WO2013084305A1

WO2013084305A1 - Virtualization multi-system configuration control method and computer system

Info

Publication number: WO2013084305A1
Application number: PCT/JP2011/078218
Authority: WO
Inventors: 成昊金; 貴広池田; 崇博大平
Original assignee: 株式会社日立製作所
Priority date: 2011-12-06
Filing date: 2011-12-06
Publication date: 2013-06-13

Abstract

In order to prevent the loss of control information when a virtual computer system is switched in the multi-system configuration control of a virtual computer, multi-system control units, which set a master/slave relationship between a pair of virtual computers belonging to different physical computers, and which are capable of changing the master/slave relationship, are installed in the operating systems of multiple physical computers. One multi-system control unit transmits packets to the other multi-system control unit and monitors the health status of the virtual computer in the physical computer that is the counterpart in terms of the master/slave relationship, and, in accordance with the monitoring result, updates management information indicating the health status of the virtual computers in the master/slave relationship. In addition, in accordance with the monitoring result, this one multi-system control unit transmits packets to other multi-system control units so as to designate a change in the master/slave relationship between the virtual computers. Furthermore, one multi-system control unit transmits input information received from an external device to the virtual computers that are in a master/slave relationship, and obtains and manages output information, which is the result of processing by the virtual computers in the master/slave relationship, and transmits to the external device the output information from the master system or the slave system virtual computer.

Description

Virtualization multisystem configuration control method and computer system

The present invention relates to a virtualized multisystem configuration control method and a computer system, and in particular, an output from a virtual machine in a virtual multisystem configuration computer system that performs multisystem configuration control between virtual machines constructed on a physical computer. The present invention relates to processing and system switching control.

Server virtualization technology has the advantage of reducing operating costs, computer installation location, power consumption, etc. by executing multiple virtual machines on one physical computer and reducing the number of physical computers. Has been. Recently, in order to ensure the same reliability as the system operated by the physical computer, the multiplexing system configuration of the virtual computer that is controlled and operated by each physical computer as well as the multiplexing of the physical computer that provides the virtual environment Is also required.

As this type of technology, Patent Document 1 discloses a cluster configuration having a system switching method using hot standby and cold standby in a server virtualization environment in which a large number of guest OSs exist. A cluster system is disclosed that realizes a suitable cluster configuration by selecting a system switching method that satisfies the high availability requirement for specifying the performance of the system and switching the system switching method.

JP 2008-269332 A

The multi-system configuration control in the virtual environment described in Patent Document 1 has a master-slave relationship between virtual machines, and the virtual machine determines whether to output control information based on its own state. For this reason, when a fault occurs in a certain virtual machine and the master / slave system is switched, it takes time for the virtual machine that was the slave to take over the calculation process, and the output of control information during that time is lost. is there.

An object of the present invention is to provide a virtualized multisystem configuration control method and a computer system that prevent loss of control information when performing system switching of virtual machines in the multisystem configuration control of a virtual machine.

In the virtual multiplex system configuration control method according to the present invention, preferably, a host OS and a plurality of physical computers having one or a plurality of virtual computers constructed on the host OS are connected via an interface, and the virtual computer Is a virtual multiplex system configuration control method in a computer system that can input and output from a commonly connected external device,
On the OS of the plurality of physical computers, a multi-system controller that can set a master-slave relationship between paired virtual computers belonging to different physical computers and change the master-slave relationship is installed.
The multi-system control unit of one transfers the packet to and from the other multi-system control unit, monitors the life / death state of the virtual machine in the physical computer of the partner in the master-slave relationship, and The management information indicating the life / death status of the virtual machines in the master-slave relationship is updated according to the result of the above, and the packet is transmitted to the other multi-system control unit according to the monitoring result. Instructing the master-detail relationship to change,
The multi-system control unit 1 transfers the input information received from the external device to the virtual machine having a master-slave relationship, and obtains output information that is a result processed by the virtual computer having a master-slave relationship And a virtual multiplex system configuration control method for transferring output information from the primary or secondary virtual machine to the external device.

In the computer system according to the present invention, preferably, a host OS and a plurality of physical computers having one or more virtual machines constructed on the host OS are connected via an interface and belong to different physical computers. A master-slave relationship is set for a virtual machine, and the virtual machine is a computer system that can input and output from a commonly connected external device,
Provided on the host OS of the plurality of physical computers is a multiplex system control unit capable of setting a master-slave relationship between paired virtual computers belonging to different physical computers and changing the master-slave relationship. The system control unit can transmit a packet to communicate management information and control information with each other, and the multiplex system control unit
First management means for managing the master-slave relationship of the virtual machine and information on the life and death of the virtual machine in correspondence with the virtual machine;
Input information received from the external device, second management means for managing output information output to the external device as a result of processing in the virtual machine having a master-slave relationship corresponding to the input information;
The packet is transferred to and from the other multi-system control unit to monitor the life / death status of the virtual machine in the physical computer of the partner in the master-slave relationship, and the master-slave relationship in accordance with the monitoring result Update the contents of the first management means indicating the life or death status of the virtual machines, and change the master-slave relationship between the virtual machines by sending packets to the other multi-system controller according to the monitoring results. First control means for controlling so as to
The input information received from the external device is transferred to the virtual machine in a master-slave relationship, and output information that is a result processed by the virtual computer in a master-slave relationship is acquired and stored in the second management unit And a second control unit that controls the output information from the primary or secondary virtual machine to be transferred to the external device.

According to a preferred example, the plurality of multi-system controllers are a VM state management table that manages master / slave relationships of the virtual machines in a paired relationship, and information on the life / death state of the virtual machines,
Input information received from the external device, a control information management table for managing output information that is a result of processing in the virtual machine having a master-slave relationship corresponding to the input information, corresponding to the virtual machine,
The multi-system control unit stores the output information obtained from the virtual machine that has transferred the input information in the control information management table, and has a master-slave relationship stored in the VM state management table. With reference to the life / death information of the virtual machine, it is decided to transfer the output information stored in the control information management table to the external device.

