JP2008305353A - Cluster system and fail-over method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To control the execution of fail-over in accordance with a processing state of a program. <P>SOLUTION: A server computer 102 constituting a cluster system 101 is provided with a service providing part 112 for providing a service to client computers 104 to 106, a cluster processing part 111 for executing fail-over when detecting operation stop of other server computer 102 or when receiving an instruction to execute fail-over, and a service providing part state monitoring part 121 for monitoring a processing state of the service providing part 112. The service providing part state monitoring part 121 periodically determines whether a processing time of the service providing part 112 exceeds a predetermined critical processing time, and instructs the cluster processing part 111 to execute fail-over when it is determined that the processing time exceeds the predetermined critical processing time. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique for controlling the execution of failover.

  As a system for improving the reliability and processing performance of a computer system, there is a cluster system in which a plurality of computers are loosely coupled. As a cluster system, for example, a load distribution type that reduces the processing load per server computer, or a business application on the server computer when a failure occurs in a server computer (active server) on which the business application operates A failover type is known in which resources are transferred to another server computer (standby server) to continue the business. For example, Patent Document 1 describes a method for executing failover by changing the execution priority of a resource in accordance with the execution priority of a resource that is taken over and a resource that is operating at a failover destination in a cluster system. .

JP-A-11-353292

  In a conventional failover type cluster system such as Patent Document 1, for example, a server computer confirms normal operation of another server computer to be monitored by heartbeat communication, and stops the server computer to be monitored, that is, Failover is executed when a down due to a hardware failure is detected.

  However, in the above cluster system, the processing status of the program running on the server computer is not monitored. In other words, failover is not executed when program responsiveness is reduced due to a shortage of resources such as CPU or memory or a software failure. Therefore, the cluster system cannot immediately recover the responsiveness of service provision to the client to a normal state.

  Therefore, an object of the present invention is to provide a technique for controlling the execution of failover in accordance with not only hardware failure but also processing status of a program in a cluster system.

  In order to solve the above problem, according to the first aspect, a server constituting a cluster system having a plurality of servers each including a failover processing unit that performs failover processing by receiving an instruction to execute failover, A storage unit for storing state information for specifying a state of a processing state for performing a failover, a state monitoring unit for specifying a processing state at the time of providing a service, and a processing state specified by the state monitoring unit are the state information. A failover execution instructing unit for instructing the failover processing unit to execute failover when the specified processing status is matched.

  Further, the storage unit stores condition information for specifying a condition for performing a failover in correspondence with each state information, and the processing status specified by the status monitoring unit is a process specified by the status information. When the processing status that matches the status and the processing status specified by the status monitoring unit satisfies the condition information corresponding to the status information determined to match, the failover execution instruction unit causes the failover processing unit to execute failover. Instruct.

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

  The first embodiment will be described below. In the first embodiment, a configuration in which the present invention is applied to a network boot system will be described as an example. First, an outline of the network boot system will be described.

  In the network boot system, a boot image composed of a file group such as an OS and an application program (hereinafter referred to as an application) necessary for booting the client computer is stored on the server computer. When the client computer is activated, the client computer reads the activation image on the server computer via the network and executes booting.

  In a network boot system composed of one server computer, if the server computer goes down due to a hardware failure or a software failure, the network boot service cannot be provided. That is, the client computer cannot be booted and activated via the network. As a result, the user cannot use the client computer. In addition, if the responsiveness of the server program for providing the network boot service on the server computer is reduced, the client computer may have a longer boot time due to the network boot and may not start due to a timeout or the like. is there. As a result, the user cannot immediately start using the client computer.

  As a network boot method, for example, PXE (Preboot Execution Environment) boot is known. In PXE boot, a server computer sets a boot image for each client computer, and stores the boot image. At the time of booting, the client computer transmits the MAC address of its own network card to the server computer. Then, the server computer transmits an activation image corresponding to the received MAC address to the client computer.

  FIG. 1 is a block diagram showing a system configuration to which the first embodiment is applied and a functional configuration of the server computer. Since the server computer 102b has the same functional configuration as the server computer 102a, the functional configuration is not shown. Further, when the server computer 102a and the server computer 102b are not distinguished, they are simply expressed as the server computer 102.

  As shown in this figure, in this system, a cluster system 101 and client computers 104, 105 and 106 are connected via a LAN (Local Area Network) 107. Of course, the connection method is not limited to the LAN, and other methods may be used.

  The cluster system 101 is composed of a server computer 102a and a server computer 102b, and is duplicated. That is, one of the server computer 102a and the server computer 102b operates as an active system, and the other operates as a standby system. The server computer 102a and the server computer 102b are connected by a LAN 108 and perform heartbeat communication via the LAN 108 to monitor the operation status of each other. Of course, the connection method between the server computer 102a and the server computer 102b is not limited to LAN, and may be, for example, serial communication.

  On the server computer 102, a control unit 150 and a storage unit 152 are constructed. The control unit 150 includes a cluster processing unit 111, a service providing unit 112, and a service providing unit status monitoring unit 121. The storage unit 152 includes a service providing unit operation history table 113, a service providing unit characteristic table 122, a service providing unit state table 123, and a failover instruction rule table 124.

  The server computer 102 configured with the above functions is, for example, a CPU 11, a main storage device 12, an auxiliary storage device 13 such as a hard disk, and a portable type such as a CD-ROM as shown in FIG. A general computer including a reading device 16 that reads information from a storage medium, an input device 14 such as a keyboard and a mouse, an output device 15 such as a display, and a communication device 17 for communicating with an external device. It is feasible.

  For example, the control unit 150 and the storage unit 152 are achieved by the CPU 11 loading a predetermined program stored in advance in the auxiliary storage device 13 into the main storage device 12 and executing it. The storage unit 152 is realized using the main storage device 12 when temporarily holding data, and using the auxiliary storage device 13 when holding data continuously. The program is downloaded from the portable storage medium via the reading device 16 or from the network via the communication device 17 to the auxiliary storage device 13, and then loaded onto the main storage device 12 and executed by the CPU 11. You may do it. Alternatively, the program may be directly loaded on the main storage device 12 from a portable storage medium via the reading device 16 or from a network via the communication device 17 and executed by the CPU 301.

