 11 to 14 via a network 19.

  Each of the nodes 11 to 14 is a web application server based on a three-layer model, for example, and provides a service for one fixed IP address ipn (n = 1 to 4) for management and clients 91 to 94. One or more virtual IP addresses vipn (n = 1 to 4) are designated. Here, the fixed IP address and the virtual IP address are examples of the fixed address and the virtual address of the present invention, respectively. Hereinafter, for convenience, the fixed IP address and the virtual IP address may be referred to as a fixed address and a virtual address, respectively.

  Each of the nodes 11 to 14 performs management communication such as notification of the latest state of the own node, health check, and database matching control through a management fixed IP address ipn, and also performs virtual communication with the client. The service is provided through the IP address vipn.

(State that the node can take)
FIG. 2 is a state transition diagram illustrating states that the node 11 can take. The nodes 12 to 14 can assume the same state, and the node 11 will be described as a representative here.

  The node 11 is in an operation state (ACT: ACTIVE) in which a service is provided through a designated virtual IP address, a standby state (SBY: StandBY) in which the service can be provided, and a stop in which the service cannot be provided. The state can be any of a state (OUS: OUt of Service) and an initial state (INIT: INITial) in which the node is activated after the power is turned on.

  Although the present invention does not limit the types of stop states, as an example, a failure state (FAIL: FAILure) that requires repair to become SBY or ACT is one typical example of a stop state, and is also a service due to periodic inspections and power outages. A stop can also be an example of a stop state.

  In the following, for the sake of brevity, the state of a node may be represented by a symbol.

  A transition is made between INIT and SBY by inputting a startup command and a shutdown command from the operator. Between SBY and ACT, in addition to the transition by the activation command and standby command being input from the operator, the transition is made by the control performed by the alternative start / end unit 36 described later. Transition from each of INIT, SBY, and ACT to FAIL occurs when a failure occurs, and transition from FAIL to INIT occurs after repair work by the operator.

  Note that the states of the nodes 11 to 14 are exchanged between the nodes and recorded.

  For example, each time a state transition occurs, the nodes 11 to 14 may broadcast information indicating the state after the transition to all the nodes. For example, when performing communication for health check, the node 11-14 You may notify to another node.

  The nodes 11 to 14 record the statuses of the other nodes thus notified in a status table described later.

(Node configuration)
FIG. 3 is a block diagram illustrating an example of a functional configuration of the node 11. The nodes 12 to 14 are also configured in the same manner, and here, the node 11 will be described as a representative.

  The node 11 includes an HDD (hard disk device) 17, a NIC (network interface controller) 18, a service execution unit 20, and a virtual address control unit 30, and is further configured by a processor, a memory, and the like not shown. System. The service execution unit 20 and the virtual address control unit 30 are software functions that are fulfilled by the processor executing a predetermined program using the memory.

  The service execution unit 20 is an execution unit that provides a service to a client (not shown) through the virtual IP address vip1, and performs, for example, a web application server function based on a three-layer model. The web layer 21, the AP layer 22, and the DB layer 23.

  Note that such a configuration of the service execution unit 20 is an example. The service execution unit 20 only needs to provide a service to the client through the virtual IP address vip1, and is not limited to the Web application server based on the three-layer model.

  The virtual address control unit 30 is a control unit that executes failover of an IP address using a virtual IP address in a unified procedure in a clustering system having a plurality of different configurations, and includes a state management unit 31 and a stop / return detection unit 33. , An alternative determination unit 34, an alternative node determination unit 35, and an alternative start / end unit 36. The state management unit 31 has a state table 32.

  Here, the stop / return detection unit 33, the substitution determination unit 34, the substitution node determination unit 35, and the substitution start / end unit 36 are respectively a stop node detection unit, a substitution determination unit, a substitution node determination unit, and a substitution start unit according to the present invention. It is an example. The stop / return detection unit 33 and the substitution start / end unit 36 are also examples of the restoration detection unit and the substitution end unit according to the present invention, respectively.

