JP2009080705A - Virtual machine system and method for restoring virtual machine in the system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To restore the state of a virtual machine up to a point of time of fault occurrence independently of a state of keeping consistency with the outside of a system without waiting an output to the outside of the system. <P>SOLUTION: A communication recording apparatus 31 records communication data input/output between an external system 40 and a virtual machine 11 in time-series order. When the restoration of the virtual machine 11 is required due to occurrence of a fault, a snapshot management mechanism 121 restores the virtual machine to an acquisition point of time (first point of time) of the snapshot on the basis of the latest snapshot of the virtual machine 11 stored in a snapshot storage part 122. A log reproduction mechanism 144 charges input data from the first point of time included in a communication log which is recorded in the communication recording apparatus 31 up to the second point of time which is the time of fault occurrence to the virtual machine 11 in the time-series order to restore the virtual machine 11 up to the state of the second point of time. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a virtual machine system including a virtual machine that operates on a physical machine, and in particular, a virtual machine system suitable for restoring the virtual machine in the event of a failure of a physical machine on which the virtual machine operates, and the system The present invention relates to a virtual machine restoration method.

  In general, in a computer system, when a failure occurs, it is required to restore the system state immediately before the failure so that the processing can be resumed. As a method for restoring such a system state, a snapshot recovery method and a checkpoint recovery method are conventionally known.

  The snapshot recovery method, for example, periodically acquires and saves the internal state of the system as a snapshot, and when a failure occurs in the system, acquires it at the time closest to the failure occurrence time (snapshot time) The system is restored to the state of the snapshot time closest to the failure occurrence time based on the made snapshot (that is, the latest snapshot).

On the other hand, as described in Patent Document 1, for example, the checkpoint recovery method is a checkpoint set when a computer system communicates with the outside of the system (another computer system) via a network. The history of input / output (transmission / reception) to / from the network is stored, and when a failure occurs in the system, the state of the system is restored using the stored history. Here, responses (outputs) to all inputs to the system from the outside of the computer system are waited until the next checkpoint.
JP 2000-194624 A

  As described above, with the conventional snapshot recovery method, the computer system state can be restored only when the latest snapshot is obtained, and the state after the latest snapshot is obtained cannot be restored. is there.

  On the other hand, the conventional checkpoint recovery method (checkpoint recovery method described in Patent Document 1) makes it possible to recover the state of the computer system based on the history after the checkpoint. However, in the checkpoint recovery method described in Patent Document 1, responses (outputs) to all inputs to the system from the outside of the computer system need to wait until the next checkpoint. For this reason, in the checkpoint recovery method, the performance in normal processing in the computer system is degraded. Also, checkpoints cannot be acquired unless the computer system can maintain consistency with the outside.

  For this reason, it is required that the computer system can be restored to the point of failure after the snapshot, regardless of the state in which consistency with the outside can be maintained without waiting for output to the outside of the system. The

  By the way, in recent years, a virtual machine system including a virtual machine (virtual machine) has been developed. A virtual computer is known as a virtualized computer that operates on a physical computer. The virtual machine includes a virtualized disk, memory, CPU (processor), and network interface. By installing an operating system (OS) and applications (application programs) on this virtual machine, the OS and applications can be executed. When any input is made from outside the system (external system) via the network, the virtual machine sends the processing result corresponding to the input to the outside of the system as an output (response) to the input.

  Therefore, even in such a virtual machine system, for example, when a failure occurs in a physical machine on which the virtual machine operates, the state of the virtual machine is changed to a point in time after the snapshot (for example, a failure occurs) It is required to be able to restore until the time of occurrence).

  The present invention has been made in consideration of the above circumstances, and its purpose is to make the state of the virtual machine independent of the state in which consistency with the outside can be maintained without waiting for the output of the virtual machine to the outside of the system. It is to provide a virtual machine system that can restore a virtual machine system to a point of failure after a snapshot and a virtual machine restoration method in the system.

  According to one aspect of the present invention, a virtual computer system is provided. The system includes a physical computer, snapshot storage means used for storing a snapshot of a virtual computer operating on the physical computer or another physical computer connected to the physical computer via a network, and And an external system using a virtual machine and communication recording means for recording communication data inputted / outputted between the virtual machines in a chronological order as a communication log. The physical computer, when the virtual computer operates on the physical computer, acquires the state of the virtual computer as a snapshot of the virtual computer, and stores the snapshot in the snapshot storage unit; The virtual machine is restored on the physical computer after a failure that prevents the virtual machine from functioning normally while the virtual machine is operating on the physical computer or the other physical computer. Snap that restores the virtual machine to the first point in time when the snapshot was acquired based on the latest snapshot of the virtual machine stored in the snapshot storage means Shot restoration means, and the virtual machine restored at the first time point, the communication recording means By inputting input data from the first time point to the second time point when the failure occurs included in the communication log recorded in the above order, the virtual machine is moved in the time series. Log replaying means for restoring to the second time point.

  According to the present invention, in addition to restoring a virtual machine to a snapshot acquisition point by snapshot, communication data input / output between the external system and the virtual machine is recorded in time series as a communication log. By inputting the input data from the snapshot acquisition time to the failure occurrence time included in the communication log to the virtual machine restored to the snapshot acquisition time in chronological order The virtual machine can be restored to the point of failure regardless of the state in which the consistency with the outside can be maintained without waiting for the output of the virtual machine to the outside of the system. ,

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a block diagram showing a configuration of a virtual machine system according to an embodiment of the present invention. The virtual computer system of FIG. 1 includes a plurality of physical computers (physical server computers), for example, two physical computers 10-1 and 10-2. The physical computers 10-1 and 10-2 are connected to the network 20.

  The physical computers 10-1 and 10-2 include well-known hardware resources (not shown) such as a CPU, an I / O device, a memory, and a network interface (network interface card: NIC). The hardware resources of the physical computers 10-1 and 10-2 are virtualized to provide an environment (virtual computer execution environment) in which the virtual computer (virtual machine) operates. FIG. 1 shows a state in which the virtual machine 11 is operating in the virtual machine execution environment of the physical machine 10-1. Note that a configuration in which a plurality of virtual machines including the virtual machine 11 operate on the physical machine 10-1 may be employed.