Preferably, when a failure occurs in the virtual machine, the multiplex system control unit refers to the control information management table and controls output information necessary for the next output processing from a processing result at the time of the failure. .
Preferably, when a failure occurs in the primary virtual machine, the multi-system control unit that manages the primary virtual machine manages the secondary virtual machine. The packet is transferred to the other multi-system control unit to notify the failure, and one multi-system control unit and the other multi-system control unit switch the master-slave relationship of the pair of virtual machines.

Preferably, the multiplex system control unit of one updates the state of the virtual machine in which the failure has occurred, and transmits the status information of the virtual machine in which the failure has occurred and the final output information at the time of the failure to the other multiplex units. Sent to the system controller
The other multiplex system control unit updates the status information of the virtual machine in the VM status management table, and refers to the control information management table to input information necessary for the next output processing of the final output information, Transfer to the virtual machine in the master-slave relationship with the virtual machine in which the failure has occurred, and further switch the master-slave relationship.

Also preferably, the other multi-system control unit outputs the output information stored in the control information management table, which is output information from the virtual machine, whose master-slave relationship has been switched, to the external device.

Preferably, when the multi-system control unit receives output information from a virtual machine within a predetermined cycle time, it refers to the VM state management table and outputs the master-slave of the virtual machine. Determine the state,
As a result of the determination, in the case of the main system, the output information obtained from the virtual machine is stored in the control information management table, and then the output information stored in the control information management table is transferred to the external device. And generating a packet for notifying the output information to the other multiplex system control unit and transferring the packet to the other multiplex system control unit,
As a result of the determination, in the case of a secondary system, check whether the output information transferred from the virtual machine of the secondary system is stored in the control information management table,
As a result of the confirmation, the stored output information is compared with the output information obtained from the virtual computer of the main system. When the two do not match, consistency between the virtual computers in the master-slave relationship is obtained. Execute fail-safe processing, assuming that it is not.

Preferably, input information from the external device is controlled to be input to the multiple system control unit of the physical computer in which the virtual computer of the main system exists, and the master system and the slave system The multiplex system control unit in the physical computer in which the virtual computer exists is configured to transmit and receive the acquired input information and a synchronization packet transferred between the multiplex system control unit including the master and slave virtual computers. Having an input information synchronization management table for managing information indicating a state corresponding to the virtual machine in a master-slave relationship;
The multiplex system control unit of 1 confirms whether the virtual computer that is the transfer source of the packet is a primary system or a slave system based on the synchronization packet transferred and received from the other multiplex system control unit,
As a result of the confirmation, in the case of the main system, waiting for the reception of the synchronization packet transferred from the other multi-system control unit,
As a result of the confirmation, in the case of the slave, the acquired input information is stored in the input information synchronization management table, and the information in the synchronous transmission state is stored in the input information synchronization management table, indicating that the reception is completed. The synchronization packet is transferred to the other multiplex system control unit as the transmission source.

According to the present invention, system switching between virtual machines is not performed by a virtual machine as in the prior art, but by a multi-system control unit (for example, a multi-system output control infrastructure program) provided on a host OS of a physical computer. Since the control information is output and the system switching process is performed from the virtual machine, the overhead of the system switching process can be reduced. In addition, it is possible to control the output of control information while preventing loss of control information when a failure occurs.
In addition, since the multi-system output control infrastructure program is aware of the multi-system configuration form and manages whether or not the control information of the virtual machine is output, the virtual machine can perform multi-configuration control without being aware of the multi-system configuration form.

1 is an overall configuration diagram of a computer system for virtual multiplex system control in one embodiment. FIG. FIG. 3 is a diagram illustrating a configuration of a VM state management table in the first embodiment. FIG. 3 is a diagram illustrating a configuration of a control information management table in the first embodiment. FIG. 3 is a diagram showing a format configuration of a packet for communication between computer nodes in the first embodiment. FIG. 3 is a diagram illustrating an outline of a virtual communication I / F according to the first embodiment. 1 is a flowchart of overall processing of a virtualized multiplex system according to Embodiment 1. FIG. FIG. 6 is a process flowchart of the input information management program according to the first embodiment. FIG. 6 is a process flowchart of an output information management program according to the first embodiment. FIG. 6 is a process flowchart of a subordinate information management program according to the first embodiment. FIG. 6 is a flowchart of control information verification / addition processing S904 in the first embodiment. The flowchart figure of the master-slave switching process S905 in Example 1. FIG. FIG. 6 is a process flowchart of the VM-HOST life and death monitoring program according to the first embodiment. The processing flowchart figure of the HOST-HOST life-and-death monitoring program in Example 1. FIG. FIG. 3 is a VM state update communication format diagram according to the first embodiment. FIG. 10 is a process flowchart of a VM state update program according to the first embodiment. FIG. 6 is an overall configuration diagram of a virtual environment multiple control computer system according to a second embodiment. FIG. 10 is an overall configuration diagram of an input information synchronization management table in Embodiment 2. The block diagram of the communication format between computer nodes for input synchronization in Example 2. FIG. FIG. 10 is a process flowchart of an input information management program according to the second embodiment. FIG. 10 is a process flowchart diagram of an input synchronization management program according to the second embodiment. The figure which shows the schematic comparison of the virtual multiplex system structure control of this invention and a prior art.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 shows an overall configuration diagram of a computer system for virtual multiplex system configuration control according to an embodiment of the present invention. This computer system is configured by connecting a terminal 9 to the virtualization multiplex system 1 via an external network 802. The virtual multiplex system 1 includes a plurality of computer nodes 100 (computer nodes 100-A and 100-B in the illustrated example) and an internal network 801 that connects these computer nodes. The computer node 100-A and the computer node 100-B (which are collectively expressed as 100) have the same configuration. Detailed illustration of the configuration contents of the computer node 100-B is omitted.

The computer node 100 includes a memory 101, a processor 102, a storage device 103, an external communication interface (I / F) 104, and an inter-node communication I / F 105 as physical computers. The memory 101 includes a host operating system (OS) 400, a multi-system output control program 300, and one or a plurality of virtual machines (VM) 200-A to 200-C constructed on the host OS 400 (collectively, 200). In the host OS 400, a virtualization program 401 and a device control program 402 are held.

