JP5728812B2 - Distributed information processing system and distributed storage system - Google Patents

Description

  The present invention relates to a distributed information processing system and a distributed storage system, and more particularly to a distributed information processing system and a distributed storage system that can install and upgrade software without stopping business processing.

Conventionally, a distributed information processing system includes a plurality of nodes. When actually performing business, a processing program (hereinafter simply referred to as “software”) is newly added to each of the plurality of nodes. It is necessary to perform a process of switching to a version of software, that is, a software update process. Generally, it is preferable that the software update process in the distributed information processing system is configured to be performed dynamically.
Usually, as a type of virtual machine corresponding to each node of a general distributed information processing system, a hypervisor type virtual machine monitor (hereinafter abbreviated as “VMM”) and business processing are performed on this VMM. There are virtual machines (hereinafter abbreviated as “VM1”) and virtual machines (hereinafter abbreviated as “VM2”) that perform installation or upgrade.

In addition, when the distributed information processing system switches software to a new version of processing software, all nodes agree on a version after the installation of the new version of the software or after the software upgrade (the new version). Therefore, it is necessary to switch all the nodes to the new version of software all at once, and this is one of the problems in the present invention.
Further, the distributed information processing system is preferably capable of continuing business processing even in a situation where nodes operating in the old version and nodes operating in the new version are mixed in the system, and this point is also a problem in the present invention. Met.
Furthermore, when the installation or upgrade of the distributed information processing system fails, it is necessary that the distributed information processing system can be restored to the state before the installation or upgrade. This is also a problem in the present invention.

As a known technique in this field, Patent Document 1 discloses that the first and second guest operating systems communicate with the client 202 using a virtualization technique that operates in each of the virtual machines 310A and 310B. Therefore, a technique is disclosed in which the updated service 312B can be tested on the same hardware 302 on which the online service 312A depends, and it is preferable that both the services are executed simultaneously.
Patent Document 2 discloses a technology that enables updating and canceling update of related servers at the same time without stopping a service when updating software that operates in a general-purpose server system.
Patent Document 3 discloses a technique for easily realizing restoration to an old version module when a failure occurs in a new version module that has started online operation after version upgrade.

In Patent Document 4, the current firmware monitoring process MPR0, which has received a response notification from the current firmware processing process PPR0, immediately requests the new firmware processing process PPR1 to resume processing synchronization points. A technique for realizing switching to firmware (that is, version upgrade) is disclosed.
Further, in Patent Document 5, one common application program or application code 50 and an executable version thereof (with modification in some cases) are included in a plurality of computers or machines M1, M2,. . . Disclosed is a technique that is performed simultaneously or in parallel over the entire Mn. In this technique, the modified structure essentially duplicates the same memory structure and content on each individual machine.

JP 2006-107500 A JP 2006-268172 A JP-A-9-62626 Japanese Patent Laid-Open No. 11-7382 Special table 2009-512030

Incidentally, as described above, one problem imposed on the distributed information processing system is software installation and upgrade. In general, when installing or upgrading the software on all nodes at once, stop the execution of business processes running on the system, stop the processing on all nodes, and then apply the software to each node. It may be necessary to restart each node depending on the software to be applied.
However, in the conventional distributed information processing system described in the above background art, in the distributed storage system configured with a plurality of independent node and storage pairs, each write is completed after completion of writing from the client. Since data is exchanged between the nodes, there is a problem that in order to stop all the nodes, it is necessary to wait until the execution of such data exchange processing is completed.

In addition, in order to apply the method of applying the software after stopping the operation of some nodes, starting the operation when the application is completed, and applying the software of other nodes sequentially, There was a problem that nodes of different versions were mixed.
In other words, data versions and communication protocols are often different between different versions.In such a case, the new version can recognize the difference from the old version, but the old version uses the new version. Since some of the functions of the version were unknown, it is difficult to incorporate compatibility functions to resolve the differences from the new version. For this reason, if the new version and the old version are mixed, However, there was a problem that data could not be exchanged correctly between nodes, and in the worst case, data in the system would be destroyed.
In addition, due to a process failure to install or upgrade software or a bug in the update program, when returning the software before updating it, it is complicated to restore each changed part of the update process so far There was a problem that work would occur.

