JP2017062841A - Information processing device having memory dump function - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information processing device for collecting a memory dump for cause investigation by quickly performing reactivation to resume operation when an abnormality requiring the reactivation is detected.SOLUTION: In the case of detecting an error of an OS in execution, a hypervisor 351 stops the OS in which the error is detected. A firmware 311 stops the hypervisor for controlling a virtual machine in which the OS with the error detected operates, changes a memory area used by the stopped OS to a second memory area different from a first memory area used by a kernel of the stopped OS, and activates the stopped hypervisor. The activated hypervisor activates the stopped OS with the second memory area as a use area. The activated OS resumes a stopping program, and writes data read from the memory area used by the hypervisor as a dump file of the hypervisor in a file in a state where the hypervisor for controlling a virtual machine operates.SELECTED DRAWING: Figure 3

Description

  The present invention relates to an information processing apparatus having a memory dump function.

  In recent years, UNIX (registered trademark) servers and IA servers have been introduced into backbone systems, and high availability of UNIX (registered trademark) servers and IA servers is regarded as important. In general, when a fatal error occurs in the system, the system is urgently stopped (panic), and a memory dump is saved on a disk in order to investigate the cause.

  Since the system cannot be used while the system is stopped, it is an important requirement to restart the system promptly.

  However, in recent years, servers with a terabyte (TB) order of installed memory have appeared, and in such a system, it takes a very long time to collect a memory dump, and the system can be restarted quickly. It is gone.

  Also, do not save the memory dump on the disk, save the memory contents at the time of emergency stop to another memory or save the memory contents at the time of failure in the dump storage area, save a part of the memory. A method of converting memory contents not saved after rebooting into a dump file is known.

Japanese Patent Laid-Open No. 11-212836 JP 2001-229053 A JP 2006-72931 A JP 2005-122334 A

  However, in the conventional method, the memory dump at the time of occurrence of an error is saved in another memory or disk. If the size of the memory dump to be saved is large, it takes time to copy the memory, and the system can be quickly restarted. There was a problem that it could not be started.

  In addition, when the operating system detects a fatal error and stops the system urgently, the operating system that detected the abnormality collects a dump, so the abnormality is detected again during the dump collection process and a hang-up occurs. There was a problem that secondary damage might occur.

  In one aspect, an object of the present invention is to promptly restart a job and collect a memory dump for investigating the cause when an abnormality that requires a restart is detected.

  In one aspect, the information processing apparatus according to the embodiment includes a storage device and an arithmetic processing device, and the processing of the hypervisor that controls a virtual machine on which an operating system is operated by the arithmetic processing device and the storage device; Firmware processing for controlling the system including the memory and the processor is executed.

  When the hypervisor detects an error in the operating system being executed, the hypervisor stops the operating system in which the error is detected.

  The firmware stops the hypervisor that controls the virtual machine on which the operating system in which the error is detected is running, and the memory area used by the stopped operating system is used by the kernel of the stopped operating system. The second hypervisor is changed to a second memory area different from the first memory area, and the stopped hypervisor is activated.

  The activated hypervisor activates the stopped operating system using the second memory area as a use area.

  The started operating system resumes the stopped program in response to the stop of the operating system that detected the error, and reads from the memory area used by the hypervisor while the hypervisor that controls the virtual machine is running. Execute the process to write the data to a file as a hypervisor dump file.

  According to the information processing apparatus of one embodiment, when an abnormality that needs to be restarted is detected, it is possible to quickly resume the business and collect a memory dump for investigating the cause.