  The state management unit 31 periodically or sequentially detects the status of the node 11 that is its own node, and records it in the detected state table 32. The situation mentioned here includes, for example, the state of the own node, the load level, and the virtual IP address specified for the own node.

  The state management unit 31 broadcasts information indicating the latest status of the own node to all the nodes whenever any change occurs in the status of the own node. For example, when the state of the own node changes, when the load level fluctuates by a certain amount, or when the virtual IP address specified for the own node is added or deleted, the latest state of the own node, Information indicating the load level and virtual IP address is broadcast to all nodes.

  The state management unit 31 refers to such information broadcast from other nodes and records the latest state, load level, and virtual IP address of each node in the state table 32.

  The state management unit 31 also performs communication for health check with the nodes 12 to 14 through the fixed IP address ipn, for example, so that the virtual IP address, state, and load level of each node are periodically The latest information collected may be recorded in the state table 32.

  FIG. 4 is a diagram illustrating an example of the state table 32 in the node 11.

  In the status table 32, for each of the node 11 that is the own node and the nodes 12 to 14 that are other nodes, the IP address specified as the node in the fixed address column 321 and the node specified in the virtual address column 322 One or more virtual IP addresses, a status column 323 indicates a node status (ACT, SBY, FAIL, or INIT symbol), and a load level column 324 indicates a node load level value (for example, a load average value). keeping.

  A similar state table 32 is also provided for the nodes 12 to 14, and the same contents regarding the states of the nodes 11 to 14 are recorded in each of the nodes 11 to 14.

(Virtual address control operation in the node)
The operation of the virtual address control unit 30 will be described with reference to the flowcharts of FIGS.

  FIG. 5 is a flowchart illustrating an example of a virtual address control process performed by the virtual address control unit 30.

  In the virtual address control unit 30, the stop / return detection unit 33 detects a node that has changed from the operation state to the stop state as a stop node (S10).

  Stop nodes are detected in several ways. For example, by directly detecting a failure occurring in the own node, the own node may be detected as a stop node, and a node for which the communication for health check performed by the state management unit 31 is interrupted is detected. You may detect as a stop node. In any case, it is desirable that the node that detects the stop node broadcasts a stop notification that identifies the detected stop node as long as communication is possible. Then, the stop node can be detected also by receiving the broadcast stop notification. By detecting the stop notification, the detection of the stop node is completed promptly in all nodes.

  When a stop node is detected (Y in S11), the substitution determination unit 34 determines whether the own node is a substitution candidate for the stop node (S20). More specifically, whether the own node is in a standby state and the virtual address designated as the own node is equal to the virtual address designated as the stop node (Y in S21), or the own node is in the operating state. In addition, when the virtual address designated as the own node is different from the virtual address designated as the stop node (Y in S22), it is determined that the own node is an alternative candidate for the stop node.

  When it is determined in step S20 that the own node is an alternative candidate, the alternative node determination unit 50 determines an alternative node that actually replaces the stop node from all the nodes that are in the same state as the own node. Node determination processing is performed (S50).

  FIG. 6 is a flowchart illustrating an example of the alternative node determination process. The main part of the alternative node determination process shown in FIG. 6 is based on a general two-phase commit procedure.

  First, the alternative node determination unit 50 sets a longer time as the load level of the own node is higher, and waits until the set time elapses (S31).

  Thereafter, the following processing is performed according to a general two-phase commit procedure. An approval request that the own node becomes an alternative node is sent to all nodes in the same state as the own node (S32), and an acceptance response returned from the node that sent the approval request is received (S33). ).

  Only when the acceptance response is received from all the nodes that transmitted the approval request (Y in S34), and there is no node having a lower load level than the own node among the nodes that transmitted the approval request (N in S35). The node is determined as an alternative node (S36), and the approval request is committed (S37). When the acceptance response is not received from all the nodes that transmitted the approval request (N in S34), and when there is a node having a lower load level than the own node among the nodes that transmitted the approval request (Y in S35), The approval request is rolled back without determining the own node as an alternative node (S38). The load level is compared by referring to the state table 32, for example.