The present invention
(1) simultaneously operating a virtual machine that performs business and a virtual machine that performs installation / upgrade on the same hardware;
(2) When the installation or upgrade is complete, switch the virtual machine that performs the work,
(3) Then, cooperate with other nodes in the system, agree on a new version that works, and switch all nodes together,
(4) As a dynamic software update method, when a phenomenon or request for canceling an upgrade occurs in the configuration requirements of (1) to (3) above, only a virtual machine that performs a job before switching the virtual machine Delete the virtual machine to be installed and upgraded, and after switching the virtual machine, copy the index table between the virtual machine that was installed or upgraded and the virtual machine that was performing the business. ,
Is the main component requirement.

In addition, as the relevant part of the present invention, the above-described configuration requirement (1) is disclosed in Patent Document 1 and Patent Document 5 described above, and the above-described configuration requirement (2) is disclosed in Patent Document 4 described above. ing. However, none of the above-mentioned patent documents discloses the configuration requirement (3). Here, the configuration requirement (3) is an essential element for solving the problem of “installing or upgrading software without stopping business” according to the present invention.
Furthermore, none of Patent Documents 1 to 5 discloses the above-described configuration requirement (4), but the configuration requirement (4) enables the software of the present invention to be returned to the state before being updated. It is an indispensable constituent requirement to solve the problem.

  An object of the present invention is to provide a distributed information processing system and a distributed storage system for solving the above problems.

In order to solve the above problems, a distributed information processing system according to the present invention is a distributed information processing system including a plurality of nodes connected to a common network, and each of the nodes is a node of its own. A first virtual machine that performs information processing using an old version of the processing function in advance, and a second virtual machine that is virtually provided in the node by an operating system installation process or upgrade process. A second virtual machine having a processing function of an old version equivalent to the processing function of the old version constituting the machine and a processing function of a new version corresponding to the processing function of the old version, and waiting for the start of information processing And the second virtual machine has the installation process or the upgrade process A data updated in the first virtual machine made in between the first virtual machine replication to the second virtual machine information indicating the storage data to be used for information processing is stored When the processing unit that performs the duplication is identified, the communication using the network connecting the other node and the own node is changed from the communication by the first virtual machine to the communication by the second virtual machine. A request to switch and a request to switch the connection from the first virtual machine to the second virtual machine to the shared storage unit that stores information indicating the storage storing the data used for the information processing Switching to the plurality of nodes based on the notification of completion of the switching from the switching unit. The second virtual machine receives the notification of the completion of switching from all of the plurality of nodes, and the processing of the new version is performed from one arbitration node of the plurality of nodes. In response to the notification of the function switching instruction, the information processing is switched to the information processing by the processing function of the new version.

Furthermore, the distributed storage system according to the present invention is characterized by comprising the storage attached to the plurality of nodes, and all configurations of the plurality of nodes and the configuration of the distributed processing system.

  According to the present invention, a virtual machine that performs business processing and a virtual machine that performs installation and upgrade are simultaneously operated on the same hardware, and when installation or upgrade is completed, the virtual machine that performs business is switched to the system. A new version that works with other nodes is agreed upon, and all nodes are configured to switch at the same time, so you can install or upgrade software without stopping business processing. There is an effect.

1 is a configuration diagram illustrating an overall configuration of a distributed information processing system according to an embodiment of the present invention. 1 is a configuration diagram illustrating a configuration example of a distributed storage system to which a distributed information processing system according to an embodiment of the present invention is applied. FIG. 3 is a configuration diagram showing a configuration near a storage in the system shown in FIG. 2. FIG. 4 is an explanatory diagram (1/2) showing an outline of a VM switching operation in the VMM 100 (single node) in the system shown in FIGS. FIG. 4 is an explanatory diagram (2/2) showing an outline of a VM switching operation in the VMM 100 (single node) in the system shown in FIGS. FIG. 4 is an explanatory diagram (1/2) showing details of a VM switching operation in the VMM 100 (single node) in the system shown in FIGS. It is explanatory drawing (2/2) which shows the detail of VM switching operation | movement in VMM100 (single node) of the system shown to FIG. It is explanatory drawing (1/2) which shows the version switching operation | movement in all the nodes of the system shown to FIG. It is explanatory drawing (2/2) which shows the version switching operation | movement in all the nodes of the system shown to FIG. It is a block diagram which shows another example of 1 structure of the distributed storage system with which the distributed information processing system which concerns on embodiment of this invention is applied.

