JP2015132979A - Information processing device and control method thereof, control program and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information processing device capable of reducing the load of collecting a memory dump.SOLUTION: A guest driver 121 notifies a host OS 11 of the occurrence of a fault by generating an interrupt in a virtual bus 21 allocated to a virtual machine 124 in which a guest OS 12 with the fault occurring therein operates. The host OS 11 acquires process identification information from a virtualization ID/process list 117 on the basis of a virtualization ID of the virtual bus 21 with the interrupt generated, acquires the virtual machine 124 in which the guest OS 12 with the fault occurring operates from a virtual machine/process list 118 on the basis of the process identification information, and collects contents of a memory 123 in a storage state storage area 114.

Description

  The present invention relates to an information processing apparatus, a control method thereof, a control program, and a recording medium.

  In an information processing system in which a plurality of virtual machines operate, for example, a virtual machine (virtual machine) is executed as a process of a host OS (operating system), and the guest OS operates on the virtual machine. When a crash occurs in the guest OS and processing stops, the contents of the memory and log information are acquired, and the cause of the crash is investigated.

  For example, in a virtual machine system, the host virtual machine has a guest storage definition for prescribing a save position in a log unit used by the host virtual machine, which is a save position for saving system information about the guest virtual machine. The monitor monitors the guest virtual machine and detects a failure that occurred in the guest virtual machine. When a failure in the guest virtual machine is detected by the panic monitoring manager, based on the guest storage definition, It has been proposed to include a log collection unit that collects system information from the retreat position.

JP 2010-086181 A

  For example, if the memory of a guest virtual machine has a capacity of 2 GB, in order to collect and store a memory dump, 2 GB is used as a storage area (hereinafter referred to as a disk area) in a storage device such as a magnetic disk device. It is necessary to assign to the virtual machine in advance. Therefore, for example, if there are 10 virtual machines, if a memory dump is collected for all 10 virtual machines, a disk area of 20 GB is required.

  However, the storage device used in the information processing system is expensive. Therefore, acquiring a large amount of disk area in preparation for a guest OS crash that cannot be predicted is not preferable from the viewpoint of memory dump collection burden (cost). Therefore, if the disk area for collecting the memory dump is reduced, even if a crash actually occurs, the memory dump may not be collected. As a result, the cause investigation becomes longer.

  An object of the present invention, according to one aspect, is to provide an information processing apparatus that can reduce the burden of collecting a memory dump.

  According to one aspect, the information processing apparatus is an information processing apparatus on which a plurality of virtual machines operate, and includes a first storage unit, a second storage unit, and a storage unit. The first storage means includes a plurality of virtual bus identification information for identifying a plurality of virtual buses connecting the virtual machine control unit for controlling the plurality of virtual machines and the plurality of virtual machines, and the plurality of virtual machines. First correspondence information that is a correspondence relationship with a plurality of process identification information for identifying processes of the corresponding virtual machine control units is stored. The second storage means stores second correspondence information that is a correspondence relationship between the plurality of process identification information and the plurality of virtual machine identification information for identifying the plurality of virtual machines. The storage means copes with the failure of the first OS on the first virtual bus corresponding to the first virtual machine from the first OS operating on the first virtual machine included in the plurality of virtual machines. When there is an interrupt, the first virtual machine is identified based on the first correspondence information and the second correspondence information, and the memory area used by the first OS operating on the first virtual machine Save the contents.

  According to one aspect, the information processing apparatus can reduce the burden of collecting a memory dump.

It is a figure which shows an example of a structure of an information processing system. It is a figure showing an example of composition of an information processing system It is explanatory drawing of an information processing system. It is explanatory drawing of an information processing system. It is an identification information creation processing flow. It is an identification information creation processing flow. It is an identification information creation processing flow. It is an identification information creation processing flow. It is a dump collection processing flow. It is a dump collection processing flow. It is a dump collection processing flow.

  FIG. 1 is a diagram illustrating an example of a configuration of an information processing system.

  A system (information processing system) on which a plurality of virtual machines operate is an information processing apparatus including a plurality of OSs (operating systems) 1, a hypervisor 2, and hardware 3, in other words, a physical computer. The plurality of OSs 1 include, for example, one host OS 11 and a plurality of guest OSs 12. For example, the host OS 11 operates on a virtual CPU 119 described later with reference to FIG. The guest OS 12 operates on, for example, a virtual machine 124 described later with reference to FIG.

  The hardware 3 includes a physical device, for example, a physical CPU, a physical memory, a physical I / O device, in other words, a physical input / output device. The physical memory is a main storage device (main memory), and the physical I / O device is a peripheral storage device such as a magnetic disk device.

  The hypervisor 2 virtualizes physical devices such as a physical CPU, a physical memory, and a physical I / O device included in the hardware 3. The hypervisor 2 transfers a virtualized physical device, in other words, a request for the virtual device to the physical device, and transfers a response from the physical device to the virtual device. As a result, the hypervisor 2 configures the virtual machine 124.

  FIG. 2 is a diagram illustrating an example of the configuration of the information processing system. In FIG. 2, the information processing system includes a plurality of virtual machines 124 and a plurality of guest OS 12, but only one virtual machine 124 and one guest OS 12 are shown.

  The information processing system includes one host OS 11 and a plurality of virtual machines 124. The host OS 11 is a virtual machine control unit and controls a plurality of virtual machines 124. On each virtual machine 124, the guest OS 12 operates.

  The host OS 11 operates on the virtual CPU 119 and includes a dump collection tool 111, a dump collection service unit 112, a host driver 113, a saved state storage area 114, and a dump storage area 115. In the example of FIG. 1, the dump collection tool 111 includes a dump collection service unit 112 and a host driver 113. The dump collection tool 111 may be provided separately from the dump collection service unit 112 and the host driver 113. The guest OS 12 operates on the virtual machine 124 and includes a guest driver 121. The virtual machine 124 includes a virtual device 122, a memory 123, and a virtual CPU 125. The hypervisor 2 includes a virtual bus 21.

  The host OS 11 is an operating system or a control program, and calls the hypervisor 2 to allocate resources such as a virtualized CPU, memory, and I / O device to the virtual machine 124. The host OS 11 is a virtual machine control unit that controls execution of the virtual machine 124. The host OS 11 operates on a host machine, for example.

  The host machine is, for example, a privileged virtual machine that exists only on the hypervisor 2. A privileged virtual machine that is a host machine can directly access a physical device, unlike other virtual machines 124, in other words, a virtual machine 124 in which the guest OS 12 operates. Therefore, the privileged virtual machine that is the host machine is completely different from the virtual machine 124 in which the guest OS 12 operates.

  The host machine may be a physical machine itself, for example. In addition, the host OS 11, in other words, the virtual machine control unit that controls the execution of the virtual machine 124 does not need to operate on the virtual machine, and does not operate on the control program that operates on the hypervisor 2 or on the virtual CPU 119. It may be a control program that runs on the CPU.

  From the above, in FIG. 2, illustration of a host machine on which the host OS 11 that is a virtual machine control unit operates is omitted. On the other hand, the host OS 11 that is the virtual machine control unit is illustrated as operating on the virtual CPU 119 that is virtualized by the hypervisor 2, but the host OS 11 is not the virtual CPU 119, as described above. It may operate directly on a physical CPU or may operate on a logical CPU corresponding to the physical CPU.

  The virtual machine 124 is configured as a virtual device assigned by the host OS 11, in other words, a guest machine. The virtual machine 124 is executed as a process on the host OS 11, specifically, as a child process. The guest OS 12 operates on the virtual machine 124. Note that the guest OS 12 does not have to exist on the virtual machine 124.

  The guest OS 12 operates on the operating virtual machine 124, in other words, the virtual CPU 125 and the virtualized memory 123, and can access a virtualized I / O device, for example, the virtual device 122. . The guest OS 12 transfers access to the virtual device to the physical device via the host OS 11 or the hypervisor 2, and the access result is returned from the physical device.

  In the guest OS 12, the guest driver 121 is a notification unit that notifies the host OS 11 of the occurrence of a failure, and is provided in each of the plurality of guest OSs 12. When a failure occurs in the guest OS 12 to which the guest driver 121 belongs, in other words, in the corresponding guest OS 12, the virtual bus 21 assigned to the host machine 11 and the virtual machine 124 in which the failed guest OS 12 operates. The occurrence of a fault is notified by generating an interrupt. In other words, when a failure occurs in the guest OS 12 that operates on the virtual machine 124, the virtual machine 124 notifies the occurrence of the failure by generating an interrupt to the virtual bus 21 assigned to itself. The guest driver 121 is provided in each of the plurality of guest OSs 12. The guest driver 121 is a kernel driver, for example, and is provided in the guest OS 12.

  Specifically, the guest driver 121 is registered in the guest OS 12 as a panic exit of the guest OS 12. For example, the guest driver 121 is called when the guest OS 12 is activated, and is incorporated in a predetermined position of the guest OS 12, specifically, a panic exit. The guest driver 121 is called from the guest OS 12 when a crash occurs in the guest OS 12, and generates an interrupt on the virtual bus 21 connecting the virtual machine and the hardware.