  In the above alternative node determination processing procedure, step S31 and step S35 are not essential, but by providing both or one of step S31 and step S35, the following additional effects can be obtained respectively. be able to.

  In other words, in step S31, the approval request is transmitted to a node with a higher load level intentionally delayed. Therefore, when a plurality of nodes become alternative candidates, an approval request is transmitted at an earlier timing to a node with a lower load level. Will do. As a result, it is possible to obtain an effect that a node having a sufficient processing capacity is more likely to become an alternative node.

  In addition, by providing step S35, only when there is no node having a load level lower than that of the own node, the own node is determined as an alternative node. Therefore, among the plurality of nodes that are alternative candidates, the load level is the highest. A reasonable decision is made that the lower node is determined to be the alternate node.

  Note that step S35 is modified so that the own node is determined as an alternative node only when there is no node whose load level is lower than the reference value obtained by subtracting a predetermined margin value from the load level of the own node. May be. Then, in a practical situation where the load level of the own node and the load level of other nodes are frequently switched, all the nodes that have become alternative candidates are determined by the difference in timing for determining the load level. It is possible to reduce the possibility of a problem that the own node is not determined as an alternative node.

  If there is no other node in the same state as the own node, it is clear that there is no need for the procedure based on the two-phase commit as described above. What is necessary is just to determine an alternative node immediately.

  Referring to FIG. 5 again, when the own node is determined to be the substitute node (Y in S39), the substitute start / end unit 36 provides the service status to the own node through the virtual address designated as the stop node. An alternative start process is performed (S40).

  FIG. 7 is a flowchart illustrating an example of an alternative start process.

  If the virtual address specified for the own node and the virtual address specified for the stop node are equal (Y in S41), the substitution start / end unit 36 maintains the virtual address specified for the own node while maintaining the status of the own node. Is changed to the operating state (S42), and if the virtual address specified for the own node and the virtual address specified for the stop node are different (N in S41), the own node is maintained in the operating state. In addition, the virtual address of the stop node is designated (S43).

  Referring to FIG. 5 again, the stop / return detection unit 33 detects that the substituting stop node has returned to the standby state (S60). As an example, the return detection includes repairing a stopped node that has been replaced and entering a standby state by executing a startup command. As a result, information indicating that the stopped node has entered a standby state is broadcast from the stopped node. You may carry out by receiving.

  When the return of the replacement stop node is detected, the replacement start / end unit 36 performs a replacement end process (S70).

  FIG. 8 is a flowchart illustrating an example of the substitution end process.

  If the virtual address specified for the own node and the virtual address specified for the stop node are different (N in S71), the substitution start / end unit 36 continues from the own node to the stop node while maintaining the state of the own node. The designation of the virtual address is deleted (S73). If the virtual address specified for the own node is equal to the virtual address specified for the stop node (N in S71), the substantial substitution end process is not performed. In this case, the own node continues to provide the service in the execution state.

(Example)
In the following, a node configured as described above is applied to a clustering system of two different operation modes without changing the configuration of the node itself, and in either operation mode, an IP address using a virtual IP address. Explain that the failover is properly performed in a unified procedure.

  As a first embodiment, an application example to a clustering system in an operation mode (a mode in which redundancy of ACT / SBY operation is increased) in which one node is in an operation state and a plurality of nodes are in a standby state will be described.

  The clustering system according to the first embodiment corresponds to the configuration in which the node 14 is excluded from the clustering system 10 of FIG. 1, the same virtual IP address A is designated for the nodes 11 to 13, and at most one of the nodes 11 to 13. One node is operated in a form that can be an ACT. The client receives services from this clustering system through the virtual IP address A.

  FIG. 9 is a sequence chart showing an example of the operation of the system in the first embodiment. In the sequence chart of FIG. 9, reference numerals of corresponding steps in the flowcharts of FIGS. 5 to 8 are shown.