  The hypervisors 12-1 and 12-2 operate on the physical computers 10-1 and 10-2, respectively. The hypervisors 12-1 and 12-2 are also called a virtual machine manager (virtual machine manager: VMM) or a virtual machine monitor (virtual machine monitor: VMM). The hypervisors 12-1 and 12-2 manage the virtual computers by managing the use of the above-described hardware resources of the physical computers 10-1 and 10-2, respectively. For example, the hypervisors 12-1 and 12-2 provide a virtual computer execution environment in which a virtual computer operates by virtualizing hardware resources of the physical computers 10-1 and 10-2. That is, the hypervisors 12-1 and 12-2 construct a virtual computer having virtualized hardware resources.

  The virtual computer 11 operating on the physical computer 10-1 includes a system unit 13 and an adapter 14. The system unit 13 is a main body of the virtual machine 11 and executes an application (application program) according to an OS (operating system) operating on the virtual machine 11. As a result, the system unit 13 performs input and output with the external system 40 via the adapter 14, the network 20, and the switch 30 described later.

  The adapter 14 constitutes an interface mechanism for the virtual machine 11 (the internal system unit 13). The adapter 14 includes an input / output mechanism 141, a log storage unit 142, a log storage processing mechanism 143, a log reproduction mechanism 144, and a log comparison mechanism 145.

  The input / output mechanism 141 performs input / output between the system unit 13 and the network 20. The input / output mechanism 141 is also a log transmission mechanism (to be described later) in the switch 30 in order to restore the state of the virtual machine 11 operating on the physical machine 10-1 as a result of the failure of the physical machine 10-1. When a log (communication history) is transmitted from 35, the log is input and passed to the log storage processing mechanism 143. The log storage unit 142 is used to temporarily store a log transmitted from the log transmission mechanism 35. In this embodiment, the log storage unit 142 is realized by allocating a memory area of the physical computer 10-1. The log storage processing mechanism 143 stores the log passed from the input / output mechanism 141 in the log storage unit 142.

  The log reproduction mechanism 144 reproduces an input to the system unit 13 and an output from the system unit 13 (that is, an input / output to / from the system unit 13) according to the input based on the log stored in the log storage unit 142 ( Reproduce. When the state of the system unit 13 is restored, the log comparison mechanism 145 compares the output from the system unit 13 with respect to the input to the system unit 13 with the output with respect to the input included in the log.

  The hypervisor 12-1 includes a snapshot management mechanism 121 and a snapshot storage unit 122. The snapshot management mechanism 121 has a snapshot acquisition function that periodically acquires the state of the virtual machine 11 including the system unit 13 as a snapshot, for example. The snapshot management mechanism 121 also has a snapshot restoration function for restoring the virtual machine 11 to the snapshot acquisition time point using the acquired snapshot. The state of the virtual machine 11 includes a CPU state (program counter and register state) and a memory state assigned to the virtual machine 11.

The snapshot storage unit 122 is used to store a snapshot acquired by the snapshot management mechanism 121. In this embodiment, the snapshot storage unit 122 is realized by allocating an area of a permanent storage device such as a disk device possessed by the physical computer 10-1. The snapshot storage unit 122 may be realized by allocating an area of a shared disk device (not shown) shared by the physical computers 10-1 and 10-2.
Although omitted in FIG. 1, the hypervisor 12-2 has the same configuration as the hypervisor 12-1.

  A switch 30 for connecting computer systems to each other via the network 20 is connected to the network 20. In the state of FIG. 1, the switch 30 connects the external system 40 to the virtual computer 11 on the physical computer 10-1 via the network 20. The external system 40 refers to a computer system external to the physical computer 10-1 (the virtual computer 11 above). In this embodiment, the external system 40 is a client terminal that receives a service provided by the physical computer 10-1 (the upper virtual computer 11), but is a computer system (physical server computer) similar to the physical computer 10-1. It doesn't matter.

  The switch 30 includes a communication recording device 31. The communication recording device 31 acquires and records a log (history) of communication via the network 20 and the switch 30. The communication recording device 31 includes a log acquisition mechanism 32, a log storage unit 33, a log storage processing mechanism 34, and a log transmission mechanism 35.

  The log acquisition mechanism 32 acquires a log of communication via the network 20 and the switch 30. Here, in order to simplify the description, the log acquired by the log acquisition mechanism 32 is between the virtual computer 11 (the system unit 13 included in the physical computer 10-1) and the external system 40. It is assumed that this is a log of communication performed via the network 20 and the switch 30. In the following description, communication (communication data) from the external system 40 to the virtual computer 11 on the physical computer 10-1 is referred to as input (input data), and vice versa, communication from the virtual computer 11 to the external system 40 is performed. (Communication data) is called output (output data).

  The log storage unit 33 is a storage unit used to store a communication log (input / output data) acquired by the log acquisition mechanism 32. The log storage processing mechanism 34 stores the input / output data acquired by the log acquisition mechanism 32 in the log storage unit 33 in chronological order. In the present embodiment, the input data stored in the log storage unit 33 by the log storage processing mechanism 34 includes time information indicating the time (communication time) when the data is sent to the virtual machine (here, the virtual machine 11). Attached. The log transmission mechanism 35 transmits a communication log stored in the log storage unit 33 to a virtual computer on a designated physical computer in response to an instruction from the control mechanism 50 described later.

  The physical computer 10-1, the physical computer 10-2, and the switch 30 are also connected to the network 21. A control mechanism 50 is also connected to the network 21. The control mechanism 50 roughly performs two controls. The first control is to synchronize a snapshot of a virtual machine operating on a physical computer connected to the network 21 and an input / output log (communication log) performed by the virtual machine. Second, when a failure occurs in a physical computer connected to the network 21, control is performed to restore the virtual computer operating on the physical computer to the state immediately before the failure.

  Note that the communication recording device 31 may be provided in a router or a proxy server (a proxy server proxying access from the external system 40 to a virtual computer operating on the physical computer 10-1 or 10-2).