The distributed information processing system according to the present invention performs a process of updating software of all nodes (that is, a dynamic software update process) for a plurality of nodes on the system without stopping business.
Each node on the distributed system includes a hypervisor type virtual machine monitor (VMM), a virtual machine (VM1) that performs business on the VMM, and a virtual machine (VM2) that performs installation or upgrade. Here, VM2 is a replica of VM1, and VM2 installs new software for itself or upgrades existing software, and after this process is completed, dynamically updated data storage on VM1 A table in which location information is stored (hereinafter referred to as “index table”) is migrated to VM2 step by step, and processing is switched from VM1 to VM2 at the stage of complete migration.

In addition, the VM 2 operating in the version before installation or upgrade (that is, the old version) in each node communicates with the VM 2 of other nodes, and all the nodes after the switch destination installation or upgraded version (that is, the new version) And all nodes will switch to the new version at the same time.
With the above processing, all nodes can be switched from the old version to the new version without stopping the business running on the distributed system.
In addition, business can be continued even in a situation where nodes operating in the old version and nodes operating in the new version coexist in the system.

Furthermore, in the unlikely event that the installation or upgrade fails, it is possible to return to the state prior to installation or upgrade without switching from VM1 to VM2 before switching from VM1 to VM2. Also, after switching from VM1 to VM2, update or update the index table by switching from VM1 to VM2 as described above, and install or upgrade by switching from VM2 to VM1 It is possible to return to the previous state.
Hereinafter, embodiments of a distributed information processing system and a distributed storage system according to the present invention will be described in detail with reference to the drawings.

FIG. 1 is a configuration diagram showing the overall configuration of a distributed information processing system according to an embodiment of the present invention.
In the figure, the distributed information processing system according to the present embodiment operates on the hardware 140 and operates on the VMM 130, which is a hypervisor type virtual machine monitor, and stores data and the virtual machine VM 100 in charge of business. Storage 200 to be configured.
The VM 100 includes VMs 110 and 120, which are virtual machines, a shared volume 132 shared between the VM 110 and the VM 120, and a network switching unit 131 that switches the network from the VM 110 to the VM 120.

The VM 110 includes a data division unit 111, a data storage unit 112, a replication unit 113 that replicates an index table 114 (to be described later) to the VM 120, an index table 114 that holds a snapshot index, a change monitoring unit 115, and a VM switching Unit 116, a change management table 117, and a key generation unit 11d.
The VM 120 includes a data division unit 121, a data storage unit 122, a replication unit 123, an index table 124, a change monitoring unit 125, a VM switching unit 126, a change management table 127, an old version operation unit 128, A VM switching completion notification communication unit 129, an old and new version switching unit 12a, a completion notification table 12b, and a version information 1 table 12c are provided.

The replication unit 113 includes a snapshot acquisition unit 113a, a snapshot transmission unit 113b, an update data acquisition unit 113c, an update data transmission unit 113d, a replication completion notification transmission unit 113e, a snapshot application unit 113f, and a snapshot. A reception unit 113g, an update data application unit 113h, an update data reception unit 113i, and a copy completion notification reception unit 113j are provided.
The replication unit 123 includes a snapshot acquisition unit 123a, a snapshot transmission unit 123b, an update data acquisition unit 123c, an update data transmission unit 123d, a replication completion notification transmission unit 123e, a snapshot application unit 123f, and a snapshot. A reception unit 123g, an update data application unit 123h, an update data reception unit 123i, and a copy completion notification reception unit 123j are provided.

Hereinafter, functions of the distributed information processing system according to the embodiment of the present invention will be described based on functions of main components.
First, the functions of the duplicating unit 113 and the duplicating unit 123 will be described based on the function of each component.
The snapshot acquisition unit 113a acquires a snapshot at a specific time. The snapshot transmission unit 113b transmits the snapshot acquired by the snapshot acquisition unit 113a to the VM 120.
The snapshot reception unit 123g receives in the VM 120 the snapshot transmitted by the snapshot transmission unit 113b.