It is a hardware block diagram of the server which concerns on embodiment. It is a figure which shows the correspondence of the server and function which concern on embodiment. It is a functional block diagram of the physical partition which concerns on embodiment. It is a block diagram of the firmware which concerns on embodiment. It is a block diagram of the hypervisor which concerns on embodiment. It is a block diagram of OS which concerns on embodiment. 3 is a flowchart of a memory dump generation process according to the first embodiment. 3 is a flowchart of a memory dump generation process according to the first embodiment. It is a flowchart of the modification of the memory dump production | generation process which concerns on 1st Embodiment. It is an example of HV dump target area information. It is a flowchart of the memory dump production | generation process which concerns on 2nd Embodiment. It is a flowchart of the memory dump production | generation process which concerns on 2nd Embodiment. It is a flowchart of the memory dump production | generation process which concerns on 2nd Embodiment. It is an example of kernel dump target area information. It is a flowchart of the memory dump generation process of the kernel by the domain for dump collection. It is a figure which shows collection of the memory dump by the domain for dump collection. It is a figure which shows the PA-RA mapping information in the memory dump collection by the domain for dump collection. 6 is a flowchart of a kernel memory dump generation process using a memory dynamic reconfiguration function. It is a figure which shows extraction of the memory dump using the memory Dynamic Reconfiguration function. It is a figure which shows the PA-RA mapping information in extraction | collection of the memory dump using a memory Dynamic Reconfiguration function. It is a flowchart of the memory dump production | generation process which concerns on 3rd Embodiment. It is a flowchart of the memory dump production | generation process which concerns on 4th Embodiment. It is a flowchart of the memory dump production | generation process which concerns on 4th Embodiment. It is a flowchart of the memory dump generation process of the hypervisor in operation.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a hardware configuration diagram of a server according to the embodiment.

  The server (information processing apparatus) 10 includes a system board 11-i (i = 1 to 3), a service processor (SP) 21, a disk unit 31, and a communication interface 41.

  The system board 11-i, service processor 21, disk unit 31, and communication interface 41 are connected via a bus 51.

  The system board 11-i includes a Central Processing Unit (CPU) 12-ik (k = 1, 2), a nonvolatile memory 14-i, and a memory 13-ik.

  The service processor 21 is a device that performs control of the server 10, control of physical partitions in the server 10, and the like. The service processor 21 includes a CPU 22 and a memory 23. The service processor 21 is an example of a control unit.

  The CPU 22 performs various processes such as control of the server 10 and control of physical partitions in the server 10.

  The memory 23 temporarily stores data used by the service processor 21. The memory 23 is, for example, a RAM.

  The disk unit 31 includes a hard disk drive (HDD) 32-i (i = 1 to 3).

  The HDD 32 is a device that stores data used by the server 10. The HDD 32 is an example of a storage unit.

  The communication interface 41 is an interface that communicates with a device connected to the server 10.

FIG. 2 is a diagram illustrating an example of a correspondence relationship between a server and a function according to the embodiment.
The server 10 is operated by being divided into two physical partitions 61-k (k = 1, 2). In the specification, the physical partitions 61-1 and 61-2 may be referred to as physical partition # 0 and physical partition # 1, respectively.

  The physical partitions # 0 and # 1 are controlled by the service processor 21. The CPUs included in the physical partitions # 0 and # 1 are an example of a processing unit.

  The physical partition # 0 composed of the system boards 11-1 and 11-2 is further divided into four logical domains # 0 to # 3 in the physical partition # 0, and independent in each logical domain # 0 to # 3. The operating system (OS) is running. The hypervisor (HV) # 0 controls the correspondence between the physical resources in the physical partition # 0 and the logical domains # 0 to # 3.

  In the physical partition # 1 configured from the system board 11-3, an operating system (OS) is operating in the logical domain # 4. The hypervisor # 1 controls the correspondence between the physical resources in the physical partition # 1 and the logical domain # 4.

FIG. 3 is a functional block diagram of a physical partition according to the embodiment.
The physical partition 61-1 includes a logical domain 201-m (m = 1 to 4), firmware (FW) 311 and a hypervisor (HV) 351.

  The physical partition 61-1 in FIG. 3 corresponds to the physical partition 61-1 in FIG.

  The logical domains 201-1 to 201-4 correspond to the logical domains # 0 to # 3 shown in FIG.