  The virtual device 122 is a storage device provided in each of the plurality of virtual machines 124. The virtual device 122 stores process identification information for identifying the process when the virtual machine 124 to which the virtual device 122 is connected is executed as a process of the host OS 11. In other words, the virtual device 122 stores process identification information that identifies the process of the host OS 11 that is the virtual machine control unit corresponding to the virtual machine 124 to which the virtual device 122 is connected.

  For example, the process identification information is written to the virtual device 122 by the dump collection service unit 112 of the host OS 11. The process identification information is written, for example, when the virtual machine 124 is created. This is because the process identification information is determined corresponding to the virtual machine 124.

  Actually, as the virtual device 122, for example, a DVD (Digital Versatile Disc) apparatus is used. For example, the virtual device 122 is dynamically incorporated into the virtual machine 124. In other words, the virtual device 122 may be a storage device that can be dynamically incorporated into the virtual machine 124. Thereby, according to the generation of the virtual machine 124, the process identification information of the process corresponding to the generated virtual machine 124 is written in the virtual device 122, and the process is stored in the guest driver 121 of the guest OS 12 operating on the generated virtual machine 124. Identification information can be notified.

  For example, a virtual DVD in which process identification information is written is created, and a virtualized DVD reader is created. The virtual DVD reading device into which the virtual DVD is inserted is connected as a virtual device 122 to the virtual machine 124 to be activated.

  The virtual device 122 is not limited to a DVD device, and other various storage devices can be used. For example, a device or resource that can be added to the definition of the virtual machine 124, data can be written to the virtual device 122 of the processed virtual machine 124, and the guest driver 121 can be written from the virtual device 122. Any device that can read data is acceptable.

  The memory 123 is a part of the main memory included in the hardware 3. The memory 123 is a memory area that is virtualized and included in each of the plurality of virtual machines 124 and is used by the guest OS 12.

  In the hypervisor 2, the virtual bus 21 is a virtual interrupt signal line that logically connects each of the plurality of virtual machines 124 and the physical CPU, for example. When a failure occurs, the guest OS 12 causes the virtual CPU of the virtual machine 124 in which the guest OS 12 is operating to execute an instruction that cannot be executed. As a result, an execution exception occurs in the virtual machine 124 in which the guest OS 12 in which the failure has occurred operates, so that the physical CPU eventually captures an interrupt to the virtual bus 21. Therefore, the virtual bus 21 generates an interrupt from the virtual machine 124 executing the guest OS 12 to the physical CPU. On the other hand, as described above, the host OS 11 operates on a host machine that is a privileged virtual machine or physical machine, and therefore processes an interrupt to the virtual bus 21 captured by the physical CPU. As a result, the virtual bus 21 logically connects each of the plurality of virtual machines 124 and the host OS 11. In other words, the plurality of virtual buses 21 connect the virtual machine control unit and each of the plurality of virtual machines. The plurality of virtual buses 21 are assigned one by one to each of the plurality of virtual machines 124 by the hypervisor 2 and given virtual bus identification information. The interrupt to the virtual bus 21 is an interrupt to the physical interrupt signal line to the physical CPU, in other words, the same priority as the hardware interrupt, in other words, the highest priority interrupt.

  When a failure occurs in the guest OS 12, the host OS 11 serving as the virtual machine control unit stores the contents of the memory 123 used by the guest OS 12 based on first correspondence information and second correspondence information described later.

  For example, it is assumed that a failure has occurred in the guest OS 12 operating on a certain virtual machine 124 included in the plurality of virtual machines 124. The virtual machine 124 on which the guest OS 12 in which the failure has occurred operates is referred to as the first virtual machine 124.

  Each of the plurality of virtual machines 124, when a failure occurs in the corresponding guest OS 12, the virtual machine 124 in which the failed guest OS 12 operates, in other words, the virtual bus allocated to the first virtual machine 124. By causing 21 to generate an interrupt, the occurrence of a failure is notified to the host OS 11 which is the virtual machine control unit.

  The host OS 11 receives an interrupt corresponding to the failure of the guest OS 12 from the guest OS 12 operating in the first virtual machine 124 included in the plurality of virtual machines 124 to the first virtual bus 21 corresponding to the first virtual machine 124. If there is, the first virtual machine 124 is identified based on the first correspondence information and the second correspondence information, and the contents of the memory 123 used by the guest OS 12 operating on the identified first virtual machine 124 are saved. To do. As will be described later, the first correspondence information includes a plurality of virtual bus identification information for identifying a plurality of virtual buses and a plurality of process identification information for identifying processes of the virtual machine control unit respectively corresponding to the plurality of virtual machines. It is correspondence. As will be described later, the second correspondence information is a correspondence relationship between a plurality of process identification information and a plurality of virtual machine identification information for identifying a plurality of virtual machines.

  In practice, in the host OS 11, the dump collection tool 111 is a memory dump storage unit that stores the contents of the memory 123 used by the guest OS 12 when the guest OS 12 crashes. The memory dump is obtained by converting the stored contents of the memory 123 into a predetermined format. The stored contents of the memory 123 can be converted into a memory dump and converted into a memory dump by the dump collection tool 111. Accordingly, the memory dump process is executed by the host OS 11 of the information processing system.

  For example, the dump collection tool 111 is provided in the host OS 11. When the dump collection tool 111 is notified of the occurrence of a failure from the guest driver 121 as a notification means, the guest having a failure is generated based on the virtual bus identification information of the virtual bus 21 in which the interrupt has occurred, for example, the virtualization ID. The process identification information of the process corresponding to the virtual machine 124 on which the OS 12 operates is obtained. A process corresponding to each of the plurality of virtual machines 124 is a process of the host OS 11 which is a virtual machine control unit.

  Furthermore, the dump collection tool 111 obtains the virtual machine 124 on which the guest OS 12 in which the failure has occurred, specifically, the virtual machine name of the virtual machine 124 is obtained based on the obtained process identification information. Further, the dump collection tool 111 saves the contents of the memory 123 of the virtual machine 124 in which the guest OS 12 in which the requested failure has occurred in a storage device. In other words, the contents of the memory 123 used by the guest OS in which the requested failure has occurred are collected in the storage device. For example, the storage device can be accessed by the host OS 11 and is a storage device different from the memory 123.

  Specifically, when the dump collection tool 111 is notified of the occurrence of a failure from the guest driver 121, the virtualization ID that is the first storage unit is based on the virtualization ID of the virtual bus 21 in which the interrupt has occurred. / From the process list 117, process identification information of the process corresponding to the virtual machine 124 in which the guest OS 12 in which the failure has occurred operates is obtained. Further, the dump collection tool 111 determines, based on the obtained process identification information, from the virtual machine / process list 118 as the second storage unit, the virtual machine 124 in which the guest OS 12 in which the failure has occurred, specifically, The virtual machine name of the virtual machine 124 is obtained. The virtualization ID / process list 117 and the virtual machine / process list 118 will be described later. The virtualization ID / process list 117 and the virtual machine / process list 118 may be collectively referred to as a virtual machine list.

  Actually, in the dump collection tool 111, the host driver 113 detects a crash notification from the guest OS 12, and notifies the dump collection service unit 112 of the crash of the guest OS 12. The notification of the crash from the guest OS 12 is detected by capturing an interrupt to the virtual bus 21 that logically connects the virtual machine and the hardware, as will be described later. The host driver 113 is a kernel driver, for example, and is provided in the host OS 11.

  Specifically, when the host driver 113 is notified of the occurrence of a failure from the guest driver 121 as a notification unit, the host driver 113 performs virtualization based on the virtualization ID of the virtual bus 21 in which the interrupt from the guest driver 121 has occurred. From the ID / process list 117, process identification information of a process corresponding to the virtual machine 124 in which the guest OS 12 in which the failure has occurred operates is obtained. The host driver 113 notifies the dump collection service unit 112 of the obtained process.

  In the dump collection tool 111, the dump collection service unit 112 identifies the virtual machine 124 in which the guest OS 12 in which the failure has occurred operates, and the memory 123 used by the virtual machine 124 in which the guest OS 12 in which the failure has occurred operates. Save the contents. The dump collection service unit 112 is a service program provided by the host OS 11.

  Specifically, the dump collection service unit 112 determines, based on the process identification information obtained by the host driver 113, from the virtual machine / process list 118, the virtual machine 124 in which the failed guest OS 12 operates, specifically, The virtual machine name of the virtual machine 124 is obtained. Then, the dump collection service unit 112 collects the contents of the memory 123 of the virtual machine 124 having the obtained virtual machine name, in other words, the contents of the memory 123 used by the guest OS 12 in which a failure has occurred. The memory 123 used by the guest OS 12 is a memory included in the virtual machine 124 on which the guest OS 12 operates.

  For example, the dump collection service unit 112 accesses the memory 123 used by the guest OS 12 in which the failure has occurred, copies the contents of the memory 123 at the time of the failure, and saves it in the storage state storage area 114. Then, the dump collection service unit 112 reads the contents of the memory 123 stored in the storage state storage area 114, converts the contents into a dump file, and stores the dump file in the dump storage area 115.