The snapshot application unit 123f applies the snapshot acquired by the snapshot reception unit 123g to the index table 124 of the VM 120.
When the update data acquisition unit 113c acquires a snapshot of the index table 114 of the VM 110 and confirms that it has been applied to the index table 124 of the VM 120, the update data acquisition unit 113c records the update part information of the index table 114 to the change monitoring unit 115. The update management table 117 is referred to instruct the acquisition of update part information and update data, and the returned update part information and update data are transmitted to the VM 120 using the update data transmission unit 113d.

The update data transmission unit 113d transmits the update partial information and update data acquired by the update data acquisition unit 113c to the VM 120.
The update data reception unit 123i receives the update partial information and update data transmitted by the update data transmission unit 113d in the VM 120.
The update data application unit 123h applies the update partial information and update data acquired by the update data reception unit 123i to the index table 124 of the VM 120.
When the change monitoring unit 115 confirms that no new update part exists, the replication completion notification transmission unit 113e notifies the VM 120 that the replication of the index table 114 has been completed.

Upon receiving the notification transmitted by the replication completion notification transmission unit 113e, the replication completion notification receiving unit 123j notifies the VM switching unit 126 that the replication of the index table 124 has been completed.
The VM switching unit 126 instructs the VMM 130 to switch the network of the VM 110 that is performing the business to the VM 120 and the connection of the shared volume 132 from the VM 110 to the VM 120, and the network and the shared volume 132 are switched. When completed, the VM switching completion notification communication unit 129 is notified.
The change monitoring unit 115 monitors the change of the index table 114 from the time when the snapshot acquisition unit 113a acquires the snapshot of the index table 114.

The VM switching unit 126 receives the notification of the completion of the replication of the index table 114 from the replication completion notification receiving unit 123j, instructs the VMM 130 to switch the VM, switches the connection of the shared volume 132 from the VM 110 to the VM 120, and the old version The operation of the operation unit 128 is started.
In response to the VM switching instruction from the VM switching unit 126 of the VM 120, the network switching unit 131 switches the network that has received data in service from the VM 110 to the VM 120.
The old version operation unit 128 operates in the same manner as the old version on the VM 120 immediately after switching from the VM 110 to the VM 120 when there is a node for which VM switching has not been completed. This eliminates a state where a node that operates in the old version and a node that operates in the new version are executed in a mixed manner.

When the notification of the VM switching completion is received from the VM switching unit 126 in its own node, the VM switching completion notification communication unit 129 notifies the VM switching completion to all the nodes in the system. Further, a VM switch completion notification from another node is recorded in the completion notification table 12b, and it is determined from the completion notification table 12b whether or not the VM switching of all the nodes is completed, and the VM switching of all the nodes is completed. If so, the switching process from the old version to the new version is started.
When the VM switching completion notification communication unit 129 notifies the old / old version switching unit 12a that the VM switching of all the nodes has been completed, all the nodes are switched by, for example, a general known agreement algorithm (two-phase commit or the like). To agree on a new version to switch to. When all the nodes agree, each node switches the processing operation from the old version operation to the new version operation. However, each node that has completed switching has suspended transmission of data to the node that is operating in the old version before switching, and when this node is notified that switching to the new version has been suspended Start sending data.

(Description of operation)
FIG. 2 is a configuration diagram showing one configuration example of a distributed storage system to which the distributed information processing system according to the embodiment of the present invention is applied.
FIG. 3 is a configuration diagram showing the configuration in the vicinity of the storage of the distributed storage system (distributed storage system according to the present invention) shown in FIG.
Each of the nodes 1 to n is a single node and has the configuration of the VMM 100 (FIG. 1).
Further, each of the nodes 1 to n is communicably connected to the wide area network 10000, and each of the nodes 1 to n is connected to storages 1 to n each having a storage number corresponding to a node code ( 1 to n are storage numbers). It is assumed that the storage 200 (FIG. 1) is applied to each of the storages 1 to n.