  Initially, node 11 is an ACT and the client is receiving service from node 11 through virtual IP address A. Further, the nodes 12 and 13 are SBY, and are waiting in a state where a service can be provided to a client (that is, in a state where the service can be an ACT at any time and the provision of the service is suspended).

  For example, a failure occurs in the hard disk device at the node 11. Thereby, the node 11 detects its own node as a stop node, becomes FAIL, and broadcasts a stop notification STOP that identifies the node 11. Here, the stop notification STOP includes information for specifying the node 11 (for example, the fixed IP address of the node 11). The nodes 12 and 13 detect the node 11 as a stop node by receiving the stop notification STOP (S10).

  The nodes 11, 12, and 13 perform substitution determination (S20). The nodes 12 and 13 stop when their own node is SBY and the virtual IP address A designated as the own node is equal to the virtual IP address A designated as the stop node (Y in S21). Judged as an alternative candidate for the node. Since the node 11 is FAIL due to a failure, the condition for substitution determination is not satisfied.

  The nodes 12 and 13 perform an alternative node determination process (S30). The nodes 12 and 13 try to transmit the respective approval requests to the nodes 13 and 12 that are SBY as the own node. If the node 12 transmits an approval request before the node 13 (S31), the node 13 stops transmitting its own approval request and returns an acceptance response to the node 12, and the node 12 accepts the approval. A response is received (S32). The node 12 determines its own node as an alternative node (S34), and transmits a commit to the node 13 (S35). Thereby, the node 12 is determined as an alternative node between the nodes 12 and 13.

  Since the virtual IP address of the node 12 is the same as that of the stop node, the node 12 performs an alternative start process for making the node ACT (S42). As a result, the client can continue to receive services from the node 12 through the virtual IP address A.

  The node 11 is repaired (S50). After the node 11 is activated and becomes INIT, the node 11 becomes SBY when a startup command is input from the operator (S55).

  The node 12 detects the return of the node 11 by receiving information representing the latest state that the node 11 broadcasts every time the state transition occurs (S60). Since the virtual IP addresses of the node 12 and the stop node coincide with each other, the node 12 does not perform a substantial replacement end process, and the node 12 continues to provide services through the virtual IP address A while remaining in ACT. As a result, the client can continue to receive services from the node 12 through the virtual IP address A.

  In this way, in an operation mode in which the clustering system provides a service through the only virtual address A, service interruption is realized.

  Next, as a second embodiment, an application example to a clustering system in an operation mode (a mode in which redundancy is added to ACT / ACT operation) in which three nodes are in an operating state and one node is in a standby state will be described. .

  The clustering system according to the second embodiment corresponds to the configuration of the clustering system 10 in FIG. 1, and different virtual IP addresses A to D are designated for the nodes 11 to 14, and the nodes 11 to 13 can all be ACT. It is operated by. Clients receive services from this clustering system through virtual IP addresses A-C.

  In this operation mode, the node 14 is maintained in SBY, and does not become a substitute candidate for the stop node in the virtual IP address control process. The node 14 is reserved as equipment for replacing the nodes 11 to 13 according to the operation of the operator, such as when the repair of the nodes 11 to 13 is prolonged.

  FIG. 10 is a sequence chart showing an example of the operation of the system in the second embodiment. In the sequence chart of FIG. 10, reference numerals of corresponding steps in the flowcharts of FIGS. 5 to 8 are shown.

  Initially, the nodes 11 to 13 are ACT, and the client receives service from the node 11 through the virtual IP address A. The node 14 is SBY and is waiting in a state where it can provide a service to the client (that is, in a state where it can become an ACT at any time and is suspending the provision of the service).

  For example, a failure occurs in the hard disk device at the node 11. Thereby, the node 11 detects its own node as a stop node, becomes FAIL, and broadcasts a stop notification STOP that identifies the node 11. Here, the stop notification STOP includes information for specifying the node 11 (for example, the fixed IP address of the node 11). The nodes 12 to 14 detect the node 11 as a stop node by receiving the stop notification STOP (S10).