  The dump collection service unit 112 is a means for stopping the virtual machine 124. When the guest driver 121, which is a notification means, notifies the occurrence of the failure, the dump collection service unit 112 displays the virtual machine 124 on which the guest OS 12 in which the failure has occurred operates. Stop. The dump collection service unit 112 is a restart unit for the virtual machine 124 and restarts or restarts the stopped virtual machine 124. Actually, the dump collection service unit 112 requests the hypervisor 2 to stop or restart the virtual machine 124. When receiving a request to stop or restart the virtual machine 124 from the dump collection service unit 112, the hypervisor 2 stops and restarts the virtual machine 124 accordingly.

  In the example of FIG. 2, the dump collection tool 111 includes the dump collection service unit 112 and the host driver 113. However, the dump collection tool 111 may be provided separately from the dump collection service unit 112 and the host driver 113.

  For example, the dump collection service unit 112 is a service program of the host OS 11, the host driver 113 is a kernel driver of the host OS 11, and the dump collection tool 111 is provided as an application that operates on the host OS 11, for example. Thereby, the user can acquire the contents of the memory 123 and execute a memory dump without changing the host OS 11.

  The storage state storage area 114 is an area of a storage device such as a magnetic disk device included in the hardware 3. A storage device such as a magnetic disk device can be accessed by the host OS 11 and is a storage device different from the memory 123. The saved state storage area 114 is one area and is common to a plurality of virtual machines 124.

  The dump storage area 115 is an area of a storage device such as a magnetic disk device included in the hardware 3. A storage device such as a magnetic disk device can be accessed by the host OS 11 and is a storage device different from the memory 123. The dump storage area 115 is a single area and is common to a plurality of virtual machines 124.

  FIG. 3 is an explanatory diagram of the information processing system. FIG. 3 shows processing executed by the host OS 11 and processing executed by the guest OS 12. Specifically, FIG. 3 illustrates processing executed by the guest driver 121 of the guest OS 12 and processing executed by the host driver 113 and the dump collection service unit 112 of the host OS 11. In FIG. 3, only one guest OS 12 is shown.

  The dump collection service unit 112 performs a pre-process of the guest control process in the initial process of the dump collection service. The pre-process for the guest control process is a process for writing process identification information in the virtual device 122 of the virtual machine 124. The written process identification information indicates the process of the host OS 11 corresponding to the designated virtual machine 124. A process corresponding to the virtual machine 124 is determined for each virtual machine 124.

  The virtual machine management service unit 116 is a VMM (Virtual Machine Monitor) that manages a plurality of virtual machines 124, and performs processes such as starting, stopping, deleting, and restarting the virtual machine 124. When requested by the dump collection service unit 112 to write process identification information, the virtual machine management service unit 116 writes the process identification information to the virtual device 122 of the designated virtual machine 124.

  For example, the dump collection service unit 112 calls the virtual machine management service unit 116 of the information processing system. The called virtual machine management service unit 116 prepares a virtual device 122 such as a virtualized DVD device and connects it to the host machine. The dump collection service unit 112 of the host OS 11 running on the host machine writes process identification information to a writable DVD in the prepared virtual DVD device. Thereafter, the virtual machine management service unit 116 disconnects the virtual DVD device from the host machine and connects it to the guest machine. The guest machine to be connected corresponds to a process of the host OS 11 corresponding to the guest machine and having process identification information to be written.

  The guest driver 121 of the guest OS 12 executes a process identification acquisition process as an initial process of the guest driver 121. The process identification acquisition process is a process for acquiring process identification information from the virtual device 122. Thereby, information for identifying the process corresponding to the virtual machine 124 from the outside is acquired.

  After the initial process, the guest driver 121 starts the crash detection process, in other words, is registered as a panic exit of the guest OS 12. The crash detection process is a process called from the guest OS 12 when a crash occurs. For this call, the guest driver 121 is registered as a panic exit of the guest OS 12. In practice, the guest OS 12 itself performs the crash detection by various means. Therefore, when the guest OS 12 detects its own crash, the guest OS 12 calls the guest driver 121 registered as its own panic exit.

  Further, the guest driver 121 executes a crash notification process. The crash notification process is a virtual link that logically connects between the virtual machine 124 in which the guest OS 12 operates and the host OS 11 as the virtual machine control unit when a crash occurs, in other words, when called from the guest OS 12. This is a process for generating an interrupt on the bus 21. The crash notification process is executed following the crash detection process, for example.

  The interruption to the virtual bus 21 occurs in the virtual bus 21 assigned to the virtual machine 124 on which the guest driver 121 operates. The virtual bus 21 is assigned to each virtual machine 124 by the hypervisor 2 when the virtual machine 124 is started, for example. Accordingly, the crash of the guest OS 12 is notified to the host driver 113 through the virtual bus 21 assigned to the virtual machine 124 in which each guest OS 12 is operating.

  In addition, the guest driver 121 executes dump output suppression processing. The dump suppression process is a process of suppressing dumping of the memory 123 by the guest OS 12 when a crash occurs, in other words, when called from the guest OS 12. The dump output suppression process is executed, for example, following the crash notification process or in parallel with the crash notification process.

  Since the guest OS 12 usually includes a dump collection function that is a function of collecting the memory dump of the memory 123 of the guest OS 12, a memory dump is collected when the guest OS 12 crashes. Therefore, the guest driver 121 suppresses dumping of the memory 123 by the guest OS 12 when a crash occurs.

  On the other hand, the host driver 113 of the host OS 11 executes an interrupt capturing process. The interrupt capturing process is a process for capturing an interrupt to the virtual bus 21 by the guest driver 121. In other words, the interrupt capturing process is a receiving process for receiving a crash notification from the guest driver 121. The priority of the interrupt for the crash notification from the guest driver 121 is the highest priority. Therefore, the host driver 113 repeatedly executes the interrupt capturing process. The interrupt capturing process is executed after the start of the first guest OS 12, for example.

  Further, the host driver 113 starts the crash determination process, in other words, registers at a predetermined position of the host OS 11. The crash determination process is a process called from the host OS 11 when a crash is determined. For this call, the host driver 113 is registered at a predetermined position of the host OS 11. The crash determination process is a process for checking all interrupts to the virtual bus 21 and determining an interrupt notifying the guest OS 12 of a crash and other interrupts. The crash determination process is executed, for example, following the interrupt capturing process.

  In addition, the host driver 113 executes process identification information resolution processing. The process identification information resolution process is a process for resolving the process corresponding to which virtual machine 124 the guest OS 12 in which the crash has occurred, in other words, corresponding to the virtual machine 124 in which the guest OS 12 in which the crash has occurred is operating. This is a process for obtaining process identification information of a process to be performed. The process identification information resolution process is executed following the crash determination process, for example, when it is determined that the interrupt is an interrupt for notifying the guest OS 12 of a crash.

  In addition, the host driver 113 executes a crash occurrence notification process. The crash occurrence notification process is a process for notifying the dump collection service unit 112 that a crash has occurred in the guest OS 12. The crash occurrence notification process is executed following the process identification information resolution process, for example, when it is determined that the interrupt is an interrupt for notifying the guest OS 12 of a crash. In the crash occurrence notification process, the process identification information of the process in which the guest OS 12 in which the crash has occurred, obtained by the process identification information resolution process, is notified to the dump collection service unit 112.

  On the other hand, the dump collection service unit 112 of the host OS 11 executes a memory content saving process, in other words, a memory state saving process. The memory state saving process is a process for saving the contents of the memory 123 used by the guest OS 12 in which the crash has occurred. The memory 123 used by the guest OS 12 is a virtualized memory included in the virtual machine 124 on which the guest OS 12 operates, in other words, a virtual memory. The memory state saving process is executed, for example, when the dump collection service unit 112 receives a notification from the host driver 113 that the guest OS 12 has crashed.

  Through the memory state saving process, the contents of the memory 123 used by the guest OS 12 in which the crash has occurred are copied and stored in a predetermined storage area, in other words, in the saved state storage area 114. The storage state storage area 114 is determined in advance by the dump collection tool 111, for example. For example, the user can access the dump collection tool 111 operating on the host OS 11 from the terminal and determine the storage state storage area 114 in advance.

  Since the memory state saving process only copies the contents of the memory 123 and stores it in the saving state storage area 114, the process can be completed in a shorter time than the creation of the dump file. When the memory state saving process is completed, the virtual machine 124 on which the guest OS 12 in which the crash occurred can be stopped and restarted without waiting for the creation of the dump file.

  After the memory content saving process, the dump collection service unit 112 executes a dump conversion process as a post process of the memory content saving process. The dump conversion process is a process of converting the contents of the memory 123 stored in the storage state storage area 114 into a dump file having a predetermined format and storing it. The dump conversion process is executed after the execution of the memory state saving process, at a timing that does not cause an obstacle to the execution of the guest control process after the guest control process or in parallel with the guest control process.

  By the dump conversion process, the contents of the memory 123 stored in the saved state storage area 114 are converted into a dump file and stored in a predetermined storage area, in other words, the dump storage area 115. The dump storage area 115 is predetermined by the dump collection tool 111, for example. For example, the user can access the dump collection tool 111 operating on the host OS 11 from the terminal and determine the dump storage area 115 in advance.