  In the specification, the logical domain 201-1 may be expressed as a control domain # 0.

  Further, in the specification, the logical domain 201-4 may be referred to as a dump dedicated domain # 3 or a dump collection domain 201-4.

Hereinafter, unless otherwise specified, a logical domain refers to a virtual machine.
The logical domain 201-m includes a CPU 202-m-k (k = 1, 2), a memory 203-m, and a disk 204-m. Hereinafter, unless otherwise specified, the CPU 202, the memory 203, and the disk 204 are a virtual CPU, a virtual memory, and a virtual disk, respectively.

The CPU 202-m-k executes various processes.
The memory 203-m stores various programs and data read from the disk 204-m.

  The firmware 311 controls the entire server 10 (a plurality of physical partitions 61-1 and 61-2), and performs, for example, hardware initialization, memory diagnosis, temperature monitoring, and the like. The firmware 311 includes a power on self test (POST) developed on the service processor 21 of FIG. 1 and the nonvolatile memories 14-1 and 14-2 of the system boards 11-1 and 11-2. Here, the Power On Self Test (POST) is a program that executes diagnosis and initialization of hardware resources when the system is activated.

  The hypervisor 351 controls the logical domain 201-m and an operating system (OS) 401-m operating on the logical domain 201-m. The hypervisor 351 is expanded on the memories 13-1-1, 13-1-2, 13-2-1 and 13-2-2 of the system boards 11-1 and 11-2 in FIG. 1, 12-1-2, 12-2-1 and 12-2-2.

The lower part of FIG. 3 shows the software in the physical partition 61-1.
The OS 401-m is operating in each of the logical domains # 0 to # 3 in the physical partition 61-1.

  The OSs 401-1 to 401-4 correspond to the operating systems of the logical domains 201-1 to 201-4, respectively.

FIG. 4 is a detailed configuration diagram of the firmware according to the embodiment.
The firmware 311 includes a dump target area information / HV dump flag storage processing unit 312, a dump target area information / HV dump flag storage area 313, an HV dump flag setting unit 314, a memory initialization processing unit 315, an HV use area changing unit 316, HV restart command unit 317, dump target area information / HV dump flag notification unit 318, PA-RA mapping notification unit 319, memory release processing unit 320, and HV dump flag reset processing unit 321 are provided.

  The dump target area information / HV dump flag storage processing unit 312 stores the dump target area information and the HV dump flag in the dump target area information / HV dump flag storage area 313.

  The dump target area information / HV dump flag storage area 313 is an area in which the dump target area information and the HV dump flag are stored. Here, the dump target area information is information indicating the dump target area, and includes information on the start address (PA Base) and size of the dump target area. The HV dump flag is control information indicating whether or not to generate a dump file in the memory area used by the hypervisor. The HV dump flag can also include information (HV live dump flag) indicating whether or not to collect a memory dump of the active hypervisor.

  The HV dump flag setting unit 314 sets the value of the HV dump flag. For example, the HV dump flag setting unit 314 sets the HV dump flag to TRUE when performing HV dump.

The memory initialization processing unit 315 initializes the memory.
The HV use area changing unit 316 changes the area of the memory used by the hypervisor 351.

The HV restart command unit 317 instructs the hypervisor 351 to restart.
The dump target area information / HV dump flag notifying unit 318 notifies the dump target area information and the HV dump flag.

  The PA-RA mapping notifying unit 319 notifies the hypervisor 351 of PA-RA mapping necessary for OS processing, and the PA-RA mapping processing unit 368 of the hypervisor 351 uses the notified PA-RA mapping. Conversion from a physical address (PA) to a real address (RA) or from a real address (RA) to a physical address (PA) is performed.

The memory release processing unit 320 performs a memory release process.
The HV dump flag reset processing unit 321 resets the HV dump flag. Specifically, the HV dump flag reset processing unit 321 sets the HV dump flag to FALSE.