Next, operations in the virtual machine system shown in FIG. 1 will be described.
First, an outline of snapshot acquisition by the snapshot management mechanism 121 arranged in the hypervisor 12-1 on the physical computer 10-1 will be described with reference to FIG. FIG. 2 shows the relationship between the state of the virtual machine 11, the snapshot time, and the snapshot.

  In this embodiment, the snapshot management mechanism 121 functions as a snapshot acquisition unit and periodically acquires a snapshot of the virtual machine 11. Here, time t1 and t2 are included as snapshot acquisition times. In this case, the snapshot management mechanism 121 acquires the state 201 of the virtual machine 11 at time t1 (the state of the virtual machine 11 including the state of the system unit 13) 201 as the snapshot 202 at time t1. Further, the snapshot management mechanism 121 acquires the state 203 of the virtual machine 11 at time t2 as the snapshot 204 at time t2. The acquired snapshots 202 and 204 are stored in the snapshot storage unit 122. The snapshot storage unit 122 also stores snapshot management information for managing the acquired snapshot generation.

  As described above, by acquiring the snapshots 202 and 204 of the virtual machine 11 at the times t1 and t2, respectively, and storing them in the snapshot storage unit 122, at any time t2 + Δt after the time t2, the virtual machine 11 can be restored to the state at time t1 or t2. Note that when applying a configuration in which only the latest snapshot is stored in the snapshot storage unit 122, the virtual machine 11 can be restored only to the state at time t2. In any case, the virtual machine 11 can be restored to at least the state at time t2.

  Next, recording of a log of communication between the virtual computer 11 (system unit 13 thereof) and the external system 40 by the communication recording device 31 in the switch 30 will be described with reference to FIG. FIG. 3 shows a flow of data (input data) at the time of communication (input) from the external system 40 to the virtual machine 11 and data (output) at the time of communication (output) from the virtual machine 11 to the external system 40 for the input data. Data) flow.

  First, data (input data) 301 addressed to the virtual machine 11 (system unit 13 thereof) from the external system 40 is sent to the adapter 14 of the virtual machine 11 via the switch 30. The adapter 14 (internal input / output mechanism 141) inputs the input data 301 from the external system 40 to the system unit 13.

  The log acquisition mechanism 32 in the communication recording device 31 provided in the switch 30 acquires the input data 301 as a communication log when the input data 301 passes through the switch 30. The log storage processing mechanism 34 in the communication recording device 31 stores the input data 301 acquired by the log acquisition mechanism 32 in the log storage unit 33 while maintaining the time series. When storing the input data, the log storage processing mechanism 34 in the present embodiment attaches time information indicating the time (communication time) at that time to the input data.

  When the system unit 13 of the virtual machine 11 receives the input data 301 input by the adapter 14 (internal input / output mechanism 141), the system unit 13 performs processing based on the input data 301 and addresses the external system 40 as the processing result. Is output to the adapter 14 (internal input / output mechanism 141). The adapter 14 (internal input / output mechanism 141) sends the output data 302 to the network 20 in order to send the output data 302 to the external system 40. The output data 302 sent to the external system 40 sent over the network 20 is sent to the external system 40 via the switch 30.

  The log acquisition mechanism 32 in the communication recording device 31 provided in the switch 30 acquires the output data 302 as a communication log when the output data 302 from the system unit 13 of the virtual computer 11 passes through the switch 30. To do. The log storage processing mechanism 34 in the communication recording device 31 stores the output data 302 acquired by the log acquisition mechanism 32 in the log storage unit 33 while maintaining the time series.

  In this way, data input / output by communication between the system unit 13 of the virtual computer 11 and the external system 40 is used as a communication log, and the log storage unit 33 in the communication recording device 31 provided in the switch 30. Are stored (recorded) in chronological order.

  FIG. 4A shows an example of a log including a column of input / output data stored in the log storage unit 33 in chronological order. In FIG. 4A, I1, I2, I3... Are input data, and O1, O2, O3... Are output data corresponding to the input data I1, I2, I3. In the example of FIG. 4A, the input / output data columns in the log are in the order of I1, O1, I2, O2, I3, O3,. However, Oi corresponding to Ii is not always output after Ii (i = 1, 2, 3,...). The input data I1, I2, I3,... Have time information indicating the times t21, t22, t23,.

  By using a communication log stored in the log storage unit 33 in chronological order as shown in FIG. 4A, the virtual machine 11 can be in the state 205 (FIG. 2) at the above-described time t2 + Δt as described below. Reference) can be restored.

  Next, as a result of the failure of the physical computer 10-1 at the time t2 + Δt after the snapshot 204 of the virtual computer 11 at the time t2 is acquired, the virtual computer 11 (the system unit 13 thereof) functions normally. Suppose that it is no longer possible. In this case, a virtual machine restoration process (system restoration process) for restoring the virtual machine 11 to the state of time (failure occurrence time) t2 + Δt is performed. The state in which the virtual computer 11 (the system unit 13 thereof) cannot function normally occurs when a failure occurs in one or both of the system unit 13 and the hypervisor 12-1 of the virtual computer 11. The failure of one or both of the system unit 13 and the hypervisor 12-1 occurs not only due to a failure of the physical computer 10-1, but also due to a software bug or the like.

  Hereinafter, the virtual machine restoration process will be described. First, it is assumed that the failure of the physical computer 10-1 has been recovered and the physical computer 10-1 has been restarted. Then, the hypervisor 12-1 is restarted on the physical computer 10-1. The hypervisor 12-1 regenerates the virtual machine 11 on the hypervisor 12-1. The control mechanism 50 requests the snapshot management mechanism 121 in the hypervisor 12-1 to restore the virtual machine 11 using the snapshot.

  Upon receiving a restoration request from the control mechanism 50, the snapshot management mechanism 121 functions as a snapshot restoration unit, and the latest snapshot stored in the snapshot storage unit 122, that is, the snapshot acquired at time t2. Based on 204 (see FIG. 2), the virtual machine 11 is restored to the state at time t2 including the system unit 13 of the virtual machine 11. That is, the snapshot management mechanism 121 restores the virtual machine 11 in the state at the time t2 (first time point) when the snapshot is acquired. Instead of the snapshot management mechanism 121, the above-described snapshot acquisition means (snapshot acquisition mechanism) and snapshot restoration means (snapshot restoration mechanism) can be provided individually.