  The dump conversion process is executed, for example, at a timing that does not hinder the execution of the guest control process. In other words, the guest control process is executed with priority over the dump conversion process. As a result, the guest control process can be executed after the end of the memory state saving process.

  Further, the dump collection service unit 112 executes guest control processing. The guest control process is a process for controlling the state of the virtual machine 124 that is executing the guest OS 12 in which the crash has occurred. For example, the dump collection service unit 112 calls the virtual machine management service unit 116 of the information processing system and requests the called virtual machine management service unit 116 to control the virtual device 122 used by the specified guest OS 12. Accordingly, the guest control process is actually executed by the virtual machine management service unit 116.

  The guest control process is executed prior to the dump conversion process following the memory state saving process when the dump collection service unit 112 receives a notification that a crash has occurred in the guest OS 12. By the guest control process, the state of the virtual machine 124 on which the guest OS 12 operates is left as it is, or is stopped or restarted.

  3 may be realized as a routine or a subroutine of the guest driver 121, the host driver 113, or the dump collection service unit 112, respectively.

  FIG. 4 is an explanatory diagram of the information processing system.

  In FIG. 4, for example, the virtual machine 124B, in other words, the virtual machine B is executed as the process Y, and the guest OS 12B operates on the virtual machine 124B. A virtual CPU 125B, in other words, virtual CPU-1 is assigned to the virtual machine 124B. The guest OS 12 and the guest driver 121 corresponding to the virtual machine 124B are represented as a guest OS 12B and a guest driver 121B, respectively. The same applies to virtual machine A and virtual machine C. In FIG. 4, it is assumed that a crash occurs in the virtual machine B.

  As described above, in order for the host OS 11 to perform a memory dump for a plurality of guest OSs 12, the host OS 11 needs to know which virtual machine 124 the crashed guest OS 12 is operating on.

  Here, in the information processing system, the virtual machine 124 exists as a process in the information processing system or the host OS 11. Therefore, the contents of the memory 123 of the virtual machine 124 can be acquired from outside the virtual machine 124. Therefore, the contents of the memory 123 when a crash occurs can be collected from outside the virtual machine 124, and a memory dump can be generated.

  However, only one virtual machine 124 is visible from the guest OS 12, while there are a plurality of virtual machines 124 on the information processing system. For example, when viewed from the host OS 11, one virtual machine 124 is one process on which the guest OS 12 operates. The virtual machine 124 viewed from the information processing system is a set of a set of virtual hardware (COP, memory, storage, network device, etc.) managed by the information processing system. In the information processing system, the virtual hardware is strictly separated from the host OS 11 and the guest OS 12 on the virtual machine 124. For this reason, for example, the guest OS 12 cannot know on which virtual machine 124 the guest OS 12 itself operates.

  From the above, when notifying the occurrence of a crash of the guest OS 12, it is necessary to be able to notify information that can identify the virtual machine 124 on which the guest OS 12 operates on the information processing system.

  Therefore, a dump collection service unit 112, a host driver 113, and a guest driver 121 are provided. The dump collection service unit 112 has information for identifying the process of the virtual machine 124, in other words, process identification information. The host driver 113 has information for identifying virtual hardware on the virtual bus 21 of the virtualization system, in other words, a virtualization ID. The guest driver 121 itself does not have process identification information and virtualization ID.

  As shown in FIG. 4, for example, the virtualization ID / process list 117 is provided in the host driver 113. The virtualization ID / process list 117 is a first storage unit and is provided in the host OS 11 and stores first correspondence information that is a correspondence relationship between a plurality of virtual bus identification information and a plurality of process identification information. The virtual bus identification information is identification information for identifying the plurality of virtual buses 21 assigned to each of the plurality of virtual machines 124, and is uniquely determined. The process identification information is identification information for identifying a process of the host OS 11 corresponding to each of the plurality of virtual machines 124, and is uniquely determined. Therefore, the process identification information identifies the process of the virtual machine control unit corresponding to each of the plurality of virtual machines 124. The virtual bus identification information is, for example, a virtualization ID, and is information for identifying a plurality of virtual buses 21 in the hypervisor 2.

  As shown in FIG. 4, for example, a virtual machine / process list 118 is provided in the dump collection service unit 112. The virtual machine / process list 118 is a second storage unit, is provided in the host OS 11, and stores second correspondence information that is a correspondence relationship between a plurality of process identification information and a plurality of machine identification information. The machine identification information is identification information for identifying a plurality of virtual machines 124. The machine identification information is, for example, the virtual machine name of the virtual machine 124 and is uniquely determined.

  First, in order for the host OS 11 to execute a memory dump for a plurality of guest OSs 12, the following pre-processing is executed.

  The dump collection service unit 112 of the host OS 11 sends process identification information of a process corresponding to the virtual machine 124 to each of the plurality of virtual machines 124. Each of the plurality of virtual machines 124 stores the process identification information received from the host OS 11 in the virtual device 122.

  Each guest driver 121 of the plurality of virtual machines 124 sends the process identification information stored in the virtual device 122 to the host driver 113 of the host OS 11.

  The host driver 113 of the host OS 11 uses the process identification information received from each of the plurality of virtual machines 124 to store the plurality of virtual buses 21 assigned to each of the plurality of virtual machines 124 in the virtualization ID / process list 117. In other words, the first correspondence information that is a correspondence relationship between the virtualization ID and the process identification information that identifies the processes that are the plurality of virtual machines 124 is stored. Furthermore, the dump collection service unit 112 of the host OS 11 includes, in the virtual machine / process list 118, process identification information for identifying processes that are a plurality of virtual machines 124, and identification information for each of the plurality of virtual machines 124, in other words, First correspondence information that is a correspondence relationship with the virtual machine name is stored.

  Next, in the operation of the information processing system, the host OS 11 performs a memory dump for the plurality of guest OSs 12 as follows.

  The guest driver 121 generates an interrupt to the virtual bus 21 of the virtualization ID assigned in advance to the guest OS 12 in order to notify the occurrence of the crash of the guest OS 12. In other words, the guest driver 121 notifies the host OS 11 of the occurrence of a failure by generating an interrupt on the virtual bus 21 assigned to the virtual machine 124 in which the guest OS 12 in which the failure has occurred operates.

  Specifically, the guest driver 121 causes the virtual CPU 125 of the virtual machine 124 in which the guest driver 121 is operating to execute an instruction that cannot be executed. As a result, an interrupt occurs in the virtual bus 21 assigned to the virtual machine 124 in which the guest driver 121 is operating.

  The host driver 113 determines the virtualization ID of the virtual bus 21 assigned to the virtual machine 124 in which the guest OS 12 in which the crash occurred operates from the interrupted virtual bus 21. In other words, the host driver 113 is notified of the occurrence of a failure from the guest driver 121 by capturing an interrupt generated on the virtual bus 21 by the guest driver 121.

  The host driver 113 refers to the virtualization ID / process list 117 based on the determined virtualization ID, and based on the virtualization ID of the virtual bus 21 assigned to the virtual machine 124 on which the guest OS 12 in which the crash occurred operates. Thus, the process identification information of the process corresponding to the virtual machine 124 is obtained.

  The dump collection service unit 112 refers to the virtual machine / process list 118 based on the obtained process identification information, and identifies the virtual machine 124 that is the target of collecting the memory dump. As a result, the contents of the memory 123 used by the virtual machine 124 on which the guest OS 12 in which the crash has occurred can be stored in the storage state storage area 114, converted into a dump file format, and stored in the dump storage area 115. it can.

  Hereinafter, processing for the host OS 11 to execute a memory dump for a plurality of guest OSs 12 will be described.

  Since the host driver 113 is a kernel driver of the host OS 11, it can know the virtualization ID in the interrupt to the virtual bus 21, but cannot know the process identification information. Only the host driver 113 can know the virtualization ID. The dump collection service unit 112 can know the process identification information, but cannot access the virtualization ID because it is not the virtual machine 124. Since the virtual machine 124 is distinguished from a process representing the virtual machine 124, in other words, a process corresponding to the virtual machine 124, the process identification information cannot be known. The guest OS 12 cannot know the process identification information and the virtualization ID.

  Therefore, the dump collection service unit 112 writes the process identification information in the virtual device 122 of the virtual machine 124 when, for example, the virtual machine 124 is generated. The guest driver 121 of the guest OS 12 operating on the virtual machine 124 reads the process identification information of the virtual device 122 from the virtual device 122 and passes it to the host driver 113. In response to this, the host driver 113 creates a virtualization ID / process list 117. In addition, the host driver 113 passes the process identification information to the dump collection service unit 112. In response to this, the dump collection service unit 112 creates the virtual machine / process list 118. As a result, when an interrupt to the virtual bus 21 having a virtual ID that is uniquely assigned to the virtual machine 124 and has a unique ID occurs, the host driver 113 can obtain the corresponding process identification information, and the dump collection service The virtual machine name corresponding to the unit 112, in other words, the virtual machine 124, specifically, the virtual machine name of the virtual machine 124 can be obtained.

  First, creation of the virtual machine / process list 118 will be described.