In the following, in the configuration of the distributed information processing system shown in FIG. 1, in particular, when a component of the node i (i is 1 to n) is shown, a node number is added before the symbol of the component shown in FIG. 1. Shall. For example, the replication unit 113 of the node 1 is referred to as a replication unit 1113. However, the description of the node number is omitted in the case where the description is applied to the whole node, not the specific description specifying a specific node.
Hereinafter, the operation of the distributed information processing system according to the present embodiment will be described with reference to FIG. 1 and the configuration diagrams shown in FIGS.
As shown in FIG. 3, for example, the OS 1110 of the VMM 1100 (node 1) includes a key generation unit 111d, a data division unit 1111, a data storage unit 1112, and an index table 1114, and performs data writing processing as described later. . When the data is written, a key for the data is created by the key generation unit 111d and registered in the index table 1114. Subsequently, the data dividing unit 1111 divides the data, the data saving unit 1112 saves the divided data in the storage 200 of each node 100, and saves the saved location in the index table n114. The other nodes (2 to n) are the same as the node 1 described above.

4 and 5 are explanatory diagrams showing an outline of the VM switching operation in the VMM 100 (single node) in the system shown in FIGS.
As shown in FIGS. 4 and 5, a hypervisor type virtual machine monitor (VMM 100) is allocated to each node, and a virtual machine (VM 110) that performs business is arranged thereon. At the time of software installation or upgrade, a virtual machine (VM 120) for applying software is newly created as a copy of the OS image of the VM 110.
After the software is applied to the VM 120, the setting information on the VM 110 and the index table 114 are transferred to the VM 120 via the shared volume 132, the network being executed in the business is switched from the VM 110 to the VM 120, and the VM performing the business is switched. Immediately after this switching, the version after the update does not operate, and the version before the update operates. After installing or upgrading the software on all nodes, switch to the new version.

6 and 7 are explanatory diagrams showing details of the VM switching operation in the VMM 100 (single node) in the system shown in FIGS.
As shown in FIGS. 6 and 7, after the VM 120 is replicated from the VM 110, when installation or upgrade processing is in progress in the VM 120, the index table 114 replication processing and change monitoring processing start-up and change management are performed in parallel with the business processing in the VM 110. Preparation for taking over work such as generation of the table 117 is started.
Upon completion of the installation process of the VM 120, the snapshot acquisition unit 113a acquires a snapshot of the index table 114 on the VM 110 (the data of the index table at that time), and the acquired snapshot is transmitted to the VM 120 by the snapshot transmission unit 113b. The received snapshot is received by the snapshot receiving unit 123g, and the received snapshot is reflected in the index table 124 on the VM 120 by the snapshot applying unit 123f.

When the snapshot of the index table 114 on the VM 110 is acquired, the change monitoring unit 115 starts recording the changed part of the index table 114.
In addition, when the snapshot application of the VM 110 index table 114 with respect to the index table 124 of the VM 120 is completed, the update data acquisition unit 113c inquires the change monitoring unit 115 about the change location, and the change location information and change data of the index table 114 are obtained. The acquired change location information and change data are transmitted to the VM 120 by the update data transmission unit 113d, the received change location information and change data are received by the update data reception unit 123i, and the received change location information and change data are received. Is reflected in the index table 124 on the VM 120 by the update data application unit 123h.

When business is being performed, the index table 114 of the VM 110 changes frequently, so the update data is applied to the index table 124 of the VM 120 as needed by the above means. When this operation is repeated and there is no more update data of the VM 110, the change monitoring unit 115 sends a replication completion notification to the replication completion notification transmission unit 113e, and the replication completion notification transmission unit 113e sends an index table to the VM 120. A copy completion notice 114 is transmitted. Upon receiving the replication completion notification, the replication completion notification receiving unit 123j of the VM 120 notifies the VM switching unit 126 of the replication completion of the index table 114.
When the VM switching unit 126 receives a notification of replication completion of the index table 114 from the replication completion notification receiving unit 123j, the VM switching unit 126 instructs the network switching unit 131 of the VMM 130 to switch the network in use in the business of the VM 110 to the VM 120. The VMM 130 is instructed to switch the connection of the shared volume 132 from the VM 110 to the VM 120. When a notification of completion of network switching is received from the network switching unit 131, it notifies the VM switching completion notification communication unit 129 that the switching of the VM has been completed.