Each VM 200 of the computer node 100-A has a virtual communication I / F 213 that communicates with the application (AP) 211, the guest OS 212, and the multi-system output control program 300. Each VM 201 of the computer node 100-B has the same configuration.
The VM 200 is a virtual execution environment created by the virtualization program 401 in which a unique program or operating system can operate. When there are a plurality of VMs 200, programs and operating systems that operate in different VMs 200 may be the same or different. In addition, programs and operating systems operating in different VMs 200 do not directly affect each other. Each VM 200 has a unique identifier within the virtualized multiplexing system 1. For example, the VM 200-A in FIG. 1 has an identifier “VM1”, and the VM 200-B has an identifier “VM2”.
The AP 211 processes the request from the terminal 9. The guest OS 212 controls the AP 211 within the VM 200 environment and transmits the processing result of the AP 211 to the virtual communication IF 213. The virtual communication IF 213 receives the processing result of the AP 211 from the guest OS 212 and outputs it to the multi-system output control infrastructure program 300.

In this embodiment, the VM 200 in the computer node 100-A and the VM 201 in the computer node 100-B constitute a multiplex system. For example, the VM 200-A and the VM 201-A form a multiplex configuration, and the VM 200-B and the VM 201-B form a multiplex configuration. In this case, which VM is the primary system and which is the secondary system depends on the management of the VM state management table. This will be described later with reference to FIG.

The terminal 9 and the computer node 100 are connected via an external network 802. Further, the computer node 100-A and the computer node 100-B in the virtual multisystem 1 are connected via an internal network 801. In the computer node 100, the external network 802 is connected to the external communication I / F 104, and the internal network 801 is connected to the inter-node communication I / F 105.

The terminal 9 transmits a request to the virtual multiplex system 1 via the external network 802, and the virtual multiplex system 1 returns a result of processing the request to the terminal 9. The request transmitted from the terminal 9 is received by the computer node 100-A and the computer node 100-B, and the same processing is performed in the VMs constituting the multiplex system in both computer nodes. As a result of the processing, correct processing data is returned to the terminal 9 from the VM 200 (eg, VM 200-A) of one computer node (eg, the primary node 100-A).

The storage device 103 of the computer node 100 stores a host OS 400, a multiplex system output control program (including programs and various table information) 300, an AP 211 that configures the VM 200, a guest OS 212, and a virtual communication I / F 213. The processor 102 expands and executes the configuration of the host OS 400, the multiplex system output control program 300, and the VM 200 from the storage device 103 to the memory 101, and further processes an interrupt from the external communication I / F 104 or the inter-node communication I / F 105. . The external communication I / F 104 transmits / receives data to / from the terminal 9 through the external network 802. The inter-node communication I / F 105 transmits / receives data to / from another computer node 100 through the internal network 801.

The virtualization program 401 is a program for creating and managing the VM 200, and the device control program 402 is a program for accessing devices such as the external communication I / F 104, the inter-node communication I / F 105, and the storage device 103. The host OS 400 uses the virtualization program 401 and the device control program 402 to control the execution of the multi-system output program 300 and the VM 200, and controls access to the external communication I / F 104, the inter-node communication I / F 105, and the storage device 103. Do.

The multi-system output control infrastructure program 300 exchanges control information with the multi-system output control infrastructure program 300 ′ operating in the other computer node 100 -B, so that the VM 200 and the VM 201 in the virtual multi-system 1 are exchanged. The master-slave relationship between the control information and the control information is controlled. In order to realize this function, the multi-system output control infrastructure program 300 includes a VM state management table 301, a control information management table 302, an input information management program 303, an output information management program 304, a subordinate information management program 305, a HOST- It consists of a HOST life / death monitoring program 306, a VM-HOST life / death monitoring program 307, and a VM status update program 308. The execution processes of the management tables 301 and 302 and the various programs 303 to 308 included in the multiplex system output control infrastructure program 300 will be described later with reference to FIGS.

The multi-system output control infrastructure program 300 operating in the computer node 100-A having the primary virtual computer determines the state of the VM 200 and updates the contents of the VM state management table 301 and the control information management table 302. In addition, when a failure is detected by the VM-HOST life / death monitoring and the HOST-HOST life / death monitoring, an instruction to switch the master / slave relationship of the VM is sent to the multi-system output control infrastructure program 300 ′ operating in the other computer node 100-B. This is performed to switch the VM master-slave relationship and update the VM life / death state. The multi-system output control base program 300 ′ operating in the computer node 100-B having the subordinate virtual computer updates the control information management table 302 ′, and switches between the main / slave from the multi-system output control base program 300 ′. And the VM state management table 301 ′ is rewritten.

Here, in order to facilitate understanding of the present invention, a difference between the virtual multiplex system configuration control of the present invention and the prior art will be described with reference to FIG.
(A) shows virtual multiplex system configuration control according to the prior art. In the prior art, management of a master-slave relationship of multiple systems and monitoring of VM life and death are performed on VM1 and VM3. Output control of information to the terminal is performed by, for example, middleware on the VM. In the illustrated example, the VM1 side on the host OS 400 side is the primary system, and the VM3 side on the host OS 400 ′ side is the secondary system. The VM3 on the secondary host OS 400 ′ side periodically sends a health check request (REQ) to the primary VM1 and receives a health check response (ACK) from the VM1 to monitor life and death. Yes. If a failure occurs in the primary VM 1 at time T0, the VM 1 processing is stopped, while the VM 3 switches itself from the secondary to the primary. Similarly, VM1 switches itself from the main system to the subordinate system.
According to this prior art, VM1 and VM3 perform VM life / death monitoring and output control, and therefore T11, T12, T21, T22, and T2 are required. Overhead. For this reason, the time for outputting the result (output information) processed by the main VM 1 to the terminal is delayed, and the output information may be lost when the system is switched.