  The dump collection service unit 112 connects to the virtual machine management service unit 116 and creates a virtual machine / process list 118. The virtual machine / process list 118 includes, as information on the virtual machine 124, a virtual machine name, a state of the virtual machine 124 (running, stopped, deleted), and process identification information of a process corresponding to the virtual machine 124. .

  The virtual machine name and the state of the virtual machine 124 are acquired from the virtual machine management service unit 116 that manages the virtual machine 124 by the dump collection service unit 112 when the virtual machine / process list 118 is created or prior to creation, for example. Is done.

  The process identification information of the process corresponding to the virtual machine 124 is unique information given to the running virtual machine 124, and identifies the process corresponding to the running virtual machine 124 independently of the definition of the virtual machine 124. Information. Accordingly, the process identification information is determined for each of the running virtual machines 124 acquired from the virtual machine management service unit 116 by the dump collection service unit 112, for example. Further, by defining the process identification information independently of the virtual machine definition, it is possible to cope with a case where a plurality of virtual machines 124 are simultaneously processed from one virtual machine definition.

  When the dump collection service unit 112 creates the virtual machine / process list 118, the dump collection service unit 112 starts monitoring the virtual machine state and starts monitoring the crash of the virtual machine 124. The dump collection service unit 112 acquires the virtual machine state from the virtual machine management service unit 116 at predetermined time intervals, and updates the virtual machine / process list 118 according to the change of the virtual machine state. If there is a newly started virtual machine 124, new process identification information is created for the newly started virtual machine 124 and written to the virtual device 122 connected to the virtual machine 124.

  When the virtual machine management service unit 116 updates the virtual machine / process list 118, the virtual machine management service unit 116 notifies the host driver 113 of a change in the virtual machine state. This notification includes the type of state change (addition, deletion, executing → stopped, stopped → executed) and process identification information of the process corresponding to the virtual machine 124.

  Next, creation of the virtualization ID / process list 117 will be described.

  The host driver 113 accepts an interrupt from the guest OS 12 and accepts a request from the dump collection service unit 112. For example, the host driver 113 creates and updates the virtualization ID / process list 117 in the processing of the request from the dump collection service unit 112. As described above, since the host driver 113 can access the virtual machine 124 only through the hypervisor 2, it can access the virtualization ID, but it cannot access the process identification information of the process corresponding to the virtual machine 124. . Therefore, the process ID information written to the virtual device 122 of the virtual machine 124 by the dump collection service unit 112 is read via the guest driver 121, and a virtualization ID / process list 117 that associates the virtualization ID with the process identification information is created. To do.

  For example, the virtualization ID is acquired from the hypervisor 2 that manages the virtual bus 21 by the host driver 113 when or before the creation of the virtualization ID / process list 117.

  The host driver 113 receives an IO request from the dump collection service unit 112. The received IO requests include types of crash monitoring start, virtual machine 124 status change (VM start, VM stop, VM deletion), and crash monitoring end.

  The host driver 113 first receives a request to start monitoring for crashes, and starts creating the virtualization ID / process list 117. Specifically, the host driver 113 acquires a virtualization ID from the hypervisor 2, and when the virtual machine 124 identified by the virtualization ID has the guest driver 121, the host driver 113 performs crash monitoring for the virtual machine 124. The start is notified, and the process identification information of the process corresponding to the virtual machine 124 is received from the guest driver 121 as a response. The virtual machine 124 for which the process identification information corresponding to the virtualization ID is obtained is added to the virtualization ID / process list 117 because it is a crash monitoring target.

  The acceptance of the interrupt from the guest OS 12 starts immediately after the initial processing, but the actual interrupt from the guest OS 12 is performed after the confirmation of the guest driver 121 of the guest OS 12 is completed and the information on the guest OS 12 Is added to the virtualization ID / process list 117.

  The host driver 113 receives a request to change the state of the virtual machine 124 after the start of crash monitoring. The host driver 113 updates the virtualization ID / process list 117 according to the contents of the status change of the virtual machine 124. At this time, the virtualization ID / process list 117 is updated using the process identification information included in the change request as a key.

  The host driver 113 receives a request to stop crash monitoring as the dump collection service unit 112 stops. If this request is received, all information in the virtualization ID / process list 117 is discarded.

  Next, delivery of process identification information from the guest driver 121 to the host driver 113 will be described.

  Upon receiving the notification of the start of crash monitoring from the host driver 113, the guest driver 121 accesses the virtual device 122 as a trigger, and acquires process identification information of the process corresponding to the virtual machine 124 on which the guest driver 121 operates. The guest driver 121 sends the acquired process identification information to the host driver 113. Since the process identification information is a value that is uniquely set by the dump collection service unit 112 at the start of the virtual machine 124, it is sufficient that this process is executed only once after the initialization of the guest driver 121.

  At this time, for example, the guest driver 121 sets the process identification information to a predetermined address. The predetermined address is, for example, a physical memory used by the host OS 11. The host driver 113 converts the address on the guest OS 12 into an address on the physical memory of the host OS 11 by the hypervisor 2 and directly refers to it. This is a direct reference by address translation.

  Note that the guest driver 121 may send the acquired process identification information to the host driver 113 using a communication function between the virtual machines 124 of the hypervisor 2. This is an indirect reference by communication between drivers.

  Next, a memory dump for a plurality of guest OSs 12 by the host OS 11 in the operation of the information processing system will be described.

  The guest driver 121B receives from the host driver 113 a notification that the guest OS 12B has started to monitor crashes, and returns process identification information to the host driver 113. Then, the guest driver 121B registers the guest driver 121B at the panic exit of the guest OS 12B.

  When the guest OS 12B detects the occurrence of a crash in the guest OS 12B, the guest driver 121B registered as a panic exit of the guest OS 12B is called. The called guest driver 121B sends a crash notification to the host driver 113. Specifically, the guest driver 121B generates an interrupt on the virtual bus 21 assigned to the virtual machine 124 in which the corresponding guest OS 12B is operating. A virtual ID is given to the virtual bus 21.

  The host driver 113 of the host OS 11 receives the notification of the crash from the guest driver 121B, and based on the virtual bus 21 in which the interrupt has occurred, the virtual bus 21 assigned to the virtual machine 124 in which the guest OS 12 in which the crash has occurred operates. The virtualization ID “ID2” is specified. Then, the host driver 113 searches the virtualization ID / process list 117 using the virtual ID “ID2” of the received notification of the crash as a key, and sends it to the virtual machine 124 on which the guest OS 12 in which the crash occurred operates. “Process Y” which is the process identification information of the corresponding process is obtained.

  The host driver 113 notifies the dump collection service unit 112 of the occurrence of the crash, and at this time, also notifies the process identification information “process Y”.

  The dump collection service unit 112 instructs the host driver 113 to start monitoring the crash of the virtual machine 124. At this time, as a response, interface information used when notifying the occurrence of the crash is sent from the host driver 113 to the dump collection service unit 112.

  The dump collection service unit 112 receives a crash occurrence notification according to the above-described interface information from the host driver 113, and extracts process identification information “process Y” included in the received crash occurrence notification. Then, the dump collection service unit 112 searches the virtual machine / process list 118 using the extracted process identification information “process Y” as a key, and the virtual machine 124 of the virtual machine 124 on which the guest OS 12 in which the crash occurred operates. The name “virtual machine B” is obtained. Thereby, the virtual machine B in which the crash has occurred and the process Y in which the virtual machine B in which the crash has occurred operate are specified.

  When the dump collection service unit 112 identifies the virtual machine B, the dump collection service unit 112 executes dump collection processing for the identified virtual machine B. For example, the dump collection service unit 112 copies the contents of the memory 123 at the time of failure of the memory 123 used by the process Y of the virtual machine B in which the crash has occurred, and saves it in the save state storage area 114. Then, the dump collection service unit 112 reads the contents of the memory 123 stored in the storage state storage area 114, converts the contents into a dump file, and stores the dump file in the dump storage area 115.

  Further, the dump collection service unit 112 requests the virtual machine management service unit 116 to stop and restart the virtual machine B.

  In order to update the virtualization ID / process list 117 according to the stop or restart of the virtual machine 124, a monitoring process for monitoring the state of the virtual machine 124 is executed. This monitoring process is executed by, for example, the virtual machine management service unit 116, and the virtual machine management service unit 116 notifies the host driver 113 of the switching of the state of the virtual machine 124 every time the state of the virtual machine 124 is switched. As a result, the virtualization ID / process list 117 is updated.

  As described above, according to the present invention, the host OS 11 can collect the memory dump of the memory 123 of the guest OS 12. Thereby, since the storage destinations of the collected dump files are collected on the host OS 11 side, the disk area can be shared among the plurality of guest OSs 12. In addition, since the dump collection function of the guest OS 12 is not used, it is possible to eliminate the need to allocate resources for dump collection to each of the virtual machines 124 on which the guest OS 12 operates.

  For example, as described above, when there are ten guest OSs 11 using the memory 123 having a capacity of 2 GB, even if a memory dump is collected for all of the ten virtual machines 124, a disk area of 20 GB is required. Instead, even if a spare disk area is considered, for example, a 4 GB disk area need only be prepared. The collected memory dump may be output as a file, for example, at a time when there is no failure in collecting the next memory dump.