FIG. 5 is a detailed configuration diagram of the hypervisor according to the embodiment.
The hypervisor 351 includes a domain emergency stop instruction unit 352, an OS panic instruction unit 353, an HV dump target area notification processing unit 354, an HV dump target area information / HV dump flag storage processing unit 355, and an HV dump target area information / HV dump flag. Storage area 356, HV restart processing section 357, OS restart instruction section 358, HV memory dump flag reading / transmitting section 359, HV dump target area reading processing section 360, memory management section 361, memory release processing section 362, HV dump Flag reset processing unit 363, HV dump flag notification unit 364, dump dedicated domain activation processing unit 365, kernel dump target area information / kernel dump flag storage processing unit 366, kernel dump target area information / kernel dump flag storage area 367, PA- RA mapping processing unit 368, PA-RA mapping information storage area 369, interrupt processing Part 370, a memory dump processing starting unit 371, the memory initialization section 372 and a kernel dump flag resetting unit 373.

The domain emergency stop instruction unit 352 instructs the domain 201 to perform an emergency stop.
The OS panic instruction unit 353 instructs the OS 401 to perform an emergency stop (panic).

  The HV dump target area notification processing unit 354 reads out and notifies the HV dump target area information from the HV dump target area information / HV dump flag storage area 356.

  The HV dump target area information / HV dump flag storage processing unit 355 stores the HV dump target area information and the HV dump flag in the HV dump target area information / HV dump flag storage area 356.

  The HV dump target area information / HV dump flag storage area 356 stores the HV dump target area information and the HV dump flag. The HV dump target area information is information indicating a memory area (HV dump target area) used by the hypervisor 351, and includes information on the start address (PA Base) and size of the memory area. The HV dump flag is control information indicating whether or not to generate a dump file in the memory area used by the hypervisor.

  The HV restart processing unit 357 stops the hypervisor 351 and restarts the hypervisor 351.

The OS restart command unit 358 instructs the OS 401 to restart.
The HV memory dump flag reading / transmitting unit 359 reads and transmits the HV dump flag.

  The HV dump target area read processing unit 360 reads and transmits the contents of the memory area indicated by the HV dump target area information. Alternatively, the HV dump target area read processing unit 360 reads and transmits the contents of the memory area used by the current hypervisor 351.

The memory management unit 361 manages the memory.
The memory release processing unit 362 performs a memory release process.

  The HV dump flag reset processing unit 363 resets the HV dump flag. Specifically, the HV dump flag reset processing unit 363 sets the HV dump flag to FALSE, for example.

The HV dump flag notification unit 364 notifies the HV dump flag.
The dump dedicated domain activation processing unit 365 activates the dump dedicated domain in the firmware mode. The firmware mode is a mode that does not start the OS, that is, a mode that stops before starting the OS.

  The kernel dump target area information / kernel dump flag storage processing unit 366 stores the kernel dump target area information and the kernel dump flag in the kernel dump target area information / kernel dump flag storage area 367.

  The kernel dump target area information / kernel dump flag storage area 367 stores kernel dump target area information and a kernel dump flag. The kernel dump target area information is information indicating a memory area (kernel dump target area) used by the kernel of the OS 401 at the time of panic, and includes information on the start address (RA Base) and size of the memory area. The kernel dump flag is information indicating whether or not to execute a kernel memory dump of the OS 401. Further, the kernel dump flag can indicate how the kernel memory dump is collected. The kernel dump flag is, for example, information such as 0: collection of a kernel memory dump, 1: collection in a dump collection domain, or 2: collection using a memory dynamic reconfiguration function. The kernel dump flag may be received from the OS 401, or may be set and held in advance by the hypervisor 351.

  The PA-RA mapping processing unit 368 performs mapping between a physical address (PA) and a real address (RA). PA is a physical address of the memory, and RA is a real address on the domain (operating system).