  Now, the virtual machine 11 restored by the snapshot management mechanism 121 has the system unit 13 in the state at time t2. However, the external system 40 is already in a state of time t2 + Δt, which is advanced by Δt from time t2. For this reason, the external system 40 communicates with the virtual computer 11 until just before the failure of the physical computer 10-1 in spite of the fact that most of the external system 40 is restored to the state at time t2. In some cases, communication cannot be performed normally. That is, the external system 40 cannot continue processing.

  Therefore, in the present embodiment, by using the log recorded by the communication recording device 31 in the switch 30, the virtual computer 11 restored to the state at time t2 can be returned to the state of t2 + Δt. Therefore, when the virtual computer 11 in the state at time t2 is restored, the log transmission mechanism 35 in the communication recording device 31 transfers the log stored in the log storage unit 33 to the adapter 14 of the virtual computer 11 via the network 20. Send through.

  The log transmitted from the log transmission mechanism 35 to the adapter 14 is received by the input / output mechanism 141 in the adapter 14 and stored in the log storage unit 142. When the input / output mechanism 141 receives the log transmitted from the log transmission mechanism 35, the input / output mechanism 141 passes the log to the log storage processing mechanism 143. Then, the log storage processing mechanism 143 stores the log passed from the input / output mechanism 141 in the log storage unit 142. The log reproduction mechanism 144 in the adapter 14 re-enters the input data included in the log stored in the log storage unit 142 into the system unit 13 of the virtual machine 11 in chronological order. Re-inputting input data included in the log into the system unit 13 in time series order is called log reproduction.

  In the present embodiment, all log data (input / output data) stored in the log storage unit 33 of the communication recording device 31 in the switch 30 is deleted when the snapshot of the virtual machine 11 is acquired. Therefore, when the virtual machine 11 is restored to the state at the time t2 based on the snapshot at the time t2, the input data re-entered in time-series order is input acquired after the time t2 (after the time t2). It becomes data. Assume that the input data after time t2 are I1, I2, I3,... Included in the log shown in FIG.

  In the present embodiment, in order to improve the accuracy when the virtual machine 11 is restored to the state of t2 + Δt, the input data Ii (i = 1, 2, 3,...) Included in the log after the time t2 is In accordance with the time information attached to the input data Ii, the input is re-input at the same input timing as that input earlier, that is, while keeping the previous input timing. As a result, the state of the virtual machine 11 during the period up to time t2 + Δt can be reproduced with high accuracy. However, it is not always necessary to observe the input timing. In this case, time information may not necessarily be attached to the input data Ii.

  The system unit 13 of the virtual computer 11 processes the input data Ii when the input data Ii included in the log is re-entered into the system unit 13 by the log reproduction mechanism 144. As a result, the system unit 13 transitions from the state when the input data Ii is input to a new state determined by the input data Ii, and outputs the output data Oi ′ for the input data Ii. For this reason, the virtual computer 11 restored to the state at time t (first time point) is restored to time t2 + Δt by re-inputting the input data Ii after time t2 included in the log into the system unit 13 in time series order. It is possible to advance to the state of (second time point).

  Here, if it can be confirmed that the virtual machine 11 has been restored to the state at time t2 + Δt, the accuracy of restoration of the virtual machine 11 can be further improved. Therefore, in this embodiment, the output data Oi ′ output from the system unit 13 is guided not only to the input / output mechanism 141 but also to the log comparison mechanism 145. The log comparison unit 145 outputs the output data Oi ′ output from the system unit 13 in response to the re-input of the input data Ii during the period during which the process of restoring the virtual computer 11 is performed (that is, the period of the restoration mode). When the input data Ii (input data Ii corresponding to the output data Oi) is input to the system unit 13 before the failure occurs, the output data included in the log stored in the The output data Oi is compared. On the other hand, during the restoration mode, the input / output mechanism 141 stops input / output and filters (blocks) the output data Oi ′ output from the system unit 13. The output data Oi corresponding to the input data Ii can be specified based on matching based on the transmission / reception destination address and port information and the analysis result of the communication session.

  When the output data Oi ′ and Oi match, the log comparison mechanism 145 sets the virtual machine 11 to the time state indicated by the time information attached to the input data Ii corresponding to the output data Oi in the log. Is determined to have been restored correctly. On the other hand, if the output data Oi ′ and Oi do not match, the log comparison mechanism 145 determines that the virtual computer 11 has not been correctly restored. In this way, the log comparison mechanism 145 monitors the reproduction status in the system unit 13 of the virtual machine 11. It is determined whether the virtual machine 11 has been correctly restored.

  When the log transmission mechanism 35 transmits the log stored in the log storage unit 33 to the adapter 14 of the virtual machine 11, the time series of the input / output data included in the log is changed to the time series of the input data. In order, the data may be processed into a set of input data and output data corresponding to the input data, and the processed log may be transmitted to the adapter 14. This processing may be performed on the adapter 14 side. FIG. 4B shows an example of the processed log when the log shown in FIG. 4A is processed.

Next, the flow of processing including snapshot acquisition and still point creation will be described with reference to the sequence chart of FIG.
The control mechanism 50 acquires the log included in the communication recording device 31 in the switch 30 when the time (for example, time t2) at which the snapshot of the virtual computer 11 operating on the physical computer 10-1 is to be acquired has arrived. A still point creation request is issued to the mechanism 32 (step 501). In response to this, the log acquisition mechanism 32 creates a quiesce point for snapshot acquisition by the snapshot management mechanism 121 (step 502). Here, time t2 is created as a stationary point. In this case, as will be described later, the log stored in the log storage unit 33 is deleted.