(B) shows virtual multiplex system configuration control according to the present invention. In the present invention, multi-system master / slave management is performed under the control of the multi-system output control infrastructure program of the

host OS

400, 400 ′, and output control to the terminal is performed under the control of the host OS. The

host OS

400, 400 ′ multi-system output control infrastructure program continuously monitors the life and death of the VM 1 and VM 3 formed in the host OS 400, and further exchanges health checks between the host OSs. In this case, if a failure occurs in the primary VM 1 at time T 0, the multi-system output control infrastructure program of the host OS 400 stops the processing of the VM 1 and switches itself from the primary system to the secondary system. On the other hand, the multiplex system output control infrastructure program of the slave host OS 400 ′ switches itself from the slave system to the master system. According to this, since all the processing is performed between the host OSs, the overhead can be reduced, and the VM does not need to be involved in the multiple configuration control. Therefore, the host OS can easily form a multiplexed system configuration of VMs. .
In order to perform this virtual multiplex system configuration control, various programs and tables in the multiplex system output control infrastructure program 300 function effectively. Details will be described below.

FIG. 2 shows the configuration of the VM state management table 301.
The VM state management table 301 manages state information regarding life / death states and master-slave relationships of all VMs created in the computer nodes 100-A and 100-B. A similar VM state management table is also held in the multi-system output control infrastructure program 300 '(not shown).
The VM state management table holds different state information for each computer node. In other words, the VM state management table 301 of the computer node 100-A holds the state information of a plurality of VMs such as the VMs 100-A and 100-B. On the other hand, the VM state management table of the computer node 100-B holds state information of the VMs 201-A, 201-B and the like.

In the VM state management table 301, the VM identifier 311 stores information for uniquely identifying the VM 200 within the computer node 100-A. The host computer node identifier 312 stores identification information of the computer node 100-A in which the VM 200 is created. The master / slave state flag 313 stores information for identifying the master / slave of the VM 200. When the target VM 200 is the primary VM, “M” is stored, and when the target VM 200 is the secondary VM, “S” is stored.
The paired VM identifier 314 stores an identifier that uniquely identifies the VM 201 of the partner (ie, in the computer node 100-B) in the VM master-slave relationship of the VM identifier 311. The paired host computer node identifier 315 stores the identification information of the computer node 100-B in which the paired VM identifier 314 is created. The paired VM life / death state flag 316 stores information on the life / death state of the paired VM. “1” indicates life and “0” indicates death.

FIG. 3 shows the configuration of the control information management table 302.
The control information management table 302 manages control information input to and output from the VM, and is prepared for each VM 200-A, 200-B, and the like. For example, when the computer node 100-A has two VMs 200, two control information management tables 302 are prepared. Similarly, this management table is prepared for each VM in the computer node 100-B.

In the control information management table 302, the VM identifier 321 stores information for identifying the VM in the computer node 100. The sequence number 322 stores information for identifying control information. The input information 323 stores input information received from the terminal 9. The period time 324 stores the time required for the VM periodic task. The output information 325 stores control information output from the VM 200. The output time 326 stores the time (date and time) when the VM 200 outputs the control information.

This control information management table can be used to prevent loss of control information output from each VM and to correct a deviation in calculation results (output control information) between VMs in a master-slave relationship. That is, VM1 and VM3 are in a master-slave relationship, and the same input information is input to the control information management table for managing both, so basically the control information managed by both is the same. However, the cycle time may be disturbed due to a failure or an interrupt, and the output information corresponding to the input information may not be output within the predicted time. In this case, by monitoring the control information management tables of each other and checking whether or not output information is output and its output time, it is confirmed that control information output from each VM is missing, and there is a master-slave relationship. Differences in calculation results between VMs can be confirmed.

FIG. 4 shows a format structure of a packet for communication between computer nodes.
The format of the packet 40 for communication between computer nodes is a format of communication data transmitted / received between the multiple system output

control infrastructure programs

300 and 300 ′ of the different computer nodes 100-A and 100-B. Is used for the transmission of the master and the switching of the master-slave.
The computer node identifier 401 stores an identifier that identifies the destination computer node 100, and the VM identifier 402 stores an identifier that identifies the destination VM. The sequence number 403 and the output information 404 respectively store the sequence number 321 of the control information management table 302 and the output information in the latest row.
The output information 404 may be empty. This is because if the output information 404 is present, the transmission is a communication for transmission of the main system output information, and if there is no output information 404, it is determined that the communication is a communication for switching between the master and the slave.

FIG. 14 shows a format structure of a packet for VM state update communication.
The VM state update communication packet 70 is communication data transmitted / received between the multiplex system output control infrastructure programs of different computer nodes 100-A and 100-B, and updates the paired VM life / death state flag in the VM state management table 301. Used to do.
The computer node identifier 701 stores an identifier that identifies the destination computer node 100, and the VM identifier 702 stores an identifier that identifies the destination VM. The life / death state 703 stores the state of the VM to be updated. When the VM is alive, “1” is stored, and when the VM is dead, “0” is stored.

FIG. 5 shows an outline of processing of the virtual communication I / F 213.
The virtual communication I / F 213 is an interface that connects between the multiplex output control infrastructure program 300 and the VM. The input information 322 stored in the control information management table 302 is received. The received input information is processed by the application 211 on the VM 200, and the control information 500 is generated as a result of the processing. The virtual communication I / F 213 receives the control information 500 and outputs it to the multi-system output control infrastructure program 300. The control information 500 is stored in the output information 325 of the control information management table 302. Note that the virtual notification interface of other VMs performs the same operation.

FIG. 6 shows a flowchart of the overall processing of the virtual multiplex system 1.
The overall processing of the virtualized multisystem 1 includes the start processing of the multisystem output control infrastructure program 300, the input information management program 303, the output information management program 304, the subordinate information management program 305, the VM-HOST life / death monitoring program 306, This is performed by executing each program of the HOST-HOST life / death monitoring program 307.
The host OS 400 and the processor 102 load the multi-system output control infrastructure program 300 and the VM 200 initially stored in the storage device 103, expand them in the memory 101 of the computer node 100, and process the multi-system control infrastructure program 300. To start.