The PA-RA mapping information storage area 369 stores mapping information between PA and RA.
The interrupt processing unit 370 causes S401 to perform interrupt processing when performing a memory dump of the OS 401 kernel. If the interrupt process is accepted, it is determined that the OS 401 kernel memory dump is possible, and the process proceeds to the OS 401 kernel memory dump process. If the interrupt process is not accepted, it is determined that the interrupt cannot be performed and the OS 401 kernel memory dump process is not performed. finish.

  The memory dump process activation unit 371 causes the control domain 201-1 to activate the memory dump process of the hypervisor 351.

The memory initialization processing unit 372 initializes the memory.
The kernel dump flag reset processing unit 373 resets the kernel dump flag. For example, the kernel dump flag reset processing unit 373 deletes the kernel dump flag or sets it to “0: Do not collect a kernel memory dump”.

FIG. 6 is a detailed configuration diagram of the OS according to the embodiment.
The OS 401-m includes a memory management unit 402-m, a file management unit 403-m, a process management unit 404-m, an interrupt processing unit 405-m, a mapping information extraction / storage processing unit 406-m, and a mapping information storage area 407-. m, HV memory dump determination unit 408-m, OS startup processing unit 409-m, HV dump target area read processing unit calling unit 410-m, kernel dump target memory read processing unit 411-m, HV dump collection processing unit 412 -M, kernel dump collection processing unit 413-m, panic processing unit 414-m, kernel dump target area notification processing unit 415-m, memory DR incorporation processing unit 416-m, memory DR detachment processing unit 417-m, free memory Additional processing unit 418-m, dump dedicated domain stop processing unit 419-m, kernel dump flag reset processing unit 420-m, Comprising a fine kernel dump flag notifying unit 421-m.

The memory management unit 402-m allocates a memory 203-m used by the OS 401-m.
The file management unit 403-m manages files that are data stored on the disk.

  The process management unit 404-m manages a process of a program executed by the OS 401-m.

The interrupt processing unit 405-m performs interrupt processing.
The mapping information extraction / storage processing unit 406-m stores information necessary for acquiring and analyzing the dump of the memory 203-m in the mapping information storage area 407-m.

  The mapping information storage area 407-m stores information necessary for acquiring and analyzing the dump of the memory 203-m. The information stored in the mapping information storage area 407-m includes, for example, a kernel text area, a data area, a heap area, a stack area, etc., mapping information (logical address, physical address, size, etc.) of each segment, an address conversion table, This is mapping information of various control tables such as a page table.

  The HV memory dump determination unit 408-m determines whether the HV dump flag is TRUE or FALSE and determines whether to perform a memory dump of the hypervisor.

The OS activation processing unit 409-m restarts the OS 401-m.
The calling unit 410-m of the HV dump target area reading processing unit calls the HV dump target area reading processing unit 360.

  The kernel dump target memory read processing unit 411-m reads the memory contents of the kernel dump target area (the memory area used by the OS 401-m kernel at the time of panic).

  The HV dump collection processing unit 412-m receives the memory contents read by the HV dump target area read processing unit 360 from the HV dump target area read processing unit 360 and generates a dump file.

  The kernel dump collection processing unit 413-m saves the read memory contents of the kernel dump target area in a file (generates a dump file).

The panic processing unit 414-m makes an emergency stop (panic) of the domain 201-m.
The kernel dump target area notification processing unit 415-m notifies the hypervisor 351 of the memory area used by the OS 401-m kernel during a panic.

The memory DR incorporation processing unit 416-m incorporates a memory area into the domain 201-m.
The memory DR separation processing unit 417-m separates the memory area from the domain 201-m.

  The free memory addition processing unit 418-m notifies the memory management unit 402-m of the dumped memory area.

  The dump dedicated domain stop processing unit 419-m stops the domain (dump dedicated domain) from which the dump is collected after collecting the dump.

  The kernel dump flag reset processing unit 420-m instructs the hypervisor 351 to reset the kernel dump flag.