  Returning to FIG. 1, the cluster processing unit 111 operates on each of the active and standby server computers 102 and performs heartbeat communication with each other via the LAN 108. The standby cluster processing unit 111 detects that the survival signal from the active cluster processing unit 111 has been interrupted, and performs failover. Also, the active cluster processing unit 111 causes the standby cluster processing unit 111 to execute a failover in response to an instruction from the active service providing unit status monitoring unit 121.

  Hereinafter, the operation of the cluster processing unit 111 will be specifically described by taking as an example the case where the server computer 102a is the active system and the server computer 102b is the standby system.

  In this case, under the control of the cluster processing unit 111, the service providing unit 112 on the server computer 102a is operating, while the service providing unit 112 on the server computer 102b is stopped. The service providing unit 112 will be described later. In addition, the cluster processing unit 111 on the server computer 102a and the cluster processing unit 111 on the server computer 102b perform heartbeat communication via the LAN 108, and monitor the operation status of each other's server computer.

  Here, for example, when the server computer 102a is stopped due to a hardware failure, the cluster processing unit 111 on the server computer 102b detects that the survival signal from the server computer 102a has been interrupted. Then, the server computer 102a is regarded as down, and resource takeover (failover) is executed. The resources taken over include the data used by the service providing unit 112 of the server computer 102a for processing, the alias IP address on the LAN 107 side, the processing of the service providing unit 112, and the like.

  In taking over the IP address, the cluster processing unit 111 on the server computer 102b uses the same alias IP address as the network interface alias on the LAN 107 side of the server computer 102a to set the network interface on the LAN 107 side of the server computer 102b. Done by adding. An alias IP address is an IP address that is an alias for an IP address and is easy to add or change. The processing of the service providing unit 112 is taken over when the cluster processing unit 111 on the server computer 102b activates the service providing unit 112 on the server computer 102b. In addition, data transfer is performed, for example, via a shared storage device to which the server computer 102 is connected.

  After the handover of resources from the server computer 102a to the server computer 102b, the server computer 102b switches from the standby system to the active system. The server computer 102a remains stopped, but switches from the active system to the standby system when returning from the stop.

  As described above, when the active server computer 102a stops, the server computer 102b takes over resources and switches from the standby system to the active system. Then, the activated service providing unit 112 starts providing services to the client computers 104 to 106 via the LAN 107 using the inherited data. Since the client computers 104 to 106 can access the server computer 102b using the same IP address, the service providing unit 112 can receive services as usual regardless of the execution of failover.

  The service providing unit 112 is, for example, a server program, an application program, a business processing program, or the like, and a plurality of service providing units 112 may be provided. In the present embodiment, the service providing unit 112 is a DHCP (Dynamic Host Configuration Protocol) server and a TFTP (Trivial File Transfer Protocol) server that are necessary as a server program for network booting. The DHCP server is a server program for automatically setting network information such as an IP address of a client computer. The TFTP server is a server program for transferring a file to a client computer without authentication of a user name or the like. The file to be transferred is a startup image composed of a group of OS and application files necessary when the client computer is started.

  The service providing unit 112 on the standby server computer 102 does not operate because it is not activated under the control of the cluster processing unit 111. When the server computer 102 is switched to the active system, the service providing unit 112 is activated and starts operating under the control of the cluster processing unit 111.

  On the other hand, the active service providing unit 112 provides services to the client computers 104 to 106, and stores an operation history in real time in the service providing unit operation history management table 113 shown in FIG.

  FIG. 2 is a diagram illustrating a configuration of the service providing unit operation history table 113. The service providing unit operation history table 113 stores operation history information in which the date 201, the time 202, the host name 203, the service providing unit name 204, and the content 205 are associated with each other. The date 201 stores the date on which the service providing unit 112 operates. The time 202 stores the time when the service providing unit 112 operates. The host name 203 stores a host name that identifies the client computer that has accessed the service providing unit 112. The service providing unit name 204 stores the name of the service providing unit that identifies the service providing unit 112. The content 205 stores details of the operation of the service providing unit 112.

  Specifically, an example of an operation history stored in the service providing unit 112 (DHCP server, TFTP server) will be described with reference to FIG. Hereinafter, “client1”, “client2”, and “client3” indicate host names of the client computers 104 to 106. “Dhcpd” indicates the program name of the DHCP server, and “tftpd” indicates the program name of the TFTP server.

  The operation history 210 in the first line to the operation history 240 in the fourth line indicate the contents of the operation history of the service providing unit 112 when the client computer 104 “client1” is network booted. In the network boot of the client computer 104 “client1”, after the DHCP server “dhcpd” provides an IP address to the client computer 104 “client1”, the TFTP server “tftpd” transfers the startup image to the client computer 104 “client1” as a file. Is executed.

  The operation history 210 indicates that the DHCP server “dhcpd” receives access from the client computer 104 “client1” at the time “12:00:01” on the date “2007 01 06” and starts providing the service “start "Indicates that you did. The operation history 220 shows that the DHCP server “dhcpd” has finished “end” providing the service to the client computer 104 “client1” at the time “12:00:03” on the date “2007 01 06”. Show. The end “end” means that the DHCP server “dhcpd” has entered a service acceptance waiting state, and is not in a state where the operation has been stopped. The operation history 230 indicates that the TFTP server “tftpd” accepts access from the client computer 104 “client1” at the time “12:00:04” on the date “2007 01 06” and starts providing the service “start "Indicates that you did. The operation history 240 shows that the TFTP server “tftpd” has finished “end” providing the service to the client computer 104 “client1” at the time “12:00:24” on the date “2007 01 06”. Show. The end “end” means that the TFTP server “tftpd” has entered a service acceptance waiting state, and is not in a state where the operation has been stopped.