When the processing is started, the multi-system output control infrastructure program 300 reads the VM state management table 301 and the control information management table 302 stored in the storage device 103 into the memory 101 (S601), and activates each VM 200 ( S602). The multiplex system output control infrastructure program 300 activates each VM 200 and then activates each program 303 to 307 (S603).
The processing operation of each program will be described below.

FIG. 7 shows a processing flowchart of the input information management program 303.
When the processing is started, the multiplex system output control infrastructure program 300 starts up each VM 200 and starts up the input information management program 303. When starting the processing, the input information management program 303 stands by until input information is received from the terminal 9 (S701). When the input information is acquired (S702), a new sequence number is added to the sequence number 322 of the control information management table 302 (S703), and the input information output from the terminal 9 is added to the added row of the sequence number. It is stored in the input information 323 (S704).

FIG. 8 is a processing flowchart of the output information management program 304.
When the processing is started, the multiplex system output control infrastructure program 300 starts up each VM 200 and starts up the output information management program 304. When starting the process, the output information management program 304 acquires the period processing time of each VM and stores it in the period time 324 of the control information management table 302 (S800). However, when the cycle time is stored in the input information from the terminal 9, the cycle time can be determined from an external input by overwriting the acquired cycle time of the cycle time 324.

The control information management table 302 is searched, and the input information 323 is selected in sequence order for the input information 323 that exists but the output information 325 does not exist, and is transferred to each VM 200 by multicast (S801). After the input information is transferred, the next input information is similarly transferred to the VM 200 after waiting for the cycle time 324. As a result, the host OS can synchronize the VMs. That is, the difference in output information between the master and slave systems can be suppressed to one at maximum.

Thereafter, when output information from the VM 200 within the cycle time is received (the result of S802 is “Yes”), the master / slave state of the output VM 200 is determined with reference to the master / slave flag 313 of the VM state management table 301. (S803).
When the result of S803 is the primary system, the output information received from the VM is written to the output information 325 of the control information management table 302 (S804), and the time when the output information is received is written to the output time 326 (S805).

Thereafter, a packet 40 is generated in order to output the calculation result of the main system (S806). The packet 40 is generated by storing an identifier and information in each field of the packet. That is, as shown in FIG. 4, the paired host computer node identifier 315 in the VM state management table 301 is stored in the computer node identifier 401 of the packet 40, and the paired VM identifier 314 is stored in the VM identifier 402. Further, the sequence number 322 of the added control information management table 302 is stored in the sequence number 403, and the output information 325 is stored in the output information 404 (S806). When the generation of the packet 40 is completed, the packet is transmitted to another multiplex system output control infrastructure program 300 ′ (S807). Thereafter, the added output information 325 is output to the external communication I / F 104 (S808).

If the result of S803 is a subordinate system, in order to maintain consistency with the data received from the main system, it is checked whether output information has already been written in the output information 325 of the control information management table 302 (S810). As a result, when it is written in the output information 325 (the result of S810 is “Yes”), the output information 325 is compared with the output information received from the VM (S811). If they match as a result of the comparison (the result of S811 is “match”), nothing is done. On the other hand, if they do not match as a result of the comparison (the result of S811 is “mismatch”), the VM is not consistent, and fail-safe processing is executed (S812). In the fail-safe process, the VM may be stopped, or may be reset and participate in the master-slave relationship again.

As a result of the confirmation in S810, if the output information 325 is not written (the result of S810 is “No”), the output information from the VM is written in the output information 325 of the control information management table 302 (S813), and the output information is The received time is written in the output time 325 (S814). This situation occurs when the slave computation is faster than the master.
If the output information from the VM 200 cannot be received within the period time as a result of the process of S802 (the result of S802 is “No”), the process proceeds to the system switching packet generation process (S1205 (FIG. 12)). .
However, the processes S806, S807, S810, S811, and S812 are necessary for achieving master-slave consistency, and these processes may be omitted if master-slave consistency is not required.

FIG. 9 is a process flowchart of the subordinate information management program 305.
When the processing is started, the multi-system output control infrastructure program 300 starts up each VM 200 and starts up the sub system information management program 305. When the secondary information management program 305 receives the packet 40 transmitted from the multi-system output control infrastructure program 300 ′ (S 901), the secondary information management program 305 acquires the output information 404 from the packet (S 902). Then, it is determined whether data is stored in the output information 404 (S903). As a result of the determination, when data is stored in the output information 404 (the result of S903 is “Yes”), the control information of the result calculated in the main system has been received. (S904) (This process will be described in detail with reference to FIG. 10). On the other hand, when no data is stored in the output information 404 of the received packet (the result of S903 is “No”), the packet is transmitted because a failure has occurred in the master system, and the master-slave switching process is performed (S905). ). (This process will be described in detail with reference to FIG. 11).
Note that if there is no need to maintain master-slave consistency, S903 is always “None”, so the process S904 is not necessary.

FIG. 10 is a flowchart showing details of the control information verification / addition processing S904.
The sequence number 402 stored in the received packet 40 is acquired (S1001), the same sequence number as that number is searched from the control information management table 302, and it is checked whether the output information 325 is stored in that row. As a result of the confirmation, when the output information 325 exists in the control information management table 302 (result of “Yes” in S1002), the slave calculation is more advanced than the master system. In this case, since it is assumed that the main system control information is correct, the output information 325 and the received output information 404 are compared and matched (the result of S1003 is “match”), and the process ends without further processing. (S1004). On the other hand, if the comparison result does not match (the result of S1003 is “mismatch”), the consistency of the output result of the slave is not satisfied, and therefore fail-safe processing S1005 such as retry processing is performed.

In the process of S1002, when the output information 325 does not exist in the sequence number 402 stored in the received packet 40 (result of S1002 is “No”), the calculation of the main system is more advanced than the subordinate system. Become. In that case, the received output information 404 is stored in the output information 325 (S1006), and the process is terminated (S1007).
If it is not necessary to maintain the consistency of the subordinate system, the control information verification / addition processing S904 is not necessary.