  The kernel dump flag notification unit 421-m notifies the hypervisor 351 of the kernel dump flag. The kernel dump flag notification unit 421-m notifies the hypervisor 351 of a kernel dump flag when it is necessary to execute a kernel memory dump. For example, if the kernel dump flag notifying unit 421-m does not collect the kernel memory dump, the kernel dump flag value is set to “0: kernel memory dump is not collected”, and the kernel memory dump is collected in the dump collecting domain. If the kernel dump flag value is "1: Collect in the dump collection domain", and the memory dynamic reconfiguration function is used to collect the kernel memory dump, the kernel dump flag value is "2: Memory dynamic reconfiguration function Use to collect ".

(First embodiment)
In the first embodiment, a memory dump of the hypervisor is collected using the control domain.

  7A and 7B are flowcharts of the memory dump generation process according to the first embodiment.

  In the initial state, it is assumed that the domains 201-1 to 201-3 and the OSs 401-1 to 401-3 are activated and in an operating state, and the domain 201-4 and the OS 401-4 are not activated.

In step S501, the hypervisor 351 detects a fatal error.
In step S502, the domain emergency stop instruction unit 352 instructs an emergency stop to the operating logical domain, that is, the control domain 201-1 and the domains 201-2 and 201-3.

  In step S503, the OS 401-i (i = 1 to 3) receives the emergency stop instruction and causes the OS 401-i to stop urgently.

  In step S504, the HV dump target area notification processing unit 354 reads out the HV dump target area information from the HV dump target area information / HV dump flag storage area 356, and notifies the firmware 311 of the information. The HV dump target area information is information indicating a memory area (dump target area) used by the hypervisor 351, and includes information on the start address (PA Base) and size of the memory area. The HV dump target area information has a format as shown in FIG. 8 and is associated with a block number, a block physical memory start address (PA Base), and a block size. Further, the HV restart processing unit 357 stops the hypervisor 351 (HV abort).

  In step S506, the dump target area information / HV dump flag storage processing unit 312 receives the HV dump target area information.

  In step S507, the dump target area information / HV dump flag storage processing unit 312 stores the received HV dump target area information as dump target area information in the dump target area information / HV dump flag storage area 313. Also, the HV dump flag setting unit 314 sets the HV dump flag to TURE and stores it in the dump target area information / HV dump flag storage area 313.

  In step S508, the firmware 311 starts the physical partition restart process while retaining the contents of the memory.

  In step S509, the memory initialization processing unit 315 starts memory initialization processing. First, for example, the top address of the memory is set as an area to be initialized.

  In step S510, the memory initialization processing unit 315 refers to the dump target area information, and determines whether the initialization target area is an area specified by the dump target area information, that is, a dump target area. If the initialization process target area is the dump target area, the control proceeds to step S512 while retaining the contents of the initialization process target area. If the initialization process target area is not the dump target area, the control proceeds to step S511.

  In step S511, the memory initialization processing unit 315 initializes a region to be initialized.

  In step S512, the memory initialization processing unit 315 determines whether initialization processing has been performed for all areas other than the dump target area. If initialization processing has been performed for all areas other than the dump target area, control proceeds to step S513. If initialization processing has not been performed for all areas other than the dump target area, an unprocessed area (for example, a dump area) The next address of the area that has been checked whether it is the target area) is set as the area to be initialized, and control returns to step S510.

  In step S513, the HV use area changing unit 316 changes the area used by the hypervisor 351 to an area other than the area indicated by the dump target area information. It should be noted that at least the changed area used by the hypervisor 351 may be set as the initialization target area.

  In step S514, the HV restart command unit 317 instructs the hypervisor 351 to restart. The dump target area information / HV dump flag notification unit 318 reads the dump target area information and the HV dump flag from the dump target area information / HV dump flag storage area 313 and notifies the hypervisor 351 of the dump target area information / HV dump flag storage area 313.

  In step S515, the HV dump target area information / HV dump flag storage processing unit 355 receives the dump target area information and the HV dump flag and stores them in the HV dump target area information / HV dump flag storage area 356. The HV dump target area information / HV dump flag storage processing unit 355 stores the received dump target area information as HV dump target area information.