  Accordingly, it is not necessary to acquire a large amount of expensive storage devices used in the information processing system in preparation for a guest OS crash that cannot be predicted, and the burden (cost) of collecting a memory dump can be reduced. it can. Even if the disk area for collecting the memory dump is reduced, the memory dump can be reliably collected when an actual crash occurs, and the cause investigation period can be shortened as a result.

  On the other hand, since the storage destinations of the collected dump files are collected on the host OS 11 side, the configuration of the information processing system can be simplified with respect to the memory dump. Thereby, a plurality of virtual machines 124 can be more easily mounted on one piece of hardware.

  Further, according to the present invention, it is necessary to provide the host driver 113 and the guest driver 121, but it is possible to eliminate the need to perform various settings for collecting the memory dump for each guest OS 12.

  Further, according to the present invention, the process of saving the contents of the memory 123 of the guest OS 12 and the process of converting the saved contents of the memory 123 into a dump file are performed at different timings, in other words, at independent timings. Can do. In addition, the storage process of the contents of the memory 123 can be completed at high speed because it is a copy of the memory 123. Therefore, when the guest OS 12 crashes, the stop processing and the restart processing of the guest OS 12 can be executed before the memory dump collection is completed when only the storage processing of the contents of the memory 123 is completed. As a result, the downtime of the virtual machine 124 can be significantly reduced as compared with the normal memory dump collection.

  Further, according to the present invention, it can be realized by using a function provided in a normal OS or hypervisor 2 such as a panic exit of the guest OS 12 or the virtual bus 21. Thus, at the OS level, it is only necessary to add the host driver 113 and the guest driver 121, and it is possible to eliminate the need to greatly change the guest OS 11 and the hypervisor 2. In addition, this prevents the memory dump collection function of each guest OS 12 from being appropriately configured or set, so that a dump is not collected or sufficient information cannot be acquired when a crash occurs. be able to.

  5 and 6 are integrated identification information creation processing flows, and show processing in the dump collection service unit of the host OS.

  The dump collection service unit 112 of the host OS 11 executes initialization processing of the dump collection service unit 112 (step S11). Thereby, the dump collection service unit 112 returns to a predetermined initial state.

  After the initialization process, the dump collection service unit 112 starts creating process identification information corresponding to the virtual machine 124 used to identify the virtual machine 124, and sets the virtual machines 124 in a predetermined order. For example, the status of the virtual machine 124 to be processed is determined as the processing target in the order of the virtual machine name (step S12).

  When the state of the virtual machine 124 is being executed, the dump collection service unit 112 creates and assigns process identification information to the process corresponding to the virtual machine 124 being executed (step S13). Then, the dump collection service unit 112 writes the assigned process identification information to the virtual device 122 of the virtual machine 124 corresponding to the process (step S14). Then, the dump collection service unit 112 adds the virtual machine 124 being executed to the virtual machine / process list 118, which is a virtual machine list (step S15).

  Thereafter, the dump collection service unit 112 checks whether or not all the virtual machines 124 have been added to the virtual machine / process list 118 (step S16). If the addition to the virtual machine / process list 118 has not been completed for all virtual machines 124, the dump collection service unit 112 repeats step S12 for the next processing target.

  When all the virtual machines 124 have been added to the virtual machine / process list 118, the dump collection service unit 112 starts monitoring the virtual machine state (step S17), and also monitors the guest OS 12 for crashes. Is started (step S18). In step S17, the dump collection service unit 112 notifies the host driver 113 of the start of monitoring of the virtual machine state.

  Thereafter, the dump collection service unit 112 enters a process of waiting for a crash waiting for a notification of a crash of the guest OS 12 (step S19).

  On the other hand, the dump collection service unit 112 starts virtual machine state monitoring processing after step S18. The virtual machine state monitoring process is executed in a state monitoring thread that monitors the state of the virtual machine 124. For example, the state monitoring thread is generated by the dump collection service unit 112 and executed by the state monitoring thread execution unit.

  After step S18, the state monitoring thread execution unit acquires a virtual machine state that is information indicating the state of the virtual machine 124 from the virtual machine management service unit 116 at predetermined time intervals (step S110).

  When the virtual machine state is acquired, the state monitoring thread execution unit checks whether there is a change in the virtual machine state in the acquired virtual machine state (step S111). Then, when there is a change in the virtual machine state, the state monitoring thread execution unit determines the content of the change in the virtual machine state (step S112).

  When the content of the machine state change is the start of a new virtual machine 124, the state monitoring thread execution unit creates and assigns process identification information to the process corresponding to the newly started virtual machine 124. (Step S113). Then, the state monitoring thread execution unit writes the assigned process identification information to the virtual device 122 of the virtual machine 124 corresponding to the assigned process (step S114). Then, the state monitoring thread execution unit adds the newly started virtual machine 124 to the virtual machine / process list 118, which is a virtual machine list (step S115). As a result, the virtual machine / process list 118 is updated in accordance with the change in the virtual machine state that has occurred in the virtual machine 124.

  Thereafter, the state monitoring thread execution unit notifies the dump collection service unit 112 that the virtual machine state has been changed (step S116).

  In step S112, when the content of the change in the machine state is the stop of the virtual machine 124, the state monitoring thread execution unit updates the machine state in the virtual machine / process list 118 for the virtual machine 124 to be stopped ( Step S117). As a result, the virtual machine / process list 118 is updated in accordance with the change in the virtual machine state that has occurred in the virtual machine 124. Thereafter, the state monitoring thread execution unit executes Step S116.

  In step S112, when the content of the change in the machine state is deletion of the virtual machine 124, the state monitoring thread execution unit deletes the virtual machine 124 to be deleted from the virtual machine / process list 118 (step S118). As a result, the virtual machine / process list 118 is updated in accordance with the change in the virtual machine state that has occurred in the virtual machine 124. Thereafter, the state monitoring thread execution unit executes Step S116.

  Thereafter, the state monitoring thread execution unit waits until a predetermined time elapses (step S119), and then repeats step S110.

  On the other hand, when there is no change in the virtual machine state in step S111, the state monitoring thread execution unit executes step S119.

  FIG. 7 is an identification information creation processing flow and shows processing in the host driver 113 of the host OS 11.

  The host driver 113 of the host OS 11 executes initialization processing for the host driver 113 (step S21). As a result, the host driver 113 returns to a predetermined initial state.

  After the initialization process, the host driver 113 starts accepting an interrupt from the guest OS 12 via the virtual bus 21 (step S22). Acceptance of an interrupt from the guest OS 12 is started immediately after the initialization process. However, the interruption from the guest OS 12 actually occurs after the guest OS 12 is added to the virtualization ID / process list 117 as described below.

  Thereafter, the host driver 113 starts accepting various requests from the dump collection service unit 112 (step S23). In the process of receiving a request from the dump collection service unit 112, the creation and update of the virtualization ID / process list 117 are executed as follows.

  After step S22, the host driver 113 enters a process of waiting for an interrupt to the virtual bus 21 that is a notification of a crash of the guest OS 12, in other words, a process of waiting for a crash to occur (step S24).

  After step S23, the host driver 113 starts service request processing and accepts an IO request from the dump collection service unit 112 (step S25). When the host driver 113 receives the IO request from the dump collection service unit 112, the host driver 113 determines the type of the received IO request and the change content (step S26).

  When the IO request received in step S26 is the stop of the virtual machine 124 or the deletion of the virtual machine 124, the host driver 113 uses the process identification information of the process corresponding to the virtual machine 124 to be stopped or deleted. Then, it is checked whether or not the virtual machine 124 to be stopped or deleted has been added to the virtualization ID / process list 117 of the virtual machine list with reference to the virtualization ID / process list 117 (step S27). When the virtual machine 124 to be stopped or deleted has been added to the virtualization ID / process list 117 (Yes in step S27), the host driver 113 determines that the virtual machine 124 to be stopped or deleted. Is deleted from the virtualization ID / process list 117 (step S28).

  Note that the host driver 113 determines that the virtual machine 124 to be stopped or deleted is not added to the virtualization ID / process list 117 (No in step S27), and the virtual machine 124 to be stopped or deleted. Do not process for.

  If the IO request received in step S26 is the start of monitoring the crash of the virtual machine 124 or the start of the virtual machine 124, the host driver 113 starts the process of creating and updating the virtualization ID / process list 117. First, the host driver 113 connects to the hypervisor 2 and acquires and enumerates the virtualization IDs of the virtual bus 21 assigned to the virtual machines 124 on which all the guest OSs 12 operate (step S29).

  Thereafter, the host driver 113 includes the guest driver 121 in the virtual machine 124 identified by the virtualization ID, with one virtualization ID as a processing target, in the ascending order of the virtualization ID, for example. It is checked whether it has been added to the cage virtualization ID / process list 117 (step S210). When the processing target virtual machine 124 has been added to the virtualization ID / process list 117 (step S210, Yes), the host driver 113 repeats step S29 for the next processing target virtual machine 124.