  When the log acquisition mechanism 32 creates a quiesce point at time t2 (step 502), the log acquisition mechanism 32 returns a response to the quiesce point creation request (response that snapshot acquisition has become possible due to the creation of the quiesce point) to the control mechanism 50 (step step). 503). Then, the control mechanism 50 issues an input / output stop request to the input / output mechanism 141 included in the adapter 14 of the virtual machine 11 (step 504). In response to this, the input / output mechanism 141 stops (freezes) input / output (communication) between the input / output mechanism 141 and the network 20 (step 505). Thereby, the quiesce point created by the log acquisition mechanism 32 is determined as a quiesce point for snapshot acquisition.

  When the input / output mechanism 141 stops the input / output to / from the network 20 (step 505), the input / output mechanism 141 returns a response to the input / output stop request to the control mechanism 50 (step 506). Then, the control mechanism 50 issues a snapshot acquisition request to the snapshot management mechanism 121 (step 507). In response to this, the snapshot management mechanism 121 acquires a snapshot of the virtual machine 11 (step 508). At this time, that is, after time t2, the input / output mechanism 141 is in a state where input / output is stopped (frozen). For this reason, the snapshot management mechanism 121 can acquire the snapshot of the virtual machine 11 at time t2 of the stationary point with high accuracy. As is apparent, the acquired snapshot is synchronized with the input / output of the input / output mechanism 141.

  If the snapshot management mechanism 121 can acquire a snapshot (step 508), it returns a response for notifying completion of snapshot acquisition to the control mechanism 50 (step 509). Then, the control mechanism 50 issues a stillness release request to the log acquisition mechanism 32 (step 510). In response to this, the log acquisition mechanism 32 deletes the log stored in the log storage unit 33, that is, the input / output data acquired before time t2, and releases the set still point (step 511). . Thereafter, if there is communication data (input / output data) passing through the switch 30, the log acquisition mechanism 32 can acquire the data as a log.

  When the log acquisition mechanism 32 releases the stationary point (step 511), it returns a response to the stationary cancellation request to the control mechanism 50 (step 512). Then, the control mechanism 50 issues an input / output restart request to the input / output mechanism 141 (step 513). In response to this, the input / output mechanism 141 releases the input / output suspension state and resumes input / output (step 514). The input / output mechanism 141 returns a response to the input / output restart request to the control mechanism 50 (step 515).

  As described above, the input / output stop / input / output restart of the input / output mechanism 141 is controlled according to the setting / cancellation of the stationary point in the communication recording device 31 and the snapshot is acquired according to the input / output stop of the input / output mechanism 141 By controlling this, it is possible to synchronize the acquired snapshot with the input / output of the input / output mechanism 141 with high accuracy. Thereby, the log reproduction mechanism 144 can cause the system unit 13 of the virtual computer 11 to perform highly accurate input / output reproduction without requiring a complicated configuration.

Next, the flow of the virtual machine restoration process described above will be described with reference to the sequence chart of FIG.
When it is necessary to restore the virtual computer 11 because a failure has occurred in the physical computer 10-1 at time t + Δt, the control mechanism 50 causes the snapshot management mechanism 121 to restore the virtual computer 11 by snapshot. A restoration request is issued (step 601). In response to this, the snapshot management mechanism 121 restores the virtual machine 11 at the time t2 based on the latest snapshot (the snapshot 204 acquired at the time t2) stored in the snapshot storage unit 122. (Step 602). Then, the snapshot management mechanism 121 returns a response to the restoration request to the control mechanism 50 (step 603).

Upon receiving the response from the snapshot management mechanism 121, the control mechanism 50 issues a log transmission request to the log transmission mechanism 35 in the communication recording device 31 (step 604). In response to this, the log transmission mechanism 35 transmits the log stored in the log storage unit 33 to the input / output mechanism 141 in the adapter 14 (step 605). When the input / output mechanism 141 receives the log transmitted from the log transmission mechanism 35, the input / output mechanism 141 stops input / output between the input / output mechanism 141 and the network 20 (step 606) and stores the log in the log storage processing mechanism 143. (Step 607). Then, the log storage processing mechanism 143 stores the log delivered from the input / output mechanism 141 in the log storage unit 142 (step 608), and notifies the input / output mechanism 141 of log storage completion (step 609). The input / output mechanism 141 notifies the log transmission mechanism 35 of the log storage completion notified from the log storage processing mechanism 143 (step 610), and the log transmission mechanism 35 completes transmission for the log transmission request in response to the log storage completion notification. A response is returned to the control mechanism 50 (step 611).
When receiving a transmission completion response from the log transmission mechanism 35 to the log transmission request, the control mechanism 50 issues a log reproduction and restoration request to the log reproduction mechanism 144 in the adapter 14 (step 612). When the log reproduction mechanism 144 receives a log reproduction and restoration request from the control mechanism 50, the log reproduction mechanism 144 converts the input data Ii (i = 1, 2, 3...) Included in the log stored in the log storage unit 142 into the virtual machine 11. Is re-entered into the system unit 13 in chronological order, thereby reproducing and restoring the log to reproduce the input / output state of the system unit 13 before the failure occurs (that is, the virtual machine 11 is sequentially restored to the state at time T2 + Δt. Log reproduction and restoration) (step 613).

  When the input data Ii is input by the log reproduction mechanism 144, the system unit 13 outputs output data Oi ′ for the input data Ii. Each time the output data Oi ′ for the input data Ii is output from the system unit 13, the log comparison mechanism 145 extracts the output data Oi corresponding to the input data Ii from the log stored in the log storage unit 142. Then, when the output data Oi corresponding to the input data Ii can be extracted, the log comparison mechanism 145 compares the output data Oi ′ for the input data Ii with the extracted output data Oi. Based on the comparison result, it is determined whether or not the virtual machine 11 has been successfully restored to the state at the time when the output data Oi corresponding to the input data Ii is output. If it is determined that the restoration cannot be performed normally, the restoration fails. In this case, for example, the restoration may be terminated while the communication is cut off, or the restoration may be performed at the time of the snapshot.

  For all input data Ii included in the log, the log comparison mechanism 145 determines that if the output data Oi ′ from the system unit 13 corresponding to the input data Ii matches the corresponding output data Oi in the log, It is determined that the state of the computer 11 has been successfully restored until the failure occurrence time t2 + Δt. Then, the log comparison mechanism 145 notifies the log reproduction mechanism 144 of the completion of restoration.