  As described above, the service providing unit 112 provides a service to the client computer, and stores the operation history in the service providing unit operation history management table 113 in real time. As will be described later, the contents of the service providing unit operation history table 113 are used for obtaining a processing time for the service providing unit 112 to provide a service to the client computer.

  Next, the service providing unit status monitoring unit 121 will be described.

  The service providing unit status monitoring unit 121 executes the following processing at a predetermined cycle. With reference to the operation history stored in the service providing unit operation history table 113, the service providing process elapsed time of the service providing unit 112 is calculated, and the time is stored in the service providing unit state table 123. Further, the life / death state of the service providing unit 112 is acquired from, for example, an OS (Operating System) and stored in the service providing unit state table 123. Then, an instruction rule corresponding to the processing status (life / death state, elapsed processing time) of the service providing unit 112 stored in the service providing unit state table 123 is searched from the failover instruction rule table 124. In addition, when the condition defined in the selected instruction rule is satisfied, an instruction to execute failover is issued to the cluster processing unit 111.

  Hereinafter, the processing of the service providing unit status monitoring unit 121 will be specifically described with reference to a table used for the processing. The service providing unit status monitoring unit 121 operates on the active server computer 102 on which the service providing unit 112 operates.

  FIG. 3 is a diagram illustrating a configuration of the service providing unit characteristic table 122. The user sets the operation characteristics of the service providing unit 112 in this table. The service providing unit characteristic table 122 stores characteristic information in which the service providing unit name 301, the average processing time 303, and the limit processing time 304 are associated with each other. The service providing unit name 301 stores the name of the service providing unit that identifies the service providing unit 112. Specifically, in this embodiment, the DHCP server program name “dhcpd” and the TFTP server program name “tftpd” are stored.

  The average processing time 303 stores an average processing time (seconds) required for the process in which the service providing unit 112 provides a service to the client computer. Specifically, for example, the processing time obtained based on the operation history of the service providing unit operation history table 113 shown in FIG. 2 is assumed to be an average value, and the value is stored. In other words, the time required for the DHCP server “dhcpd” to provide a service to the client computer is “2” based on the difference between the time 202 of the operation history 210 and the operation history 220. Similarly, from the difference between the time 202 of the operation history 230 and the operation history 240, the time required for the TFTP server “tftpd” to provide a service to the client computer is “20”.

  The limit processing time 304 stores a limit processing time (seconds) as the required performance of the service providing unit 112 when the response time of the service providing unit 112 decreases due to some abnormality. For example, here, three times the average processing time 303 is set as the limit processing time, and the limit processing time 304 stores “6” and “60”, respectively. When the processing time of the service providing unit 112 is proportionally increased N times by the total number N of simultaneous accesses from the client computers 104 to 106, the limit processing time 304 is expressed as a proportional relational expression, for example, “6 × N “,“ 60 × N ”can be stored. Here, the total number of simultaneous accesses refers to the number of accesses within the average processing time from the start of a certain process. In the following example, it is assumed that the processing time is independent of the total number N of simultaneous accesses.

  FIG. 4 is a diagram illustrating a configuration of the service providing unit state table 123. The service providing unit status table 123 is periodically updated by the service providing unit status monitoring unit 121 and stores the latest information on the life / death status of the service providing unit 112 and the elapsed processing time. The service providing unit state table 123 stores state information in which the service providing unit name 401, the life / death state 403, and the processing elapsed time 404 are associated with each other. The service providing unit name 401 stores the name of the service providing unit that identifies the service providing unit 112. In this embodiment, the program name “dhcpd” of the DHCP server to be monitored and the program name “tftpd” of the TFTP server are stored.

  The life / death state 403 stores the life / death state (operating or stopped) of the service providing unit 112. Specifically, “alive” is stored as active and “dead” is stored as stopped. The life / death state is acquired when the service providing unit status monitoring unit 121 makes an inquiry to the OS, for example.

  The process elapsed time 404 stores the elapsed time (seconds) after the service providing unit 112 starts the process of providing a service to the client computer.

  Specifically, the service providing unit status monitoring unit 121 refers to the service providing unit operation history table 113 illustrated in FIG. 2 and performs processing based on the difference between the date 201 and time 202 of the referenced history information and the current date and time. The elapsed time 404 is obtained. For example, the current date and time referring to the service providing unit operation history table 113 is a date of “12:00:14” on the date of “January 06, 2007” “12:00:14” (history information 240 is still Suppose that it is not stored). In this case, in the history information 220 of the DHCP server “dhcpd”, the content 205 is “end”, and the time “12:00:03” is before the current time, so the DHCP server “dhcpd” The elapsed processing time is “0”. On the other hand, since the content 205 is “start” in the history information 230 of the TFTP server “tftpd”, the elapsed processing time is the difference between the current time “12:00:14” and the time “12:00:04”. "10". As described above, the elapsed processing time of each service providing unit 112 is calculated and stored in the elapsed processing time 404.

  FIG. 5 is a diagram showing the configuration of the failover instruction rule table 124. The user sets a rule for designating the failover operation according to the processing status of the service providing unit 112 in this table. Since the user can arbitrarily set the failover, it is possible to finely control the failover. The failover instruction rule table 124 stores instruction rule information in which a life / death state 501 for each service providing unit 112, a processing elapsed time 502 for each service providing unit 112, and a rule 503 are associated with each other.

  The life / death state 501 stores the life / death state for the service providing unit name 401 of the monitored service providing unit 112 stored in the service providing unit state table 123. Similarly, the processing time 502 stores the processing elapsed time (seconds) for the service providing unit name 401 of the monitored service providing unit 112. These values (life / death status, processing elapsed time) are data used as a reference value or threshold value for comparison with the processing status of the service providing unit 112 stored in the service providing unit status table 123. In this embodiment, the life / death state 501 stores the life / death states 504 and 505 of the DHCP server “dhcpd” and the TFTP server “tftpd” to be monitored. The process elapsed time 502 stores the process elapsed times 506 and 507 of the DHCP server “dhcpd” and the TFTP server “tftpd” to be monitored. The user sets an arbitrary combination of the life / death state 501 and the processing elapsed time 502 and a rule 503 associated with them.