The network switching unit 131 switches the network of the VM 110 currently connected in business to the VM 120 in accordance with an instruction from the VM switching unit. When the network switching is completed, the VM switching unit 126 is notified that the network switching is completed.
8 and 9 are explanatory diagrams showing version switching operations in all nodes in the system shown in FIGS.
The old version operation unit 128 starts the operation at the same time when the network is switched, and performs the same processing as that of a node that has not yet been installed. This process allows business to continue without problems even if different versions are mixed.
When the VM switching completion notification is notified from the VM switching unit 126, the VM switching completion notification communication unit 129 generates the completion notification table 12b, and the VM switching of its own node is completed for all nodes. To be notified. When a VM switch completion notification is received from another node, the completion is recorded in the node corresponding to the completion notification table 12b.

When the VM switching completion notification unit 129 of each node refers to the completion notification table 12b and determines that all the nodes in the system have completed the switching of the VM, the VMs of all the nodes are displayed in the new / old version switching unit 12a. Notify that it has been switched.
The old and new version switching unit 12a uses the version information table 12c managing the version of each node and a general agreement algorithm (such as two-phase commit), and makes an agreement on a new version to be switched at each node. In the process of the agreement, an arbitrary node is first set as an arbitrator, and the arbitrator node 120 sends VM switching version information and a switchable message to all nodes (1120 to n120).

  The node (1120 to n120) that has received the sent version information determines whether or not it matches the version information of its switching destination, and if it is correct, returns information indicating OK to the mediator node 120. Upon confirming that the OK information has been returned from all nodes (1120 to n120), the mediator node 120 instructs all nodes (1120 to n120) to switch the operation to the new version. Then, the mediator node 120 also switches to the operation to the new version. When the switching is completed, each node notifies all the nodes that the switching to the new version has been completed, and each node that has received this notification stores the corresponding node and the new node in the version information table 12c. Record the version information of the version. Since the switching timing of each node is different, a node operating in the new version and a node operating in the old version are mixed. Therefore, referring to the version information table 12c, the transmission of data to a node that has not been switched is suspended, and the transmission of data is resumed when the switching is completed.

FIG. 10 is a configuration diagram showing another configuration example of the distributed storage system to which the distributed information processing system according to the embodiment of the present invention is applied.
In the system having the configuration shown in FIG. 10, when installation or upgrade processing fails due to a program bug or communication delay, or when installation or upgrade processing is canceled, installation or upgrade processing is canceled on all nodes. For the node that has not been switched to the VM 120, the processing for switching to the VM 120 is stopped, the processing of the task in the VM 110 is continued, and the index table 124 of the VM 120 for the node that has been switched to the VM 120. Is copied to the VM 110, and the business processing is continued in the VM 110. As a result, the system can be restored to the state before the installation or upgrade.

When a phenomenon or request for canceling installation or upgrade occurs in a specific node, the cancellation of installation or upgrade processing is notified from the node to all nodes.
The node that has not received the cancellation notification of the installation or upgrade process and has not yet switched to the VM 120 stops the switch process to the VM 120, continues the business process in the VM 110, and extinguishes the VM 120.
The node that has been switched to the VM 120 that has received the notification of cancellation of the installation or upgrade process replicates the index table 124 of the VM 120 to the VM 110. Therefore, a system in which the VM 110 and the VM 120 are replaced in FIG. The index table 124 is replicated step by step in the system having the configuration shown in FIG.

The snapshot acquisition unit 123a acquires a snapshot of the index table 124 on the VM 120, transmits the acquired snapshot to the VM 110 by the snapshot transmission unit 123b, and receives the transmitted snapshot by the snapshot reception unit 113g. The received snapshot is reflected in the index table 114 on the VM 110 by the snapshot application unit 113f.
The change monitoring unit 125 starts recording the changed part of the index table 124 when a snapshot of the index table 124 on the VM 120 is acquired.
The update data acquisition unit 123c inquires the change monitoring unit 125 about the change location when the snapshot application of the index table 124 of the VM 120 with respect to the index table 114 of the VM 110 is completed, and acquires the change location information and the change data of the index table 124. The acquired change location information and change data are transmitted to the VM 110 via the update data transmission unit 123d.