  If the processing target virtual machine 124 has not been added to the virtualization ID / process list 117 (No in step S210), the host driver 113 starts the crash monitoring for the processing target virtual machine 124. Notification is made (step S211), and as a response to this notification, the process identification information is acquired from the virtual machine 124 to be processed (step S212). Then, the host driver 113 adds the acquired virtualization ID and process identification information corresponding to the acquired virtual machine 124 to the virtualization ID / process list 117 (step S213). Step S29 is repeated until the process for the virtualization ID is completed. As a result, the virtual machine 124 for which the virtualization ID and the process identification information corresponding to the virtualization ID can be acquired is registered in the virtualization ID / process list 117 and is subject to crash monitoring.

  If the IO request received in step S26 is to stop monitoring the crash of the virtual machine 124, the host driver 113 discards information on all virtual machines 124 in the virtualization ID / process list 117 (step S214). .

  FIG. 8 is an identification information creation processing flow, and shows processing in the guest driver 121 of the guest OS 12.

  The guest driver 121 executes initialization processing for the guest driver 121 (step S31). As a result, the guest driver 121 returns to a predetermined initial state.

  After the initialization process, the guest driver 121 waits for a crash monitoring start notification from the host driver 113 (step S32).

  Thereafter, the waiting guest driver 121 receives a notification of the start of crash monitoring from the host driver 113 (step S33). Upon receiving the notification of the start of monitoring the crash, the guest driver 121 accesses the virtual device 122 used by the corresponding guest OS 12, and the process of the process corresponding to the virtual machine 124 that operates itself from the virtual device 122. Process identification information is read and acquired (step S34).

  Thereafter, the guest driver 121 notifies the host driver 113 of the process identification information acquired from the virtual device 122 (step S35). Then, the guest driver 121 registers the guest driver 121 at the panic exit of the guest OS 12 (step S36). For example, the guest driver 121 is registered by registering a function called from a panic exit of the guest OS 12 when a crash occurs. Thereafter, the guest driver 121 executes a waiting process for waiting for the guest OS 12 to end or a crash occurrence (step S37).

  FIG. 9 is a dump collection processing flow and shows processing in the guest driver 121 of the guest OS 12.

  The guest driver 121 generates an OS state monitoring thread and a crash waiting thread after the initialization process and the virtual machine identification information creation process of the guest driver 121 (step S41). The OS state monitoring thread is executed by the OS state monitoring thread execution unit. The client waiting thread is executed by the client waiting thread execution unit. The OS state monitoring thread and the client waiting thread are executed in parallel. The process in step S41 is the process shown in FIG.

  The guest driver 121, in other words, the OS state monitoring thread execution unit enters an event waiting state in the information processing system after step S41 (step S42). That is, the guest driver 121 receives a notification of the start of crash monitoring from the host driver 113, returns process identification information to the host driver 113, registers the guest driver 121 in the panic exit of the guest OS 12, and then waits for an event. become.

  Thereafter, when an event occurs, the OS state monitoring thread execution unit checks whether the generated event is an event of termination of the guest OS 12 (step S43). When the generated event is not an event of termination of the guest OS 12 (No at Step S43), the OS state monitoring thread execution unit repeats Step S42.

  When the generated event is an event of termination of the guest OS 12 (step S43, Yes), the OS state monitoring thread execution unit cancels the registration of the guest driver 121 at the panic exit of the guest OS 12, and the guest OS 12 itself Is terminated (step S44).

  On the other hand, the guest driver 121, in other words, the crash waiting thread execution unit, when the guest OS 12 detects the occurrence of a crash in itself, calls the guest driver 121 registered in the panic exit of the guest OS 12 (step S45).

  The called guest driver 121 interrupts the host OS 11 via the virtual bus 21, thereby notifying the host driver 113 of the occurrence of a crash in the guest OS 12 (step S46).

  Then, the guest driver 121 calls a processing routine for suppressing the output of the memory dump by the guest OS 12 subsequent to step S46 or in parallel with step S46, and suppresses the output of the memory dump (step S47). This processing routine is an infinite loop that prevents the contents of the memory 123 of the virtual machine 124 from being changed until the contents of the memory 123 used by the guest OS 12 are stored from outside the virtual machine 124.

  FIG. 10 is a dump collection processing flow and shows processing in the host driver 113 of the host OS 11.

  The host driver 113 generates an OS state monitoring thread and an interrupt waiting thread after the initialization process of the host driver 113 and the virtual machine identification information creation process (step S51). The OS state monitoring thread is executed by the OS state monitoring thread execution unit. The interrupt waiting thread is executed by the interrupt waiting thread execution unit. The OS state monitoring thread and the interrupt waiting thread are executed in parallel. Note that the processing in step S51 is the processing shown in FIG.

  By generating the interrupt waiting thread, the host driver 113 can accept an interrupt from the guest driver 121 of the guest OS 12. The host driver 113 receives a crash monitoring start request from the dump collection service unit 112 and creates a crash monitoring target virtualization ID / process list 117, and then starts actual crash monitoring. In other words, the host driver 113 enters an event waiting state and an interrupt waiting state in the information processing system.

  The host driver 113, in other words, the OS state monitoring thread execution unit enters a state of waiting for an event in the information processing system after step S51 (step S52).

  Thereafter, when an event occurs, the OS state monitoring thread execution unit checks whether the generated event is an event of termination of the host OS 11 (step S53). When the generated event is not an event of termination of the host OS 11 (No at Step S53), the OS state monitoring thread execution unit repeats Step S52.

  When the generated event is an event of termination of the host OS 11 (step S53, Yes), the OS state monitoring thread execution unit releases the interrupt waiting thread (step S54), and the virtual machine list virtualization ID / process The list 117 is discarded (step S55), and the process is terminated.

  On the other hand, the host driver 113, in other words, the interrupt waiting thread execution unit waits for an interrupt from the information processing system (step S56).

  When an interrupt occurs in the host OS 11 via the virtual bus 21 by the guest driver 121 in response to the occurrence of a crash in the guest OS 12, the interrupt waiting thread execution unit captures the generated interrupt (step S57). Then, the interrupt waiting thread determines whether or not the interrupt is generated when the guest OS 12 crashes by referring to information accompanying the interrupt such as the virtualization ID (step S58).

  If the interrupt is not triggered by the crash of the guest OS 12, it is an interrupt that was triggered by an event that occurred in the host OS 11, so the interrupt-waiting thread execution unit sends the captured interrupt to the interrupt for the host OS 11. After transferring to the waiting thread (step S59), step S56 is repeated. The interrupt waiting thread is a thread that processes an interrupt caused by the crash of the guest OS 12, whereas the interrupt waiting thread execution unit for the host OS 11 is a thread that processes an interrupt caused by the host OS 11.

  When an interrupt occurs when the guest OS 12 crashes, the interrupt waiting thread execution unit resolves the process identification information based on the virtualization ID (step S510). Specifically, the interrupt waiting thread execution unit identifies the virtualization ID of the virtual bus 21 assigned to the virtual machine 124 in which the guest OS 12 in which the crash occurred operates based on the virtual bus 21 in which the interrupt has occurred. Then, by referring to the virtualization ID / process list 117 using the specified virtualization ID as a key, process identification information of a process corresponding to the virtual machine 124 on which the guest OS 12 in which the crash has occurred is obtained. Thereby, the process of the host OS 11 corresponding to the virtual machine 124 in which the guest OS 12 in which the crash occurred operates can be specified.

  Thereafter, the interrupt waiting thread execution unit notifies the identified virtual machine 124 to the dump collection service unit 112 (step S511), and repeats step S56. This notification includes process identification information of a process corresponding to the virtual machine 124 in which the guest OS 12 in which the crash has occurred operates.

  FIG. 11 is a dump collection processing flow and shows processing in the dump collection service unit 112 of the host OS 11.

  The dump collection service unit 112 generates an OS state monitoring thread and a crash event waiting thread after the initialization process and the virtual machine identification information creation process of the dump collection service unit 112 (step S61). The OS state monitoring thread is executed by the OS state monitoring thread execution unit. The crash event waiting thread is executed by the crash event waiting thread execution unit. The OS state monitoring thread and the crash event waiting thread are executed in parallel. The process of step S61 is the process shown in FIGS.

  After step S61, the dump collection service unit 112, in other words, the OS state monitoring thread execution unit enters an event waiting state in the information processing system (step S62).

  Thereafter, when an event occurs, the OS state monitoring thread execution unit checks whether the generated event is an event of termination of the host OS 11 (step S63). When the event that has occurred is not an event of termination of the host OS 11 (No in step S63), the OS state monitoring thread execution unit repeats step S62.

  If the event that has occurred is an event that the host OS 11 has ended (step S63, Yes), the OS state monitoring thread execution unit stops receiving the crash event by the crash event waiting thread (step S64), and the host driver 113 Is instructed to stop crash monitoring (step S65), the virtual machine / process list 118 in the virtual machine list is discarded (step S66), and the process is terminated.

  On the other hand, the dump collection service unit 112, in other words, the crash event waiting thread execution unit enters an event waiting state of the information processing system (step S67).