  The rule 503 stores the condition 508 and the execution content 509 in association with each other. As described above, the rule 503 corresponds to each combination of the life / death states 504 and 505 for each service providing unit 112 and the processing elapsed times 506 and 507 for each service providing unit 112. The condition 508 stores a condition for executing the execution content 509.

  For example, when the condition 508 is “none”, the service providing unit status monitoring unit 121 executes the content of the execution content 509 unconditionally. When the condition 508 is “immediately after completion of processing by the service providing unit”, the service providing unit status monitoring unit 121 indicates that the specified service providing unit has started providing service information (the content 205 is “start”). ), The content of the execution content 509 is executed when it is detected that the history information (content 205 is “end”) is stored in the service providing unit operation history table 113. When the condition 508 is “after the limit processing time of the service providing unit”, the service providing unit status monitoring unit 121 indicates that the processing elapsed time 404 of the service providing unit state table 123 of the specified service providing unit is the service providing unit characteristic. When it is detected that the limit processing time 304 of the table 122 has been exceeded, the contents of the execution contents 509 can be executed. These condition contents may be combined with an AND condition or an OR condition.

  The execution content 509 stores the content of processing to be executed by the service providing unit status monitoring unit 121. For example, when the execution content 509 is “normal processing”, the service providing unit status monitoring unit 121 does nothing. When the execution content 509 is “failover execution”, the service providing unit status monitoring unit 121 instructs the cluster processing unit 111 to execute failover.

  Using the failover instruction rule table 124 described above, the service providing unit status monitoring unit 121 periodically performs processing according to the instruction rule. First, the service providing unit status monitoring unit 121 acquires the life / death state 403 and the elapsed processing time 404 of each service providing unit 401 stored in the service providing unit state table 123. After that, the instruction rule information that matches the contents of the life / death state 501 and the processing elapsed time 502 is selected from the acquired life / death state and processing elapsed time and the instruction rule information 511 to 541. Then, it is determined whether or not the condition 508 of the selected instruction rule information is satisfied. If it is determined that it is established, the content of the execution content 509 is executed. If it is determined that it is not established, the content of the execution content 509 is not executed. Hereinafter, each instruction rule information and the operation of the service providing unit status monitoring unit 121 will be described in detail.

  The instruction rule information 511 indicates that the live state 403 of both the DHCP server “dhcpd” and the TFTP server “tftpd” is “alive”, and the processing elapsed time 404 is the limit processing time “6” seconds, “ This is the case for 60 "seconds or less. In this case, since the condition 508 is “none” and the execution content 509 is “normal processing”, the service providing unit status monitoring unit 121 does not perform any particular processing.

  The instruction rule information 521 is “dead” when the live state 403 of the DHCP server “dhcpd” is stopped, and the processing elapsed time 404 of the TFTP server “tftpd” is “0” seconds (waiting for service acceptance). Is the case. Further, the instruction rule information 531 is a case where the live state 403 of the TFTP server “tftpd” is “dead” when stopped, and the processing elapsed time 404 of the DHCP server “dhcpd” is “0” seconds. The instruction rule information 541 corresponds to a case where the survival state 403 of both the DHCP server “dhcpd” and the TFTP server “tftpd” is “dead” during stoppage. In these cases, since the condition 508 is “none” and the execution content 509 is “failover execution”, the service providing unit status monitoring unit 121 issues a failover execution instruction to the cluster processing unit 111.

  The instruction rule information 512 indicates that the survival state 403 of both the DHCP server “dhcpd” and the TFTP server “tftpd” is “alive” during operation, and the processing elapsed time 404 of the DHCP server “dhcpd” is the limit processing time ” This is a case where 6 ”seconds are exceeded and the processing elapsed time 404 of the TFTP server“ tftpd ”is“ 0 ”seconds. In addition, the instruction rule information 514 indicates that the live state 403 of both the DHCP server “dhcpd” and the TFTP server “tftpd” is “alive” during operation, and the processing elapsed time 404 of the TFTP server “tftpd” is the limit process. This is a case where the time exceeds “60” seconds and the processing elapsed time 404 of the DHCP server “dhcpd” is “0” seconds. In these cases, since the condition 508 is “none” and the execution content 509 is “failover execution”, the service providing unit status monitoring unit 121 issues a failover execution instruction to the cluster processing unit 111. According to such instruction rule information, failover is executed when at least one processing time of the service providing unit 112 exceeds the limit processing time. Therefore, even if the responsiveness of the service providing unit 112 is lowered, it is possible to avoid the responsiveness from being lowered.

  The instruction rule information 513 and 515 are obtained by setting the contents of the instruction rule information 512 and 514 to finer conditions.

  The instruction rule information 513 indicates that the live state 403 of both the DHCP server “dhcpd” and the TFTP server “tftpd” is “alive” during operation, and the processing elapsed time 404 of the DHCP server “dhcpd” is the limit processing time ” This is a case where 6 "seconds are exceeded and the processing elapsed time 404 of the TFTP server" tftpd "is longer than" 0 "seconds (service is being provided to the client computer). The condition 508 is “immediately after the completion of tftpd processing or after the limit processing time 60 s of tftpd”, and the execution content 509 is “failover execution”. In this case, the service providing unit status monitoring unit 121 has history information (content 205 is “end”) corresponding to history information (content 205 is “start”) indicating that service provision of the TFTP server “tftpd” has started. Is detected in the service providing unit operation history table 113, or when it is detected that the processing elapsed time 404 exceeds the limit processing time “60” seconds, a failover execution instruction is issued to the cluster. Output to processing unit 111. If this condition 508 is not satisfied, no failover execution instruction is issued.