The update data receiving unit 113i receives the transmitted change location information and change data, and reflects the received change location information and change data in the index table 114 on the VM 110 by the update data application unit 113h.
While the business is being performed, the index table 114 of the VM 120 changes frequently. Therefore, the update data is applied to the index table 114 of the VM 110 as needed by the above means. This operation is repeated, and when there is no update data of the VM 120, the change monitoring unit 125 sends a replication completion notification to the replication completion notification transmission unit 123e, and the replication completion notification transmission unit 123e sends an index table to the VM 110. A copy completion notification 124 is transmitted.
Upon receiving this replication completion notification, the replication completion notification receiving unit 113j of the VM 110 notifies the VM switching unit 116 of the completion of index table replication.

When the VM switching unit 116 receives a notification of replication completion of the index table 124 from the replication completion notification receiving unit 113j, the VM switching unit 116 instructs the network switching unit 131 of the VMM 130 to switch the network used in the business of the VM 120 to the VM 110. The VMM 130 is instructed to switch the connection of the shared volume 132 from the VM 120 to the VM 110. Further, the network switching unit 131 receives a notification of network switching completion.
In response to an instruction from the VM switching unit 116, the network switching unit 131 switches the network of the VM 120 currently connected in business to connection with the VM 110. When the network switching is completed, the network switching unit 131 notifies the VM switching unit 116 that the network switching is completed.

  According to the distributed information processing system according to the present embodiment, a virtual machine that performs business on the same hardware and a virtual machine that performs installation and upgrade are simultaneously operated, and when installation or upgrade is completed, the virtual machine that performs business , Switch to other nodes in the system, agree on a new version that works, and configure all nodes to switch at the same time, so it is possible to install and upgrade software without stopping operations Yes.

  In addition, after it has been agreed that the virtual machine that has been installed or upgraded is to run on the old version first, and that all nodes will switch from running on the old version to running on the new version, each node will When switching the operating version, the node that switched to the operation in the new version holds the data to be sent to the node that operates in the old version, and is notified that the node that operates in the old version has switched to the new version. Since the transmission of data is resumed at the point in time, the operation can be continued even in a situation where nodes operating in the old version and nodes operating in the new version coexist in the system. Yes.

  Furthermore, before switching the virtual machine, it is only necessary to delete the virtual machine to be installed and upgraded, leaving only the virtual machine that performs the business, and after switching the virtual machine, the installation or upgrade is performed. Since the index table is replicated between the performed virtual machine and the virtual machine that has been operating, there is an effect that the software can be returned to the state before the update.

  Each of the above devices has a computer system inside. Each process described above is stored in a computer-readable recording medium in the form of a program, and the above process is performed by the computer reading and executing the program. Here, the computer-readable recording medium means a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, a semiconductor memory, or the like. Alternatively, the computer program may be distributed to the computer via a communication line, and the computer that has received the distribution may execute the program.

  The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.

  The distributed information processing system according to the present invention is applicable to the construction of a distributed information processing system and a distributed storage system to which the system is applied, and in particular, installs and upgrades software without stopping business processing. It is suitable for the construction of a distributed information processing system that can be used and a distributed storage system to which the system is applied.

1 to n nodes 110, 120 VM (virtual machine)
111, 121 Data division unit 112, 122 Data storage unit 113, 123 Replication unit 114, 124 Index table 115, 125 Change monitoring unit 116, 126 VM switching unit 117, 127 Change management table 11d Key generation unit 128 Old version operation unit 129 VM switching completion notification communication unit 130 VMM (virtual machine monitor)
131 Network switching unit 132 Shared volume

Claims (6)