  The nodes 11 to 14 make an alternative determination (S20). The nodes 12 and 13 have their own nodes being ACT, and because the virtual IP addresses B and C designated as the own nodes are different from the virtual IP addresses A designated as the stop nodes (Y in S22), Is an alternative candidate for the stop node. Since the node 11 is FAIL due to a failure, the condition for substitution determination is not satisfied. Since the node 14 is SBY and the virtual IP address D designated as the own node is different from the virtual IP address A designated as the stop node, the alternative determination condition is not satisfied.

  The nodes 12 and 13 perform an alternative node determination process (S30). The nodes 12 and 13 try to transmit the respective approval requests to the nodes 13 and 12 that are ACTs in the same manner as the own node. If the node 12 transmits an approval request before the node 13 (S31), the node 13 stops transmitting its own approval request and returns an acceptance response to the node 12, and the node 12 accepts the approval. A response is received (S32). The node 12 determines its own node as an alternative node (S34), and transmits a commit to the node 13 (S35). Thereby, the node 12 is determined as an alternative node between the nodes 12 and 13.

  Since the node 12 has a different virtual IP address between the own node and the stop node, the node 12 performs an alternative start process for additionally specifying the stop node's virtual IP address A to the own node. The node 12 enters a state in which the virtual IP addresses B and A are designated (S43). As a result, the client can continue to receive services from the node 12 through the virtual IP address A.

  The node 11 is repaired (S50). After the node 11 is activated and becomes INIT, the node 11 becomes SBY when a startup command is input from the operator (S55).

  The node 12 detects the return of the node 11 by receiving information representing the latest state broadcasted every time the node 11 changes state (S60). Since the virtual IP address is different between the own node and the stop node, the node 12 performs an alternative end process of deleting the virtual IP address A of the stop node from the own node. The virtual IP address specified for the node 12 is only B again (S73).

  As a result, the node 12 provides the service only through the virtual IP address B, and the load level of the node 12 is immediately normalized.

  Immediately thereafter, the node 11 becomes ACT when an activate command is input from the operator (S80). As a result, the client can receive service from the node 11 again through the virtual IP address A.

(Summary)
Comparing the first example and the second example described above, the node of the present embodiment has a configuration in which redundancy of ACT / SBY operation is increased without changing the configuration of the node itself, and Applicable to clustering systems in two different operation modes, with redundancy added to ACT / ACT operations. In either operation mode, failover of IP addresses using virtual IP addresses is appropriate with a unified procedure. It can be seen that

  In particular, in the node according to the present embodiment, whether or not the own node becomes an alternative candidate for the stop node, whether the own node and the stop node are different in virtual IP address, and the status of the own node is SBY or ACT. Judgment is made according to a characteristic condition defined by a combination of comparisons.

  This judgment condition is that in the ACT / SBY operation in which the same virtual IP address is specified for all nodes and at most one node can be ACT, different virtual IP addresses are specified for all nodes, and all nodes are ACT. In ACT / ACT operation that can be used, and in addition to the ACT / ACT operation, in addition to the node that should be reserved with SBY, it functions in common and effectively, so it is excellent for configuring a clustering system. A node with high versatility and high adaptability to different operation modes is realized.

  As mentioned above, although the node of this invention, the clustering system comprised from such a some node, and the control method of a clustering system were demonstrated based on embodiment, this invention is limited to this embodiment. is not. Unless it deviates from the meaning of the present invention, those in which various modifications conceived by those skilled in the art are applied to the present embodiment are also included in the scope of the present invention.

  The node, clustering system, clustering system control method, and program of the present invention can be suitably used for a high-availability clustering system that performs IP address failover.