  The instruction rule information 515 indicates that the live state 403 of both the DHCP server “dhcpd” and the TFTP server “tftpd” is “alive” during operation, and the processing elapsed time 404 of the DHCP server “dhcpd” is “0” seconds. This is a case where the processing elapsed time 404 of the TFTP server “tftpd” exceeds the limit processing time “60” seconds. The condition 508 is “immediately after completion of dhcpd processing or after the limit processing time 6 s of dhcpd”, and the execution content 509 is “failover execution”. In this case, the service providing unit status monitoring unit 121 has history information (content 205 is “end”) corresponding to history information (content 205 is “start”) indicating that the DHCP server “dhcpd” has started providing the service. Is detected in the service providing unit operation history table 113, or when it is detected that the processing elapsed time 404 exceeds the limit processing time "6" seconds, a failover execution instruction is issued to the cluster. Output to processing unit 111. If this condition 508 is not satisfied, no failover execution instruction is issued.

  According to the instruction rule information such as the instruction rule information 513 and 515, when at least one processing time of the service providing units 112 exceeds the limit processing time, the other service providing units 112 that are not in the stopped state are processing. Then, failover is executed after the processing of the service providing unit 112 being processed is completed. Therefore, service provision to the client computer by the service providing unit 112 is not interrupted. If failover is to be executed unconditionally, the condition 508 of the instruction rule information 513 and 515 is set to “none”.

  The instruction rule information 522 and 532 are obtained by setting the contents of the instruction rule information 521 and 531 to more detailed conditions, respectively.

  The instruction rule information 522 indicates that the live state 403 of the DHCP server “dhcpd” is “dead” and the processing elapsed time 404 of the TFTP server “tftpd” is longer than “0” seconds (provides service to the client computer) Middle). Similarly to the instruction rule information 513, the condition 508 is “immediately after completion of tftpd processing or after tftpd limit processing time 60 s”, and the execution content 509 is “failover execution”. In this case, the service providing unit status monitoring unit 121 has history information (content 205 is “end”) corresponding to history information (content 205 is “start”) indicating that service provision of the TFTP server “tftpd” has started. Is detected in the service providing unit operation history table 113, or when it is detected that the processing elapsed time 404 exceeds the limit processing time “60” seconds, a failover execution instruction is issued to the cluster. Output to processing unit 111. If this condition 508 is not satisfied, no failover execution instruction is issued.

  The instruction rule information 532 indicates that the live state 403 of the TFTP server “tftpd” is “dead” and the processing elapsed time 404 of the DHCP server “dhcpd” is longer than “0” seconds (provides service to the client computer) Middle). The condition 508 is “immediately after the completion of dhcpd processing or after the dhcpd limit processing time 6 s” as in the case of the instruction rule information 515, and the execution content 509 is “failover execution”. In this case, the service providing unit status monitoring unit 121 has history information (content 205 is “end”) corresponding to history information (content 205 is “start”) indicating that the DHCP server “dhcpd” has started providing the service. Is detected in the service providing unit operation history table 113, or when it is detected that the processing elapsed time 404 exceeds the limit processing time "6" seconds, a failover execution instruction is issued to the cluster. Output to processing unit 111. If this condition 508 is not satisfied, no failover execution instruction is issued.

  Even if at least one of the service providing units 112 is stopped by the instruction rule information such as the instruction rule information 522 and 532, if another service providing unit 112 that is not in the stopped state is processing, Failover is executed after the processing of the service providing unit 112 is completed. Therefore, service provision to the client computer by the service providing unit 112 is not interrupted. If failover is to be executed unconditionally, the condition 508 of the instruction rule information 522 and 532 is set to “none”.

  By setting the instruction rule as described above, in the network boot system, even when the active server computer, the DHCP server or the TFTP server is down, the situation where the client computer cannot be network booted is avoided. be able to. Moreover, even when the responsiveness of the DHCP server or the TFTP server is lowered, it is possible to avoid a situation in which the client computer cannot be started due to a timeout of the startup time due to the network boot.

  Next, the operation of the service providing unit status monitoring unit 121 configured as described above will be described.

  FIG. 6 is a flowchart showing a processing flow of the service providing unit status monitoring unit 121. The service providing unit status monitoring unit 121 periodically executes the following flow at a cycle specified by the user. For example, the period is preferably about 1 second since the service providing unit for network booting (DHCP server, TFTP server) is a second order process.

  First, the service providing unit status monitoring unit 121 reads the operation history of the service providing unit 112 stored in the service providing unit operation history table 113 (S601).

  Then, the service providing unit status monitoring unit 121 updates the service providing unit state table 123 (S602). Specifically, the operating state of the service providing unit 112 (operating “alive” or stopped “dead”) is inquired of the OS, and the acquired operating state is stored in the life / death state 403. Further, the read operation history is checked in order, and is classified for each name of the host name 203 and for each service providing unit name 204. Then, the content 205 of the operation history for each service providing unit is referred to. When there is an operation history in which an end “end” corresponding to the start “start” is recorded, “0” seconds are stored in the process elapsed time 404. When only the start “start” is recorded, the time obtained from the difference between the date 201 and time 202 of the operation history and the current date is stored in the process elapsed time 404.

  Next, in the service providing unit status monitoring unit 121, the life / death state 403 of each service providing unit 112 in the service providing unit state table 123 updated in S602 matches the life / death state 501 of each instruction rule information in the failover instruction rule table 124. The matching instruction rule information is selected (S603).

  Also, the service providing unit status monitoring unit 121 determines whether the processing elapsed time 404 of each service providing unit 112 in the service providing unit state table 123 updated in S602 matches the processing elapsed time 502 of the instruction rule information selected in S603. Determination is made and matching instruction rule information is selected (S604). Through the processing of S603 and S604, any one of the instruction rule information 511 to 541 is selected.