FIG. 11 is a flowchart showing details of the master-slave switching process S905.
This process is a process for switching the primary system and the secondary system and outputting the calculation result of the secondary system because a failure has occurred in the primary system.
A sequence number 403 (assumed to be n) is acquired from the received packet 40 (S1101). Then, the control information management table 302 is searched to check whether the output information 325 is stored in the row having the same sequence number 322 as the received sequence number 403 (S1102). As a result, when the output information 325 is stored in the row of the same sequence number 322 (the result of S1102 is “Yes”), the stored control information is transmitted to the external communication I / F 104 (S1103).
When there is no output information previously calculated and stored in the secondary control information management table (the result of S1102 is “none”). The master-slave update flag in the VM state management table is updated to “M” (S1107), and the process ends (S1108).

FIG. 12 is a process flowchart of the VM-HOST life and death monitoring program.
When the multi-system output control infrastructure program 300 starts processing, each VM 200 is started and the VM-HOST life / death monitoring program 306 is started. The VM life / death monitoring packet is transmitted to each VM 200 (S1201), and it is confirmed whether there is a response to the “life monitoring” within the life / death monitoring time (S1202). Here, the VM life / death monitoring packet has a destination VM identifier and a command code for confirmation of life / death. As a result of the confirmation, if there is a response related to “life / death monitoring” within the life / death monitoring time (the result of S1202 is “Yes”), nothing is done and life / death monitoring is continued.

On the other hand, when there is no response regarding “life monitoring” within the life / death monitoring time (the result of S1202 is “no”), the master / slave flag 313 of the VM state management table 301 is referred to and the state of the master / slave of the VM 200 not responding is determined. . When the VM 200 that does not respond is the main system (the result of S1203 is “main”), the pair's VM life / death state flag 316 of the VM state management table 301 is referred to in order to examine the life / death state of the pair of VMs (S1204). If the paired VM is dead (the result of S1204 is “dead”), it is determined that a failure has occurred in both the primary and secondary VMs, and a fail-safe process is performed (S1209).

On the other hand, as a result of referring to the paired VM life / death state flag 316 in the VM state management table 301, if the paired VM is alive (result of S1204 is “live”), the system switching packet 40 is generated (S1205). . Here, the system switching packet 40 stores the paired host computer node identifier 315 in the VM state management table in the computer node identifier 401, stores the paired VM identifier 314 in the VM identifier 402, and stores the latest control information. The sequence number 322 of the management table is stored in the sequence number 403. The generated system switching packet 40 is transmitted to the multiplex system output control infrastructure program 300 'of the other computer node 100-B (S1206).
Thereafter, the master-slave flag 313 of the VM state management table 301 is updated to “S” (S1207), and the fail-safe process S1208 is executed.

When the VM 200 that does not respond in S1203 is a slave (the result of S1203 is “slave”), a VM state update packet 70 (FIG. 14) is generated (S1210). In order to update the pair VM live / dead state flag 316 of the VM state management table 300 of the paired VM, the host computer node identifier 315 of the notification destination pair and the VM identifier 314 of the pair are acquired from the VM state management table 300, and the VM Store in the corresponding location of the status update packet 70. Thereafter, the VM state update packet 70 is transmitted to the multi-system output control infrastructure program 300 ′ of the other computer node 100-B (S1211), and the fail-safe process is executed (S1212).

FIG. 13 is a process flowchart of the HOST-HOST life and death monitoring program.
When the multi-system output control infrastructure program 300 starts processing, each VM is started up and the HOST-HOST life / death monitoring program 306 is started up. Then, the packet for life and death monitoring is transmitted to the multiplex system output control infrastructure program 300 ′ of the other computer node 100-B (S1301). Here, the VM life / death monitoring packet has a host computer node identifier, a destination VM identifier, and a command code for confirmation of life / death. If a response to “life / death monitoring” is received within the life / death monitoring time (the result of S1302 is “Yes”), nothing is done and life / death monitoring is continued. On the other hand, if a response to “life / death monitoring” is not received within the life / death monitoring time (the result of S1302 is “none”), all the master-slave flags 313 of the VM state management table 301 are updated to “M” (S1303). Thereafter, the multi-system output control infrastructure program 300 ′ of the other computer node 100-B executes fail-safe processing.

FIG. 15 is a flowchart of the VM state update program.
When the multi-system output control infrastructure program 300 starts processing, the VM 200 is started up and the VM state update program 308 is started up. When a VM state update packet is received from the multiplex system output control infrastructure program 300 'of another computer node 100-B (S1501), a life / death state 703 is acquired from the VM state update packet (S1502). Then, the VM life / death state flag 316 of the VM state management table 301 is updated to the life / death state 703 (S1503).
It should be noted that the above processing operation is executed in the same manner for the same type of program that the multi-system output control infrastructure program 300 'of the other computer node 100-B has.

The first embodiment is an example in which information from the external terminal 9 is input to both the primary and secondary computer nodes via the external communication I /

Fs

104 and 104 ′. On the other hand, in the second embodiment, the input information from the external terminal 9 is input only to the primary computer node 100-A, and is input from the primary computer node 100-A to the secondary computer node 100-B. This is an example of synchronizing information.
Since processing other than the synchronization of input information is the same as that in the first embodiment, the description thereof is omitted. In FIG. 16 and subsequent figures, the same reference numerals are given to the same parts of the drawings in the first embodiment.

FIG. 16 shows the overall configuration of a virtual environment multiplex control computer system.
The difference from the computer system shown in FIG. 1 is that an input information synchronization management table 161 and an input synchronization management program 162 are added. This input information synchronization management table 161 exists in the multi-system output control infrastructure program of the computer nodes constituting the primary system and the secondary system.