  When the crash event waiting thread execution unit receives a crash occurrence notification from the host driver 113, the crash event waiting thread execution unit identifies the virtual machine 124 in which the guest OS 12 in which the crash occurred operates based on the crash occurrence notification. The reason for the occurrence of the crash is that the virtual machine 124 and the guest OS 12 cannot know as which process the virtual machine 124 is executed. Then, when specifying the virtual machine 124, the interrupt waiting thread execution unit activates a dump collection thread corresponding to the specified virtual machine 124 (step S68). The dump collection thread is activated for each virtual machine 124. Therefore, a plurality of dump collection threads may be activated. The dump collection thread is executed by the dump collection thread execution unit. Since the memory dump process takes a certain amount of time, the process of the dump collection service unit 112 and the process of the crash event waiting thread are executed asynchronously.

  After starting the dump collection thread, the crash event waiting thread execution unit returns to the event waiting state in step S67.

  The activated dump collection thread execution unit stores the contents of the memory 123 used by the guest OS 12 in which the crash has occurred in the storage state storage area 114 as a temporary file (step S69). Then, the dump collection thread execution unit activates a dump conversion thread that executes dump conversion processing (step S610).

  Thereafter, the dump collection thread execution unit requests the virtual machine management service unit 116 to stop and restart the virtual machine 124 (step S611).

  On the other hand, the activated dump conversion thread execution unit converts the temporary file having the contents of the memory 123 stored in the storage state storage area 114 into a dump file format and stores it in the dump storage area 115 (step S612). .

  As mentioned above, although this invention was demonstrated based on the Example, this invention can be variously deformed within the range of the main point.

  For example, a notification other than the virtual bus 21 may be used for notification of a crash from the guest OS 12 to the host OS 11. In this case, the crash notification may be captured with the same priority as the hardware interrupt. Further, identification information other than the virtualization ID may be used in the crash notification.

  Further, not only the crash of the guest OS 12 but also the crash of the host OS 11 may be detected, and a dump file of the memory used by the host OS 11 as the virtual machine control unit may be acquired by the same means.

  Further, the virtual machine 124 on which the guest OS 12 operates may hold the process identification information by means other than the virtual device 122.

  Further, the host driver 113 and the dump collection service unit 112 may be integrally formed. In this case, the virtualization ID / process list 117 and the virtual machine / process list 118 may be integrally formed.

  Further, the host OS 11 may include a virtualization ID / process list 117 and a virtual machine / process list 118. The host OS 11 includes a list in which the virtualization ID / process list 117 and the virtual machine / process list 118 are integrally formed, and provides necessary information in response to requests from the host driver 113 and the dump collection service unit 112. You may make it do.

1 Virtual machine 2 Hypervisor 3 Hardware 11 Host OS
12 Guest OS
21 Virtual Bus 111 Dump Collection Tool 112 Dump Collection Service Unit 113 Host Driver 114 Storage State Storage Area 115 Dump Storage Area 116 Virtual Machine Management Service Section 117 Virtualization ID / Process List 118 Virtual Machine / Process List 121 Guest Driver 122 Virtual Device 123 Memory 124 Virtual machine 125 Virtual CPU

Claims (9)

An information processing apparatus that operates a plurality of virtual machines,
A plurality of virtual bus identification information for identifying a plurality of virtual buses connecting the virtual machine control unit for controlling the plurality of virtual machines and the plurality of virtual machines, and the virtual machine control corresponding to the plurality of virtual machines, respectively. First storage means for storing first correspondence information that is a correspondence relationship with a plurality of process identification information for identifying a process of a part;
Second storage means for storing second correspondence information which is a correspondence relationship between the plurality of process identification information and the plurality of virtual machine identification information for identifying the plurality of virtual machines;
There was an interrupt corresponding to the failure of the first OS on the first virtual bus corresponding to the first virtual machine from the first OS operating on the first virtual machine included in the plurality of virtual machines. A first virtual machine is identified based on the first correspondence information and the second correspondence information, and the contents of the memory area used by the first OS operating on the first virtual machine are stored. An information processing apparatus comprising: storage means. The plurality of virtual buses are included in a hypervisor that configures the plurality of virtual machines by virtualizing a physical device of the information processing apparatus, and are assigned to each of the plurality of virtual machines by the hypervisor, The information processing apparatus according to claim 1, wherein bus identification information is given. The storage means is notified of the occurrence of a failure by causing an interrupt to a virtual bus assigned to the first virtual machine when a failure occurs in the first guest OS from the first virtual machine. The process identification of the process of the virtual machine control unit corresponding to the virtual machine on which the first guest OS operates from the first storage unit based on the virtualization identification information of the virtual bus in which the interrupt occurred A virtual machine having the machine identification information obtained by obtaining the machine identification information of the virtual machine on which the first guest OS operates from the second storage unit based on the obtained process identification information; The content of a memory area used by the operating first OS is stored in a storage device different from the memory area. The information processing apparatus described. The storage means includes a host driver provided in a second OS that is the virtual machine control unit, and a dump collection service unit that is a service program of the second OS,
When the host driver is notified of the occurrence of a failure from the first virtual machine, the first OS stores the first OS from the first storage unit based on virtual bus identification information of the virtual bus in which the interrupt has occurred. Obtaining the process identification information of the process of the virtual machine control unit corresponding to the operating virtual machine;
The dump collection service unit obtains the machine identification information of the virtual machine on which the first OS operates from the second storage unit based on the obtained process identification information, and includes the obtained machine identification information. The information processing apparatus according to claim 3, wherein contents of the memory area of the machine are stored in the storage device. The information processing apparatus includes:
A virtual device for storing process identification information for identifying a plurality of processes of the virtual machine control unit corresponding to the plurality of virtual machines;
Each of the plurality of virtual machines sends the process identification information stored in the virtual device to the virtual machine control unit,
The storage unit uses the process identification information received from each of the plurality of virtual machines, and stores the correspondence between the plurality of virtual bus identification information and the plurality of process identification information in the first storage unit. The first correspondence information is stored, and the second correspondence information that is a correspondence relationship between the plurality of process identification information and the plurality of machine identification information is stored in the second storage unit. The information processing apparatus according to claim 1. The storage means sends the process identification information for each of the processes of the virtual machine control unit corresponding to each of the plurality of virtual machines to each of the plurality of virtual machines,
The information processing apparatus according to claim 5, wherein each of the plurality of virtual machines stores the process identification information received from the virtual machine control unit in the virtual device. A method for controlling an information processing apparatus in which a plurality of virtual machines operate,
A plurality of virtual bus identification information for identifying a plurality of virtual buses connecting the virtual machine control unit and each of the plurality of virtual machines; Storing first correspondence information, which is a correspondence relationship with a plurality of process identification information for identifying processes of the corresponding virtual machine control units, in a first storage unit;
The virtual machine control unit stores second correspondence information, which is a correspondence relationship between the plurality of process identification information and the plurality of virtual machine identification information for identifying the plurality of virtual machines, in a second storage unit;
The virtual machine control unit causes a failure of the first OS to a first virtual bus corresponding to the first virtual machine from a first OS operating on a first virtual machine included in the plurality of virtual machines. Is used, the first OS that operates on the first virtual machine is used by specifying the first virtual machine based on the first correspondence information and the second correspondence information. A method for controlling an information processing apparatus, characterized by storing the contents of a memory area. A control program for an information processing apparatus that operates a plurality of virtual machines,
The control program is stored in a computer.
A plurality of virtual bus identification information for identifying a plurality of virtual buses connecting the virtual machine control unit and each of the plurality of virtual machines; A process of storing first correspondence information, which is a correspondence relationship with a plurality of process identification information for identifying processes of the corresponding virtual machine control units, in a first storage unit;
Processing in which the virtual machine control unit stores second correspondence information, which is a correspondence relationship between the plurality of process identification information and the plurality of virtual machine identification information for identifying the plurality of virtual machines, in a second storage unit. When,
The virtual machine control unit causes a failure of the first OS to a first virtual bus corresponding to the first virtual machine from a first OS operating on a first virtual machine included in the plurality of virtual machines. Is used, the first OS that operates on the first virtual machine is used by specifying the first virtual machine based on the first correspondence information and the second correspondence information. A control program for executing a process for saving the contents of a memory area. A computer-readable recording medium for recording a control program of an information processing apparatus on which a plurality of virtual machines operate,
The control program is stored in a computer.
A plurality of virtual bus identification information for identifying a plurality of virtual buses connecting the virtual machine control unit and each of the plurality of virtual machines; A process of storing first correspondence information, which is a correspondence relationship with a plurality of process identification information for identifying processes of the corresponding virtual machine control units, in a first storage unit;
Processing in which the virtual machine control unit stores second correspondence information, which is a correspondence relationship between the plurality of process identification information and the plurality of virtual machine identification information for identifying the plurality of virtual machines, in a second storage unit. When,
The virtual machine control unit causes a failure of the first OS to a first virtual bus corresponding to the first virtual machine from a first OS operating on a first virtual machine included in the plurality of virtual machines. Is used, the first OS that operates on the first virtual machine is used by specifying the first virtual machine based on the first correspondence information and the second correspondence information. A recording medium that executes processing for saving the contents of a memory area.

Images