  When the log comparison mechanism 145 is notified of the restoration completion from the log comparison mechanism 145, that is, when the log comparison mechanism 145 determines that the state of the virtual machine 11 has been successfully restored until the failure occurrence time t2 + Δt, the log comparison mechanism 145 has not been restored as in this example. If the input data Ii to be input is not included in the log, a response for notifying the completion of restoration in response to the log reproduction and restoration request from the control mechanism 50 is returned to the control mechanism 50 (step 614).

  Then, the control mechanism 50 issues an input / output restart request to the input / output mechanism 141 (step 615). In response to this, the input / output mechanism 141 releases the input / output stop state and resumes input / output (step 616). The input / output mechanism 141 returns a response to the input / output restart request to the control mechanism 50 (step 617).

  By the way, even if the output data Oi ′ corresponding to the input data Ii is output from the system unit 13, the output data Oi corresponding to the input data Ii is not included in the log. Sometimes it cannot be taken out. In this case, the log comparison mechanism 145 determines that a failure has occurred before the output data Oi is output from the system unit 13. In the log comparison mechanism 145, the output data in the log for which the most recent match with the output data from the system unit 13 is the output data immediately before the occurrence of the failure, and therefore the state of the virtual machine 11 is determined as the failure occurrence time t2 + Δt It is determined that it has been successfully restored. Then, the log comparison mechanism 145 notifies the log reproduction mechanism 144 of the completion of restoration.

  When the log reproduction mechanism 144 is notified of the restoration completion from the log comparison mechanism 145, the log reproduction mechanism 144 notifies the completion of restoration in response to the log reproduction and restoration request from the control mechanism 50 even if the input data Ii that has not been input is included in the log. Is returned to the control mechanism 50 (step 614). Then, steps 615 and 616 are performed, and input / output by the input / output mechanism 141 is resumed. At this time, if the input data Ii that has not been input is included in the log, the log reproduction mechanism 144 re-inputs the input data Ii that has not been input to the system unit 13.

  When the input data Ii (that is, input data Ii whose corresponding output data Oi is not included in the log) is re-entered into the system unit 13 by the log reproduction mechanism 144 after the restoration is completed, the input data Ii Output data Oi output from the system unit 13 is sent from the input / output mechanism 141 to the external system 40 via the network 20 and the switch 30. Here, when the output data Oi passes through the switch 30, the output data Oi is acquired by the log acquisition mechanism 32 and added to the log storage unit 33 by the log storage processing mechanism 34.

[First Modification]
Next, a first modification of the above embodiment will be described.
In the above embodiment, the control mechanism 50 directly controls the input / output mechanism 141 and the log reproduction mechanism 144. The feature of the first modification is that the log acquisition mechanism 32 is controlled by the log acquisition mechanism 32 in the processing including snapshot acquisition and quiesce point creation, and the control of the input / output mechanism 141 in the virtual machine restoration processing is the log transmission mechanism. 35 and the log reproduction mechanism 144, and the log transmission mechanism 35 is controlled by the log reproduction mechanism 144 in the virtual machine restoration process.

  Hereinafter, with reference to the sequence charts of FIG. 7 and FIG. 8, the above-described implementation of the processing flow including snapshot acquisition and creation of a quiesce point and the flow of virtual machine restoration processing applied in the first modification example will be described. A brief description will be given centering on differences from the embodiment (sequence charts of FIGS. 5 and 6). 7 and 8, the same reference numerals are assigned to the same parts as those in FIGS. 5 and 6.

  First, it is assumed that a request for creating a quiesce point is issued from the control mechanism 50 to the log obtaining mechanism 32 in the process including snapshot acquisition and quiesce point creation applied in the first modification (step 501). Then, the log acquisition mechanism 32 creates a quiesce point for snapshot acquisition (step 502), and issues an input / output stop request to the input / output mechanism 141 instead of the control mechanism 50 (step 701). That is, the log acquisition mechanism 32 propagates the still point creation request from the control mechanism 50 to the input / output mechanism 141 as an input / output stop request. The input / output mechanism 141 stops input / output with the network 20 in response to the input / output stop request from the log acquisition mechanism 32 (step 505). Thereby, the set still point is determined as a still point for acquiring a snapshot.

  When the input / output mechanism 141 stops the input / output (step 505), the input / output mechanism 141 returns a response to the input / output stop request to the log acquisition mechanism 32 (step 702). In response to this, the log acquisition mechanism 32 returns a response to the previous still point creation request to the control mechanism 50 (step 503). Then, the control mechanism 50 requests the snapshot management mechanism 121 to acquire a snapshot (step 507). When the snapshot management mechanism 121 acquires a snapshot (step 508), the control mechanism 50 receives the snapshot management mechanism 121 from the snapshot management mechanism 121. A response is returned to (step 509).

  In response to a response from the snapshot management mechanism 121, the control mechanism 50 issues a stillness release request to the log acquisition mechanism 32 (step 510). In response to this, the log acquisition mechanism 32 sets the set still point. Release (step 511). The log acquisition mechanism 32 issues an input / output restart request to the input / output mechanism 141 instead of the control mechanism 50 (step 703). That is, the log acquisition mechanism 32 propagates the static release request from the control mechanism 50 to the input / output mechanism 141 as an input / output restart request.

  In response to the input / output restart request from the log acquisition mechanism 32, the input / output mechanism 141 releases the input / output stop state and restarts input / output (step 514). The input / output mechanism 141 returns a response to the input / output restart request from the log acquisition mechanism 32 to the log acquisition mechanism 32 (step 704). When the response to the input / output restart request is returned from the input / output mechanism 141, the log acquisition mechanism 32 returns the response to the previous static release request to the control mechanism 50 (step 512).

  Next, in the virtual machine restoration process applied in the first modified example, a restoration request for restoring the virtual machine 11 by a snapshot is issued from the control mechanism 50 to the snapshot management mechanism 121. (Step 601). Then, the snapshot management mechanism 121 restores the virtual machine 11 at the time t2 based on, for example, the snapshot 204 acquired at the time t2 stored in the snapshot storage unit 122 (step 602). A response to the restoration request is returned to the control mechanism 50 (step 603).