17, in the input information synchronization management table 161, the VM identifier 1711 stores an identifier for identifying the VM in the computer node 100. The sequence number 1712 identifies input information to be synchronized. The input information 1713 stores the input information received from the terminal 9 when the VM identifier is “M” from the VM state management table 301. In the case of “S”, the input information from the computer node (for example, computer node 100-A) that controls the main VM transmitted by executing the input information management program 303 (see FIG. 19) is stored. In the case of a computer node that controls the primary VM, the synchronous transmission state 1714 stores “1” indicating transmission completion after transmission to the slave. In the case of a computer node that controls a subordinate VM, “1” is stored upon reception from the main system. In the case of a computer node that controls the primary VM, the synchronous reception state 1715 stores “1” indicating that it has been received upon reception from the secondary.

In the case of a computer node that controls a subordinate VM, processing is performed by the input synchronization management program 162, but “1” is stored by transmission to the main system. In the synchronous transmission state 1714 and the synchronous reception state 1715, the states of the primary and secondary input information synchronization management tables 161 are matched. Therefore, even when the main system is abnormal, the sub system can take over the state of the input information synchronization management table 161 without requiring a separate process.

FIG. 18 shows a format configuration of a packet 180 for communication between computer nodes for input synchronization.
This packet 180 is communicated between computer nodes in order to synchronize input between the master and slave systems. The transmission source computer node identifier 1811 stores the transmission source computer node identifier, and the transmission source VM identifier 1812 stores the transmission source VM identifier. The computer node identifier 1813 stores a reception source computer node identifier, and the VM identifier 1814 stores a reception source VM identifier. As for the sequence number 1815 and the input information 1816, information from the external terminal 9 is stored in the main system. On the other hand, in the slave system, information received from the master system is copied and stored.

Next, processing of the input information management program 303 will be described with reference to FIG.
When the processing is started, the multiplex system output control infrastructure program 300 starts up each VM and starts up the input information management program 303. The input information management program 303 stands by until input information is received from the terminal 9 (S1901). When the input information from the terminal 9 is acquired (S1902), it is confirmed whether the VM that processes the input information is the primary system or the secondary system (S1903).

As a result of the confirmation, in the case of the slave, it waits until it receives the input information from the terminal 9 (S1901). On the other hand, in the case of the main system, a new sequence number is added to the sequence number 1612 of the input information synchronization management table 161 (S1911). Thereafter, the input information output from the terminal 9 is stored in the input information 1713 of the input information synchronization management table 161 in the added sequence number row (S1912). Then, in order to transmit to the computer node that controls the subordinate VM, the communication packet 180 for communication between computer nodes for input synchronization is created (S1913). The packet 180 is transmitted to the computer node that controls the subordinate VM (S1914). Then, “1” is stored in the synchronous transmission state 1614, and the transmission is completed (S1915).

Thereafter, when the packet 180 is received from the computer node that controls the subordinate VM (S1916), "1" is stored in the synchronous reception state 1615 of the input information synchronization management table 161, and the reception state is set (S1917). Further, the sequence number 1612 is stored in the sequence number 322 of the control information management table 302 (S1918), and the input information 1613 is also stored in the input information 323 (S1919). This completes the input information management process.

Next, processing of the input synchronization management program 162 will be described with reference to FIG.
When the processing is started, the multiplex system output control infrastructure program 300 starts up each VM and starts up the input synchronization management program 162. The input synchronization management program 162 stands by until it receives the packet 180 for communication between input synchronization computer nodes from the other computer node 100-B (its multiple system output control infrastructure program) (S2011). When the input information is acquired from the other computer node 100-B (S2012), it is confirmed whether the VM identifier 1714 is the primary system (S2013). In the case of the main system as a result of the confirmation, it waits until it receives a packet 180 for communication between computer nodes for input synchronization from another computer node (S2011).

On the other hand, in the case of the slave as a result of the confirmation, the sequence number 1815 of the packet 180 is stored in the sequence number 1712 of the input information synchronization management table 161 (S2031). Thereafter, the input information 1816 of the packet 180 is stored in the input information 1713 in the row of the added sequence number (S2032). “1” is stored in the synchronous transmission state 1714, indicating that transmission has been made from the computer node that controls the VM of the transmission source VM identifier 1712. The packet 180 is sent back to notify the transmission source computer node that the reception has been completed (S2035). In order to indicate that the data has been sent back, “1” is stored in the synchronous reception state 1715 of the input information synchronization management table 161. The stored sequence number 1712 is stored in the sequence number 322 of the control information management table 302 (S2037). The input information 1613 is stored in the input information 323 of the control information management table 302 (S2038). This completes the input synchronization management process.

1: Virtual multiple system 801: Internal network 802: External network 9: Terminal
100-A, 100-B, 100: Computer node 101: Memory 102: Processor 103: Storage device 104: External communication I / F 105: Inter-node communication I / F
200, 201: Virtual machine (VM) 211: Application 212: Guest OS 213: Virtual communication I / F
300: Multiple system output control infrastructure program 301: VM state management table 302: Control information management table 303: Input information management program 304: Output information management program 305: Subordinate information management program 306: HOST-HOST life / death monitoring program 307: VM -HOST Life / death monitoring program 308: VM state update program 400: Host OS 401: Virtualization program 402: Device control program