  Next, the service providing unit status monitoring unit 121 determines whether or not the state of the service providing unit 112 satisfies the condition 508 of the instruction rule information selected in S604 (S605). Specifically, when the condition 508 is “none”, the process proceeds to S606 (“YES” in S605). When the condition 508 is other than “None”, it is determined whether the processing of the designated service providing unit is completed or whether the processing time of the designated service providing unit exceeds the limit processing time. If it is detected immediately after the completion of processing or after the elapse of the limit processing time, the condition 508 is satisfied, and the process proceeds to S606 (“YES” in S605). On the other hand, if the condition 508 is not satisfied, the processing is terminated (“NO” in S605, END).

  In S606, the service providing unit status monitoring unit 121 executes the content of the execution content 509 of the selected instruction rule information (S606). Specifically, when the execution content 509 is “normal processing”, the service providing unit status monitoring unit 121 does nothing. When the execution content 509 is “failover execution”, the service providing unit status monitoring unit 121 instructs the cluster processing unit 111 to execute failover. Upon receiving the instruction, the cluster processing unit 111 takes over resources from the active server computer 102 to the standby server computer 102 and switches between the standby system and the active system.

  The above steps are executed, and the service providing unit status monitoring unit 121 ends the process. Similarly, the above-described flow is periodically executed thereafter.

  The first embodiment has been described above. According to the first embodiment, in the cluster system, when at least one of the server programs on the active server computer has not finished processing within the limit processing time that is a performance requirement, the standby server computer Failover is executed. As a result, even when a heavy load occurs in the server program, failover is performed on the standby server computer, so that not only hardware failure of the server computer but also a decrease in responsiveness of the server program can be dealt with. it can.

  In addition, if at least one of the server programs on the active server computer has not finished processing within the limit processing time, it waits for the other server programs to complete and then performs failover to the standby server computer. To do. Thereby, it is not necessary to interrupt the processing between the server program and the client computer at the time of failover. In this way, a highly reliable and highly responsive cluster system that can realize non-stop service can be provided.

  Next, a modification of the first embodiment will be described. In this modification, a part of the first embodiment has a simpler configuration. Hereinafter, this modification will be described with reference to FIGS. 8, 9, and 10, focusing on differences from the first embodiment. Similar to the first embodiment, a configuration in which the present invention is applied to a network boot system will be described as an example. The system configuration and the functional configuration of the server computer (FIG. 1), the service providing unit operation history table 113 (FIG. 2), the service providing unit characteristic table 122 (FIG. 3), and the hardware configuration of the server computer (FIG. 7) are as follows. Since it is the same as that of the embodiment, the description is omitted.

  Unlike the first embodiment, this modification has a service providing unit state table 123 shown in FIG. 8 and a failover instruction rule table 124 shown in FIG. That is, the processing of the service providing unit status monitoring unit 121 using these tables is different.

  FIG. 8 is a diagram illustrating a configuration of the service providing unit state table 123 according to the modification. The difference from the first embodiment (FIG. 4) is that the life / death state 403 is omitted. The service providing unit status table 123 is periodically updated by the service providing unit status monitoring unit 121 and stores the latest information on the elapsed processing time of the service providing unit 112. The service providing unit state table 123 stores state information in which the service providing unit name 401 and the processing elapsed time 404 are associated with each other. The contents of the processing elapsed time 404 are the same as those in the first embodiment.

  FIG. 9 is a diagram illustrating a configuration of a failover instruction rule table 124 according to a modification. The difference from the first embodiment (FIG. 5) is that the life / death state 501 for each service providing unit 112 is omitted. The failover instruction rule table 124 stores instruction rule information in which an elapsed processing time 502 for each service providing unit 112 and a rule 503 are associated with each other according to user settings. The contents of the processing elapsed time 502 and the rule 503 are the same as those in the first embodiment.

  Using the failover instruction rule table 124 described above, the service providing unit status monitoring unit 121 periodically performs processing according to the instruction rule. First, the service providing unit status monitoring unit 121 acquires the processing elapsed time 404 of each service providing unit 401 stored in the service providing unit state table 123. Thereafter, the instruction rule information that matches the content of the process elapsed time 502 is selected from the acquired process elapsed time and the instruction rule information 911 to 915. Then, it is determined whether or not the condition 508 of the selected instruction rule information is satisfied. If it is determined that it is established, the content of the execution content 509 is executed. If it is determined that it is not established, the content of the execution content 509 is not executed. Hereinafter, each instruction rule information and the operation of the service providing unit status monitoring unit 121 will be described in detail.

  The instruction rule information 911 is when the processing elapsed time 404 of both the DHCP server “dhcpd” and the TFTP server “tftpd” is the limit processing time “6” seconds and “60” seconds or less, respectively. In this case, since the condition 508 is “none” and the execution content 509 is “normal processing”, the service providing unit status monitoring unit 121 does not perform any particular processing.

  The instruction rule information 912 is a case where the processing elapsed time 404 of the DHCP server “dhcpd” exceeds the limit processing time “6” seconds and the processing elapsed time 404 of the TFTP server “tftpd” is “0” seconds. The instruction rule information 914 indicates that the process elapsed time 404 of the TFTP server “tftpd” exceeds the limit process time “60” seconds and the process elapsed time 404 of the DHCP server “dhcpd” is “0” seconds. is there. In these cases, since the condition 508 is “none” and the execution content 509 is “failover execution”, the service providing unit status monitoring unit 121 issues a failover execution instruction to the cluster processing unit 111. According to such instruction rule information, failover is executed when at least one processing time of the service providing unit 112 exceeds the limit processing time. Therefore, even if the responsiveness of the service providing unit 112 is lowered, it is possible to avoid the responsiveness from being lowered.