A distributed information processing system comprising a plurality of nodes connected to a common network,
A first virtual machine that performs information processing in advance by the processing function of the previous version in each of the nodes, and a second virtual machine that is virtually provided in the node by operating system installation processing or upgrade processing And having an old version processing function equivalent to the old version processing function constituting the first virtual machine and a new version processing function corresponding to the old version processing function, and waiting for the start of information processing A second virtual machine,
The second virtual machine stores data used for information processing, which is data updated in the first virtual machine performed with the first virtual machine during the installation process or the upgrade process. When the processing unit that performs the duplication of the information indicating the storage that has been identified to the second virtual machine is identified, the communication using the network connecting the other node and its own node is performed. A request for switching from communication by a virtual machine to communication by the second virtual machine and connection to a shared storage unit that stores information indicating a storage in which data used for the information processing is stored are connected from the first virtual machine. A request to switch to the second virtual machine is sent to the switching unit of its own node, and the switching unit Notifies completion of switching to the plurality of nodes based on the notification of completion of the switching of the al,
The second virtual machine is instructed to switch from one arbitration node of the plurality of nodes to the processing function of the new version based on receiving notification of switching completion from all of the plurality of nodes. The information processing system is switched to the information processing by the processing function of the new version. After the second virtual machine notifies the arbitration node of the completion of switching, the second virtual machine performs the information processing by the old version processing function equivalent to the old version processing function constituting the first virtual machine. ,
When notification of switching instruction to the processing function of the new version is received from one of the plurality of nodes based on receiving notification of switching completion from all of the plurality of nodes The information processing by the processing function of the old version equivalent to the processing function of the old version constituting the first virtual machine is switched to the information processing by the processing function of the new version. 2. A distributed information processing system according to 1. The second virtual machine completes switching to the information processing by the new version processing function for each of the plurality of nodes after the switching to the information processing by the new version processing function is completed. To the node that has not received the notification that the switching to the information processing by the processing function of the new version has been completed among the plurality of nodes, the information by the processing function of the new version The data transmission in the processing is suspended, and the data transmission to the node is performed when the node receives notification that the switching to the information processing by the processing function of the new version is completed. Item 3. A distributed information processing system according to Item 2. The processing function of the new version is a processing function provided in the second virtual machine by an operating system installation process or an upgrade process,
The second virtual machine is
When the installation process or the upgrade process is failed or canceled, the cancellation is notified to all of the other nodes, and the processing function of the old version of the first virtual machine is the time when the failure or cancellation is detected. Before the switch to the information processing by the processing function of the new version of the second virtual machine from the information processing of the first virtual machine to the information processing from the information processing by the processing function of the old version of the first virtual machine. The switching to the information processing by the processing function of the new version of the virtual machine of No. 2,
When the installation process or the upgrade process is canceled, the cancellation is notified to all of the other nodes, and the time when the cancellation is performed is the information processing by the processing function of the old version of the first virtual machine. From the information processing by the processing function of the new version of the second virtual machine to the first virtual machine after switching to the information processing by the processing function of the new version of the second virtual machine Switching the information processing by the processing function of the old version from the information processing by the processing function of the old version of the first virtual machine to the information by the processing function of the new version of the second virtual machine The distribution according to any one of claims 1 to 3, wherein the distribution is performed in the same manner as switching to processing. Type information processing system. The processing function of the new version is a processing function provided in the second virtual machine by an operating system installation process or an upgrade process,
The second virtual machine is
If the installation or upgrade process fails or is canceled, all of the other nodes are notified of the cancellation,
Upon receiving the cancellation information from said other node, to the first of the said information processing by the new version of the processing function of the virtual machine from the information processing by the old version of the processing functions said second virtual machine Before the switching, the switching from the information processing by the processing function of the old version of the first virtual machine to the information processing by the processing function of the new version of the second virtual machine is stopped,
The point in time when the cancellation information is received from the other node is changed from the information processing by the processing function of the old version of the first virtual machine to the information processing by the processing function of the new version of the second virtual machine. Is switched from the information processing by the new version processing function of the second virtual machine to the information processing by the old version processing function of the first virtual machine. The switching from the information processing by the processing function of the old version of the virtual machine to the information processing by the processing function of the new version of the second virtual machine is performed. The distributed information processing system according to any one of the above.   The storage system connected to the plurality of nodes is provided, and all the configurations of the plurality of nodes are the configuration of the distributed information processing system according to any one of claims 1 to 5. Distributed storage system.

Images