10 Clustering system 11-14 Node 17 HDD
18 NIC
DESCRIPTION OF SYMBOLS 19 Network 20 Service execution part 21 WEB layer 22 AP layer 23 DB layer 30 Virtual address control part 31 State management part 32 State table 33 Stop / return detection part 34 Alternative judgment part 35 Alternative node determination part 36 Alternative start end part 91-94 Client

Claims (9)

One of a plurality of nodes constituting a clustering system that provides services to clients,
In each of the plurality of nodes, one fixed address for management and one or more virtual addresses for providing the service to the client are designated, and the service is provided through the virtual address. Can be any of a state, a standby state waiting in a state where the service can be provided, and a stopped state where the service cannot be provided,
In the clustering system, the first virtual clustering system in which the same virtual address is specified for all nodes and at most one node can be in an operating state, and different virtual addresses are specified in all nodes and all nodes are in an operating state. Any one of a second clustering system to be obtained and a third clustering system obtained by adding a node to be reserved in a standby state to the second clustering system,
The node that is the one node is
A stop node detection unit that detects a node that is in a stop state from an operation state among the plurality of nodes as a stop node;
When the stop node is detected, the own node is in a standby state and the virtual address specified for the own node is equal to the virtual address specified for the stop node, or the own node is in an operating state and An alternative determination unit that determines that the own node is an alternative candidate for the stop node because the virtual address specified for the own node is different from the virtual address specified for the stop node;
When it is determined that the own node is the alternative candidate, an alternative is to determine an alternative node that actually replaces the stop node from one or more nodes in the same state as the own node among the plurality of nodes. A node determination unit;
When the own node is determined to be the alternative node, if the virtual address specified for the own node and the virtual address specified for the stop node are equal, the virtual address specified for the own node is maintained while maintaining the virtual address specified for the own node. When the status is changed to the operating status, and the virtual address specified for the own node is different from the virtual address specified for the stopped node, the virtual node of the stopped node is set to the own node while maintaining the status of the own node. A node comprising: an alternative start unit that, by additionally specifying an address, causes the local node to enter an operation state in which the service is provided through the virtual address specified by the stop node. further,
A return detection unit for detecting that the stop node is in a standby state;
When the standby state is detected, if the virtual address specified for the own node is different from the virtual address specified for the stop node, the state of the stop node is maintained from the own node while maintaining the state of the own node. The node according to claim 1 , further comprising: an alternative end unit that deletes designation of a virtual address. The alternative node determination unit transmits an approval request that the own node becomes the alternative node to the one or more nodes in the same state as the own node among the plurality of nodes, and The node according to claim 1, wherein when an acceptance response to the approval request is received from all the transmitted nodes, the node is determined as the alternative node and the approval request is committed. The node according to claim 3 , wherein the alternative node determination unit sets a longer time as the load level of the own node is higher, and transmits the approval request after the set time has elapsed. The alternative node determination unit determines the own node as the alternative node and commits the approval request only when there is no node having a lower load level than the own node among the nodes that transmitted the approval request. The node according to claim 3 . A clustering system composed of a plurality of nodes and providing services to clients,
In each of the plurality of nodes, one fixed address for management and one or more virtual addresses for providing the service to the client are designated, and the service is provided through the virtual address. Can be any of a state, a standby state waiting in a state where the service can be provided, and a stopped state where the service cannot be provided,
In the clustering system, the first virtual clustering system in which the same virtual address is specified for all nodes and at most one node can be in an operating state, and different virtual addresses are specified in all nodes and all nodes are in an operating state. Any one of a second clustering system to be obtained and a third clustering system obtained by adding a node to be reserved in a standby state to the second clustering system,
Each of the nodes
A stop node detection unit that detects a node that is in a stop state from an operation state among the plurality of nodes as a stop node;
When the stop node is detected, the own node is in a standby state and the virtual address specified for the own node is equal to the virtual address specified for the stop node, or the own node is in an operating state and An alternative determination unit that determines that the own node is an alternative candidate for the stop node because the virtual address specified for the own node is different from the virtual address specified for the stop node;
When it is determined that the own node is the alternative candidate, an alternative is to determine an alternative node that actually replaces the stop node from one or more nodes in the same state as the own node among the plurality of nodes. A node determination unit;