  The instruction rule information 913 indicates that the processing elapsed time 404 of the DHCP server “dhcpd” exceeds the limit processing time “6” seconds and the processing elapsed time 404 of the TFTP server “tftpd” is greater than “0” seconds (in the client computer). Service). The condition 508 is “immediately after the completion of tftpd processing or after the limit processing time 60 s of tftpd”, and the execution content 509 is “failover execution”. In this case, the service providing unit status monitoring unit 121 has history information (content 205 is “end”) corresponding to history information (content 205 is “start”) indicating that service provision of the TFTP server “tftpd” has started. Is detected in the service providing unit operation history table 113, or when it is detected that the processing elapsed time 404 exceeds the limit processing time “60” seconds, a failover execution instruction is issued to the cluster. Output to processing unit 111. If this condition 508 is not satisfied, no failover execution instruction is issued.

  The instruction rule information 915 is a case where the processing elapsed time 404 of the DHCP server “dhcpd” is greater than “0” seconds and the processing elapsed time 404 of the TFTP server “tftpd” exceeds the limit processing time “60” seconds. The condition 508 is “immediately after completion of dhcpd processing or after the limit processing time 6 s of dhcpd”, and the execution content 509 is “failover execution”. In this case, the service providing unit status monitoring unit 121 has history information (content 205 is “end”) corresponding to history information (content 205 is “start”) indicating that the DHCP server “dhcpd” has started providing the service. Is detected in the service providing unit operation history table 113, or when it is detected that the processing elapsed time 404 exceeds the limit processing time "6" seconds, a failover execution instruction is issued to the cluster. Output to processing unit 111. If this condition 508 is not satisfied, no failover execution instruction is issued.

  According to the instruction rule information such as the instruction rule information 913 and 915, when at least one processing time of the service providing units 112 exceeds the limit processing time, another service providing unit 112 that is not in a stopped state is processing. Then, failover is executed after the processing of the service providing unit 112 being processed is completed. Therefore, service provision to the client computer by the service providing unit 112 is not interrupted.

  By setting the instruction rule as described above, even if the response of the DHCP server or the TFTP server is lowered in the network boot system, the client computer cannot be started due to a timeout of the boot time due to the network boot. The situation can be avoided.

  Next, the operation of the service providing unit status monitoring unit 121 configured as described above will be described.

  FIG. 10 is a flowchart showing a process flow of the service providing unit status monitoring unit 121 according to the modification. The difference from the first embodiment (FIG. 6) is that the process is omitted from S603. In addition, processing (S602 ′, S604 ′) corresponding to the service providing unit state table 123 (FIG. 8) and the failover instruction rule table 124 (FIG. 9) is executed. Since S601, S605, and S606 are the same as those in the first embodiment, description thereof is omitted.

  After S601, the service providing unit status monitoring unit 121 updates the service providing unit state table 123 (S602 ′). Specifically, the read operation history is checked in order, and is classified into each name of the host name 203 and each service providing unit name 204. Then, the content 205 of the operation history for each service providing unit is referred to. When there is an operation history in which an end “end” corresponding to the start “start” is recorded, “0” seconds are stored in the process elapsed time 404. When only the start “start” is recorded, the time obtained from the difference between the date 201 and time 202 of the operation history and the current date is stored in the process elapsed time 404.

  Next, the service providing unit status monitoring unit 121 determines that the processing elapsed time 404 of each service providing unit 112 in the service providing unit state table 123 updated in S602 ′ is the processing elapsed time of each instruction rule information in the failover instruction rule table 124. It is determined whether it matches 502, and the matching instruction rule information is selected (S604 ′). Thereafter, the service providing unit status monitoring unit 121 executes the processes of S605 and S606.

  In the above, the modification of 1st Embodiment was demonstrated. According to this modification, in the cluster system, when at least one of the server programs on the active server computer has not finished processing within the limit processing time required for performance, the failover is performed on the standby server computer. Is executed. As a result, when a high load state occurs in the server program, a failover is executed on the standby server computer, so it is possible to cope with a decrease in the responsiveness of the server program.

  In addition, if at least one of the server programs on the active server computer has not finished processing within the limit processing time, it waits for the other server programs to complete and then performs failover to the standby server computer. To do. Thereby, it is not necessary to interrupt the processing between the server program and the client computer at the time of failover. In this way, a highly reliable and highly responsive cluster system that can realize non-stop service can be provided.

  The present invention has been described in connection with exemplary embodiments. Obviously, many alternatives, modifications, and variations will be apparent to practitioners skilled in this art. Accordingly, the above-described embodiments of the present invention are intended to illustrate and not limit the gist and scope of the present invention.

The block diagram which shows the system configuration | structure which concerns on 1st Embodiment, and the function structure of a server computer. The figure for demonstrating the structure of the service provision part operation | movement log | history management table which concerns on 1st Embodiment. The figure for demonstrating the structure of the service provision part characteristic table which concerns on 1st Embodiment. The figure for demonstrating the structure of the service provision part state table which concerns on 1st Embodiment. The figure for demonstrating the structure of the failover instruction | indication rule table which concerns on 1st Embodiment. The flowchart which shows the flow of a process of the service provision part condition monitoring part which concerns on 1st Embodiment. The block diagram which shows the hardware constitutions of the server computer which concerns on 1st Embodiment. The figure for demonstrating the structure of the service provision part state table which concerns on the modification of 1st Embodiment. The figure for demonstrating the structure of the failover instruction | indication rule table which concerns on the modification of 1st Embodiment. The flowchart which shows the flow of a process of the service provision part condition monitoring part which concerns on the modification of 1st Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 ... Cluster system, 102 ... Server computer, 104 ... Client computer, 105 ... Client computer, 106 ... Client computer, 107 ... LAN, 108 ... LAN, 111 ... Cluster processing unit, 112 ... service providing unit, 113 ... service providing unit operation history table, 122 ... service providing unit characteristic table, 123 ... service providing unit state table, 124 ... failover instruction Rule table, 150 ... control unit, 152 ... storage unit,
201 ... Date, 202 ... Time, 203 ... Host name, 204 ... Service provider name, 205 ... Contents, 210-240 ... Operation history,
301 ... Service providing unit name, 303 ... Average processing time, 304 ... Limit processing time, 310-320 ... Characteristic information,
401 ... Service providing unit name, 403 ... Life / death state, 404 ... Process elapsed time, 410-420 ... Status information,
501 ... Life / death state, 502 ... Process elapsed time, 503 ... Rule, 504 to 505 ... Life / death state for each service providing unit, 506 to 507 ... Process elapsed time for each service providing unit, 508 ... condition, 509 ... execution content, 511-541 ... instruction rule information,
DESCRIPTION OF SYMBOLS 11 ... CPU, 12 ... Main memory, 13 ... Auxiliary memory, 14 ... Input device, 15 ... Output device, 16 ... Reading device, 17 ... Communication device.