  Upon receiving the response from the snapshot management mechanism 121, the control mechanism 50 issues a log restoration request to the log transmission mechanism 35 (step 801). In response to this, the log transmission mechanism 35 transmits the log stored in the log storage unit 33 to the input / output mechanism 141 (step 605). Then, the input / output mechanism 141 stops input / output (step 606). The log transmitted to the input / output mechanism 141 is transferred to the log storage processing mechanism 143 and stored in the log storage unit 142 by the log storage processing mechanism 143 (steps 607 and 608). Then, the log storage processing mechanism 143 notifies the input / output mechanism 141 of the log storage completion (step 609), and the log storage completion is further propagated from the input / output mechanism 141 to the log transmission mechanism 35 (step 610).

  When receiving the log storage completion notification propagated from the input / output mechanism 141, the log transmission mechanism 35 issues a log reproduction and restoration request to the log reproduction mechanism 144 in place of the control mechanism 50 (step 802). When the log reproduction mechanism 144 receives the log reproduction and restoration request from the log transmission mechanism 35, based on the log stored in the log storage unit 142, the input / output state (here, time T2 + Δt) of the system unit 13 before the failure occurs. Log reproduction and restoration for reproducing (state) (step 613).

  The log reproduction mechanism 144 issues an input / output restart request to the input / output mechanism 141 when the state of the virtual machine 11 can be normally restored until time t2 + Δt (step 803). In response to this, the input / output mechanism 141 releases the input / output stop state and resumes input / output (step 616). The input / output mechanism 141 returns a response to the input / output restart request to the log reproduction mechanism 144 (step 804).

  When the log reproduction mechanism 144 receives a response to the input / output restart request from the input / output mechanism 141, it returns a response to the previous log reproduction and restoration request to the log transmission mechanism 35 (step 805). When receiving a response to the log reproduction and restoration request from the log reproduction mechanism 144, the log transmission mechanism 35 returns a response to the restoration request based on the previous log to the control mechanism 50 (step 806).

[Second Modification]
Next, a second modification of the above embodiment will be described.
In the above embodiment, all input data Ii included in the log transmitted by the log transmission mechanism 35 is input to the system unit 13 of the virtual machine 11 by the log reproduction mechanism 144. However, when a plurality of applications operate in the system unit 13, in order to restore the virtual machine 11 at high speed, it may be necessary to restore only a state in which a specific application among the plurality of applications is operating. That is, partial restoration of the state of the virtual machine 11 may be requested. In such a case, it is useless to re-input all input data Ii included in the log. Therefore, the feature of the second modification is that only the input data passed to a specific application is re-input among the input data Ii included in the log.

  FIG. 9 is a block diagram showing a configuration of the log reproduction mechanism 144 in the adapter 14 of the virtual machine 11. The log reproduction mechanism 144 includes a filter mechanism 144a. The filter mechanism 144a narrows down input data to be reproduced in the log in accordance with, for example, the filter setting 144b from the control mechanism 50. Note that the filter mechanism 144a may be provided outside the log reproduction mechanism 144.

  In the present embodiment, the filter setting 144b is a port assigned to the specific application that operates on the virtual computer 11 (system unit 13 thereof), and is between the system unit 13 of the virtual computer 11 and the specific application. Contains the port number of the port used to send and receive input / output data. If this port number is PNs, among the input data transmitted from the external system 40, the input data addressed to the specific application on the virtual machine 11 is the IP address assigned to the virtual machine 11 as the destination IP address. And PNs as the destination port number.

  Therefore, the filter mechanism 144a selects only the input data whose destination port number is PNs from the input data addressed to the virtual machine 11 included in the log. The log reproduction mechanism 144 sequentially inputs the input data selected by the filter mechanism 144 a to the system unit 13 of the virtual computer 11. Thereby, the input data input to the system unit 13 of the virtual machine 11 can be narrowed down to input data to be passed to the specific application. Further, by narrowing down the input data input to the system unit 13 of the virtual computer 11, it is possible to narrow down the output data output from the system unit 13, that is, the output data to be compared by the log comparison mechanism 145. Therefore, according to the second modification, partial restoration of the state of the virtual machine 11 can be performed efficiently.

[Third Modification]
Next, a third modification of the above embodiment will be described.
FIG. 10 is a block diagram showing a configuration of the log comparison mechanism 145 in the adapter 14 of the virtual machine 11. The third modification differs from the second modification in that, instead of providing the filter mechanism 144a in the log reproduction mechanism 144, as shown in FIG. 10, a filter mechanism 145a corresponding to the filter mechanism 144a is This is provided in the log comparison mechanism 145. The filter mechanism 145a may be provided outside the log comparison mechanism 145.

  In the third modification, the log reproduction mechanism 144 sequentially inputs all input data included in the log stored in the log storage unit 142 to the system unit 13 as in the above embodiment. Then, the system unit 13 outputs output data corresponding to the input data input by the log reproduction mechanism 144.

  In the third modification, the filter mechanism 145a narrows down the output data to be compared by the log comparison mechanism 145 in accordance with the filter setting 145b from the control mechanism 50 among the output data output from the system unit 13, for example. Here, assuming that the port number PNs is included in the filter setting 145b, only output data from a specific application whose transmission source port number is PNs is selected as output data to be compared by the log comparison mechanism 145. .

  In the third modification, output data to be compared by the log comparison mechanism 145 is narrowed down to output data from a specific application. For this reason, the load on the log comparison mechanism 145 can be reduced. In addition, since all input data included in the log stored in the log storage unit 142 is input to the system unit 13, the virtual computer 11 including the system unit 13 can be restored with high accuracy. .