When the own node is determined to be the alternative node, if the virtual address specified for the own node and the virtual address specified for the stop node are equal, the virtual address specified for the own node is maintained while maintaining the virtual address specified for the own node. When the status is changed to the operating status, and the virtual address specified for the own node is different from the virtual address specified for the stopped node, the virtual node of the stopped node is set to the own node while maintaining the status of the own node. A clustering system comprising: an alternative start unit that places an own node into an operation state in which the service is provided through a virtual address specified as the stop node by additionally specifying an address. A control method of a clustering system configured by a plurality of nodes and providing services to clients,
In each of the plurality of nodes, one fixed address for management and one or more virtual addresses for providing the service to the client are designated, and the service is provided through the virtual address. Can be any of a state, a standby state waiting in a state where the service can be provided, and a stopped state where the service cannot be provided,
In the clustering system, the first virtual clustering system in which the same virtual address is specified for all nodes and at most one node can be in an operating state, and different virtual addresses are specified in all nodes and all nodes are in an operating state. Any one of a second clustering system to be obtained and a third clustering system obtained by adding a node to be reserved in a standby state to the second clustering system,
The own node which is one of the plurality of nodes includes a stop node detection unit, an alternative determination unit, an alternative node determination unit, and an alternative start unit.
In the stop node detection unit, a node that is in a stop state from an operation state among the plurality of nodes is detected as a stop node,
When the stop node is detected, the substitution determination unit determines that the own node is in a standby state and the virtual address specified for the own node is equal to the virtual address specified for the stop node, or the own node Is determined to be an alternative candidate for the stop node because the virtual address specified for the own node and the virtual address specified for the stop node are different,
If it is determined that the node is the replacement candidate, the replacement node determination unit actually replaces the stop node from one or more nodes in the same state as the node among the plurality of nodes. Determine an alternate node to
When the own node is determined to be the alternative node, if the virtual address specified for the own node and the virtual address specified for the stop node are equal, the virtual address specified for the own node is maintained while maintaining the virtual address specified for the own node. When the status is changed to the operating status, and the virtual address specified for the own node is different from the virtual address specified for the stopped node, the virtual node of the stopped node is set to the own node while maintaining the status of the own node. A control method in which, by specifying an additional address, the substitution start unit sets its own node to an operation state in which the service is provided through a virtual address specified as the stop node. A computer-executable program for controlling a clustering system configured with a plurality of nodes and providing services to clients,
In each of the plurality of nodes, one fixed address for management and one or more virtual addresses for providing the service to the client are designated, and the service is provided through the virtual address. Can be any of a state, a standby state waiting in a state where the service can be provided, and a stopped state where the service cannot be provided,
In the clustering system, the first virtual clustering system in which the same virtual address is specified for all nodes and at most one node can be in an operating state, and different virtual addresses are specified in all nodes and all nodes are in an operating state. Any one of a second clustering system to be obtained and a third clustering system obtained by adding a node to be reserved in a standby state to the second clustering system,
In its own node that is one of the plurality of nodes,
A stop node detection step of detecting, as a stop node, a node that is in a stop state from an operation state among the plurality of nodes;
When the stop node is detected, the own node is in a standby state and the virtual address specified for the own node is equal to the virtual address specified for the stop node, or the own node is in an operating state and An alternative determination step for determining that the own node is an alternative candidate for the stop node because the virtual address specified for the own node is different from the virtual address specified for the stop node;
When it is determined that the own node is the alternative candidate, an alternative is to determine an alternative node that actually replaces the stop node from one or more nodes in the same state as the own node among the plurality of nodes. A node determination step;
When the own node is determined to be the alternative node, if the virtual address specified for the own node and the virtual address specified for the stop node are equal, the virtual address specified for the own node is maintained while maintaining the virtual address specified for the own node. When the status is changed to the operating status, and the virtual address specified for the own node is different from the virtual address specified for the stopped node, the virtual node of the stopped node is set to the own node while maintaining the status of the own node. A program that causes a computer to execute an alternative start step of setting an own node to an operation state of providing the service through a virtual address specified by the stop node by additionally specifying an address. A computer-readable recording medium storing the program according to claim 8 .

Images