Claims (15)

A server constituting a cluster system having a plurality of servers each including a failover processing unit that performs a failover process of taking over the provision of the service from one server that provides the service to another server by receiving a failover execution instruction Because
A storage unit for storing state information for specifying a state of a processing state for performing a failover;
A status monitoring unit that identifies the processing status when providing the service;
A failover execution instruction unit that instructs the failover processing unit to execute failover when the processing status specified by the status monitoring unit matches the processing status specified by the status information;
The server characterized by having. The server according to claim 1,
In the storage unit, condition information for specifying a condition for performing failover is stored corresponding to each state information,
The processing status specified by the status monitoring unit matches the processing status specified by the status information, and the processing status specified by the status monitoring unit satisfies the condition information corresponding to the status information determined to match. The failover execution instruction unit instructs the failover processing unit to execute failover,
A server characterized by The server according to claim 1,
The processing status is a processing time of a predetermined process,
The state information is information for specifying that a processing time of a predetermined process has exceeded a predetermined limit time;
A server characterized by The server according to claim 2,
The processing status is a processing time of a predetermined process,
The state information is information for specifying that the processing time of a predetermined process has exceeded a predetermined limit time,
The condition information indicates that processing other than the predetermined processing has been completed and that the processing time of processing other than the predetermined processing has exceeded a predetermined limit time. Information identifying at least one of
A server characterized by The server according to claim 2,
The processing status is an operation state indicating operation or stop of a predetermined process,
The state information is information for specifying that an operation state of a predetermined process is stopped,
The condition information indicates that processing other than the predetermined processing has been completed and that the processing time of processing other than the predetermined processing has exceeded a predetermined limit time. Information identifying at least one of
A server characterized by A server constituting a cluster system having a plurality of servers each including a failover processing unit that performs a failover process of taking over the provision of the service from one server that provides the service to another server by receiving a failover execution instruction The failover execution method in
The server
Stores status information that identifies the status of the processing status for failover,
A status monitoring step for identifying the processing status when providing the service;
A failover execution instruction step for instructing the failover processing unit to execute a failover when the processing status specified in the status monitoring step matches the processing status specified in the status information;
A failover execution method characterized in that is executed. The failover execution method according to claim 6, wherein
The server
The condition information for specifying the condition for performing the failover for each state information is stored in association with each other,
The failover execution instruction step includes:
The processing status specified by the status monitoring unit matches the processing status specified by the status information, and the processing status specified by the status monitoring unit satisfies the condition information corresponding to the status information determined to match. Instructing the failover processing unit to execute failover,
Failover execution method characterized by this. The failover execution method according to claim 6, wherein
The processing status is a processing time of a predetermined process,
The state information is information for specifying that a processing time of a predetermined process has exceeded a predetermined limit time;
Failover execution method characterized by this. The failover execution method according to claim 7, comprising:
The processing status is a processing time of a predetermined process,
The state information is information for specifying that the processing time of a predetermined process has exceeded a predetermined limit time,
The condition information indicates that processing other than the predetermined processing has been completed and that the processing time of processing other than the predetermined processing has exceeded a predetermined limit time. Information identifying at least one of
Failover execution method characterized by this. The failover execution method according to claim 7, comprising:
The processing status is an operation state indicating operation or stop of a predetermined process,
The state information is information for specifying that an operation state of a predetermined process is stopped,
The condition information indicates that processing other than the predetermined processing has been completed and that the processing time of processing other than the predetermined processing has exceeded a predetermined limit time. Information identifying at least one of
Failover execution method characterized by this. A cluster system having a plurality of servers each provided with a failover processing unit that performs failover processing to take over the provision of the service from one server that provides the service to another server by receiving a failover execution instruction for the computer A program for functioning as a server to be configured,
The computer,
A storage unit for storing state information for specifying a state of a processing state for performing a failover;
A status monitoring unit that identifies the processing status when providing the service;
When the processing status specified by the status monitoring unit matches the processing status specified by the status information, as a failover execution instruction unit that instructs the failover processing unit to execute failover,
A program characterized by functioning. The program according to claim 11,
In the storage unit, condition information for specifying a condition for performing failover is stored corresponding to each state information,
The processing status specified by the status monitoring unit matches the processing status specified by the status information, and the processing status specified by the status monitoring unit satisfies the condition information corresponding to the status information determined to match. The failover execution instruction unit instructs the failover processing unit to execute failover,
A program characterized by The program according to claim 11,
The processing status is a processing time of a predetermined process,
The state information is information for specifying that a processing time of a predetermined process has exceeded a predetermined limit time;
A program characterized by A program according to claim 12,
The processing status is a processing time of a predetermined process,
The state information is information for specifying that the processing time of a predetermined process has exceeded a predetermined limit time,
The condition information indicates that processing other than the predetermined processing has been completed and that the processing time of processing other than the predetermined processing has exceeded a predetermined limit time. Information identifying at least one of
A program characterized by A program according to claim 12,
The processing status is an operation state indicating operation or stop of a predetermined process,
The state information is information for specifying that an operation state of a predetermined process is stopped,
The condition information indicates that processing other than the predetermined processing has been completed and that the processing time of processing other than the predetermined processing has exceeded a predetermined limit time. Information identifying at least one of
A program characterized by

Images