  In addition, this invention is not limited to the said embodiment or its modification example as it is, A component can be deform | transformed and embodied in the range which does not deviate from the summary in an implementation stage. For example, in the above embodiment and its modification, after a failure occurs in the physical computer 10-1, the failure of the physical computer 10-1 is recovered, the physical computer 10-1 is restarted, and the physical computer 10- When the hypervisor 12-1 is restarted on 1, the virtual machine 11 that has been operating until the time of the failure is restored on the hypervisor 12-1. However, if the physical computers 10-1 and 10-2 constitute, for example, a cluster system, the physical computer 10-1 provides the external system 40 when a failure occurs in the physical computer 10-1. In order to allow the physical computer 10-2 to continue the service that has been performed, the virtual computer 11 can be restored on the hypervisor 12-2 operating on the physical computer 10-2.

  In the above embodiment and its modification, the control mechanism 50 is provided independently of the physical computers 10-1 and 10-2. However, a control mechanism corresponding to the control mechanism 50 is provided in each of the hypervisors 12-1 and 12-2 operating on the physical computers 10-1 and 10-2, for example, the physical computers 10-1 and 10-2. It may be provided. In this case, the control mechanism provided in the hypervisor 12-1 or 12-2 to which the virtual computer 11 should be restored may perform the same control as the control mechanism 50.

  In addition, the log reproduction mechanism 144 may be provided with the filter mechanism 144a and the log comparison mechanism 145 may be provided with the filter mechanism 145a by combining the second and third modifications.

  In addition, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment or its modification. For example, you may delete a some component from all the components shown by embodiment or its modification.

1 is a block diagram showing a configuration of a virtual machine system according to an embodiment of the present invention. The figure which shows the relationship between the state of a virtual machine, snapshot time, and a snapshot. The figure which shows the flow of the data (input data) at the time of communication (input) from the external system to the virtual machine, and the flow of data (output data) at the time of communication (output) from the virtual machine to the external system for the input data . The figure which shows the example of the log containing the row | line | column of the input-output data stored in time series in the log storage part, and the example of the processed log when the said log is processed. The sequence chart which shows the flow of a process including the snapshot acquisition applied to the embodiment, and a still point creation. The sequence chart which shows the flow of the virtual machine restoration process applied in the embodiment. The sequence chart which shows the flow of a process including the snapshot acquisition applied in the 1st modification of the embodiment, and still point creation. The sequence chart which shows the flow of the virtual machine restoration process applied in the 1st modification of the embodiment. The block diagram which shows the structure of the log reproduction | regeneration mechanism applied in the 2nd modification of the embodiment. A block diagram showing a configuration of a log comparison mechanism applied in a third modification of the embodiment

Explanation of symbols

  10-1 and 10-2 ... physical computers, 11 ... virtual computers, 12-1,12-2 ... hypervisor, 13 ... system unit, 14 ... adapter, 20,21 ... network, 30 ... switch, 31 ... communication record Device 32... Log acquisition mechanism 33... Log storage unit 34... Log storage processing mechanism 35 35 log transmission mechanism 40 external system 50 control mechanism 141 input / output mechanism 142 log storage unit 143 ... log storage processing mechanism, 144 ... log reproduction mechanism, 145 ... log comparison mechanism.

Claims (5)

A physical computer,
Snapshot storage means used for storing a snapshot of a virtual machine operating on the physical computer or another physical computer connected to the physical computer via a network;
Communication recording means for recording communication data input / output between the external system using the virtual machine and the virtual machine as a communication log in chronological order, and
The physical computer is
When the virtual machine operates on the physical computer, a snapshot acquisition unit that acquires the state of the virtual machine as a snapshot of the virtual computer and stores the snapshot in the snapshot storage unit;
The virtual machine is restored on the physical computer after a failure that prevents the virtual machine from functioning normally while the virtual machine is operating on the physical computer or the other physical computer. Snap that restores the virtual machine to the first point in time when the snapshot was acquired based on the latest snapshot of the virtual machine stored in the snapshot storage means Shot restoration means,
From the first time point included in the communication log recorded by the communication recording unit to the second time point when the failure occurs, in the virtual machine restored at the first time point And a log reproducing unit for restoring the virtual machine to the second time point by inputting input data in time series.   The physical computer, when the log reproduction means inputs the input data included in the communication log into the virtual computer, outputs the output data output from the virtual computer as a result of the virtual computer for the input data By comparing the input data included in the communication log with the output data output from the virtual machine when the input data is input to the virtual machine before the failure occurs, 2. The virtual computer system according to claim 1, further comprising log comparison means for determining whether or not the state of the computer is correctly restored.   Control means for synchronizing the snapshot and the recorded input / output by controlling the acquisition of the snapshot by the snapshot acquiring means and the recording of the input / output communication data by the communication recording means. The virtual machine system according to claim 1, further comprising: The input data includes a port number of a port assigned to an application that operates on the virtual machine and processes the input data as a destination port number;
The log reproduction means restores the virtual machine to the second time point at least with respect to the operation of a predetermined specific application among the plurality of applications when a plurality of applications operate on the virtual machine. Therefore, from the time series of input data included in the communication log, only input data including a port number of a port assigned to the specific application as a destination port number is input to be input to the virtual machine. The virtual computer system according to any one of claims 1 to 3, further comprising filter means for selecting data. Utilizing a physical computer, snapshot storage means used to store a virtual computer operating on the physical computer or another physical computer connected to the physical computer via a network, and the virtual computer A virtual machine restoration method applied to a virtual machine system comprising a communication recording means for recording communication data input / output between the external system and the virtual machine in time series as a communication log,
When the virtual machine operates on the physical computer, the physical computer acquires the state of the virtual machine as a snapshot of the virtual machine, and stores the snapshot in the snapshot storage unit;
The virtual machine is restored on the physical computer after a failure that prevents the virtual machine from functioning normally while the virtual machine is operating on the physical computer or the other physical computer. When the state becomes possible, the physical computer is based on the latest snapshot of the virtual computer stored in the snapshot storage unit, and the virtual computer has the first snapshot from which the snapshot has been acquired. A step to restore to the point in time,
From the first time point included in the communication log recorded by the communication recording unit to the second time point when the failure occurs, in the virtual machine restored at the first time point And a step of restoring the virtual computer to the second time point by inputting the input data in time series by the physical computer.

Images