Claims

A plurality of physical computers having a host OS and one or a plurality of virtual computers constructed on the host OS are connected via an interface, and the virtual computers can input / output from a commonly connected external device. A virtual multiple system configuration control method in a system,
On the OS of the plurality of physical computers, a multi-system controller that can set a master-slave relationship between paired virtual computers belonging to different physical computers and change the master-slave relationship is installed.
The multi-system control unit of one transfers the packet to and from the other multi-system control unit, monitors the life / death state of the virtual machine in the physical computer of the partner in the master-slave relationship, and The management information indicating the life / death status of the virtual machines in the master-slave relationship is updated according to the result of the above, and the packet is transmitted to the other multi-system control unit according to the monitoring result. Instructing the master-detail relationship to change,
The multi-system control unit 1 transfers the input information received from the external device to the virtual machine having a master-slave relationship, and obtains output information that is a result processed by the virtual computer having a master-slave relationship A virtual multiplex system configuration control method, wherein the output information from the primary or secondary virtual machine is transferred to the external device.
A plurality of multi-system control units, a VM state management table for managing information on the master-slave relationship of the virtual machines in a pair relationship and the life / death state of the virtual machines;
Input information received from the external device, a control information management table for managing output information that is a result of processing in the virtual machine having a master-slave relationship corresponding to the input information, corresponding to the virtual machine,
The multi-system control unit stores the output information obtained from the virtual machine that has transferred the input information in the control information management table, and has a master-slave relationship stored in the VM state management table. Deciding to transfer the output information stored in the control information management table to the external device with reference to the life / death information of the virtual machine;
The virtual multiplex system configuration control method according to claim 1.
The multi-system control unit, when a failure occurs in the virtual machine, refers to the control information management table and controls output information necessary for the next output processing from the processing result at the time of the failure.
The virtual multiplex system configuration control method according to claim 1 or 2.
When a failure occurs in the primary virtual machine, the one multi-system control unit that manages the virtual machine in the main system, the other multi-system that manages the secondary virtual machine. The packet is transferred to the control unit to notify the failure, and the one multi-system control unit and the other multi-system control unit switch the master-slave relationship of the pair of virtual machines.
The virtual multiplex system configuration control method according to any one of claims 1 to 3.
The multi-system control unit of 1 updates the state of the virtual machine in which the failure has occurred, and transmits the status information of the virtual machine in which the failure has occurred and the final output information at the time of the failure to the other multi-system control unit And
The other multiplex system control unit updates the status information of the virtual machine in the VM status management table, and refers to the control information management table to input information necessary for the next output processing of the final output information, Transfer to the virtual machine in the master-slave relationship with the virtual machine in which the failure occurred, and further switch the master-slave relationship;
The virtual multiplex system configuration control method according to any one of claims 2 to 4.
The other multiplex system control unit outputs the output information stored in the control information management table, which is output information from the virtual machine whose master-slave relationship has been switched, to the external device,
The virtual multiplex system configuration control method according to claim 5.
The multi-system control unit, when receiving output information from a virtual machine within a predetermined cycle time, refers to the VM state management table, determines the master-slave state of the output virtual machine,
As a result of the determination, in the case of the main system, the output information obtained from the virtual machine is stored in the control information management table, and then the output information stored in the control information management table is transferred to the external device. And generating a packet for notifying the output information to the other multiplex system control unit and transferring the packet to the other multiplex system control unit,
As a result of the determination, in the case of a secondary system, check whether the output information transferred from the virtual machine of the secondary system is stored in the control information management table,
As a result of the confirmation, the stored output information is compared with the output information obtained from the virtual computer of the main system. When the two do not match, consistency between the virtual computers in the master-slave relationship is obtained. Execute fail-safe processing on the assumption that
The virtual multiplex system configuration control method according to any one of claims 2 to 6.
Input information from the external device is controlled so as to be input to the multiple system control unit of the physical computer in which the primary virtual computer exists, and the primary and secondary virtual computers exist. The multiplex system control unit in the physical computer includes the acquired input information and information indicating a transmission / reception state of a synchronization packet transferred between the multiplex system control unit having a master and a slave virtual computer. An input information synchronization management table for managing the virtual machines in a master-slave relationship,
The multiplex system control unit of 1 confirms whether the virtual computer that is the transfer source of the packet is a primary system or a slave system based on the synchronization packet transferred and received from the other multiplex system control unit,
As a result of the confirmation, in the case of the main system, waiting for the reception of the synchronization packet transferred from the other multi-system control unit,
As a result of the confirmation, in the case of the slave, the acquired input information is stored in the input information synchronization management table, and the information in the synchronous transmission state is stored in the input information synchronization management table, indicating that the reception is completed. Transfer the synchronization packet to the other multiplex system control unit as the transmission source.
The virtual multiplex system configuration control method according to any one of claims 2 to 7.
A plurality of physical computers having a host OS and one or a plurality of virtual computers constructed on the host OS are connected via an interface, and a master-slave relationship is set for virtual computers belonging to different physical computers. A computer system capable of inputting and outputting from a commonly connected external device,
Provided on the host OS of the plurality of physical computers is a multiplex system control unit capable of setting a master-slave relationship between paired virtual computers belonging to different physical computers and changing the master-slave relationship. The system control unit can transmit a packet to communicate management information and control information with each other, and the multiplex system control unit
First management means for managing the master-slave relationship of the virtual machine and information on the life and death of the virtual machine in correspondence with the virtual machine;
Input information received from the external device, second management means for managing output information output to the external device as a result of processing in the virtual machine having a master-slave relationship corresponding to the input information;
The packet is transferred to and from the other multi-system control unit to monitor the life / death status of the virtual machine in the physical computer of the partner in the master-slave relationship, and the master-slave relationship in accordance with the monitoring result Update the contents of the first management means indicating the life or death status of the virtual machines, and change the master-slave relationship between the virtual machines by sending packets to the other multi-system controller according to the monitoring results. First control means for controlling so as to
The input information received from the external device is transferred to the virtual machine in a master-slave relationship, and output information that is a result processed by the virtual computer in a master-slave relationship is acquired and stored in the second management unit And a second control means for controlling the output information from the primary or secondary virtual machine to be transferred to the external device.
The second management means is a control information management table for managing, for each virtual computer, input information input from the external device and output information corresponding to the input information, which is a processing result of the virtual computer,
The second control means, when a failure occurs in the virtual machine, refers to the control information management table and controls output information necessary for the next output processing from a processing result at the time of the failure. Computer system.
The second control means inputs from the control information management table to the virtual machine in consideration of the overall processing time for the input of each virtual machine in the master-slave relationship and the time required for synchronization between the virtual machines in the master-slave system The computer system according to claim 9 or 10, wherein output information that is lost at the time of failure is prevented by generating a processing cycle for transmitting information.
When a failure occurs in the primary virtual machine, the first control means in the multi-system control unit managing the virtual machine in the primary system manages the virtual machine in the secondary system. The computer system according to any one of claims 9 to 11, which performs control so as to transfer a packet to the multiplex system control unit, to notify a failure, and to switch the master-slave relationship of the paired virtual machines.