  In step S516, the HV restart processing unit 357 restarts the hypervisor 351. However, the memory area specified by the HV dump target area information is not used.

  In step S517, the OS restart command unit 358 instructs the OS 401-1 to 401-3 to restart.

  In step S518, the OS restart processing units 409-2 and 409-3 restart the OS 401-2 and 401-3, respectively.

In step S519, the OSs 401-2 and 401-3 resume the business.
In step S520, the OSs 401-2 and 401-3 are in a normal operation state.

  In step S521, the OS restart processing unit 409-1 restarts the OS 401-1.

In step S522, the OS 401-1 resumes the business.
In step S523, the HV memory dump determination unit 408-1 requests the hypervisor 351 to transmit an HV dump flag.

  In step S524, when receiving the request, the HV memory dump flag reading / transmitting unit 359 reads out the HV dump flag from the HV dump target area information / HV dump flag storage area 356 and transmits it to the OS 401-1.

  In step S525, the HV memory dump determination unit 408-1 receives the HV dump flag and determines whether or not the HV dump flag is TRUE. If the HV dump flag is TRUE, the control proceeds to step S527, and if it is FALSE, the control proceeds to step S531.

  In step S526, the calling unit 410-1 of the HV dump target area reading processing unit calls the HV dump target area reading processing unit 360.

  In step S527, the HV dump target area read processing unit 360 reads the contents of the memory area indicated by the HV dump target area information and transmits the contents to the control domain.

  In step S528, the HV dump collection processing unit 412-1 receives the memory contents read by the HV dump target area reading processing unit 360 from the HV dump target area reading processing unit 360, and writes the received memory contents to a file for dumping. Generate a file. Thereafter, the processes of steps S529 and S530 and step S531 are executed in parallel.

  In step S529, the memory release processing unit 362 releases the memory area specified by the HV dump target area information. The HV dump flag reset processing unit 363 resets the HV dump flag, that is, sets it to FALSE. The HV dump flag notification unit 364 notifies the firmware 311 of the HV dump flag.

  In step S530, the memory release processing unit 320 clears the dump target area information. The HV dump flag reset processing unit 321 resets the HV dump flag, that is, sets it to FALSE.

In step S531, the OS 401-1 is in a normal operation state.
According to the memory dump generation processing according to the first embodiment, when an error is detected and the hypervisor and the operating system are restarted, copying is not performed to another memory or the like even if the size of the memory dump is large So you can quickly reboot the hypervisor and operating system. Thereby, business stop time can be shortened.

Here, a modification of the memory dump generation process according to the first embodiment will be described.
In the modified example, a memory dump of a running hypervisor is collected (referred to as a hypervisor live dump).

  FIG. 7C is a flowchart of a modification of the memory dump generation process according to the first embodiment.

  In the flowchart of the modified example, steps S532 and S533 are added to the flowchart of the memory dump generation process according to the first embodiment of FIGS. 7A and 7B, and if NO is determined in step S525, the control proceeds to step S532. It is something that goes on.

  In FIG. 7C, the changed part with respect to FIGS.

  In the modification, for example, the data structure of the HV dump flag can be changed to 0: not collected, 1: HV dump at the time of abnormality, 2: HV live dump. The HV memory dump determination unit 408-1 determines that the HV dump flag is TRUE when the HV dump flag is 1, and determines that the HV dump flag is FALSE when the HV dump flag is 0 or 2. In addition, when the HV dump flag is 2, the HV memory dump determination unit 408-1 determines that the HV dump live flag is TRUE.

  In step S532, the HV memory dump determination unit 408-1 determines whether or not the HV live dump flag is TRUE. When the HV dump flag is TRUE (that is, when the HV dump flag is 2), the control proceeds to step S533, and when it is FALSE, the control proceeds to step S531.

  In step S533, HV live dump processing is performed. Specifically, the calling unit 410-1 of the HV dump target area read processing unit calls the HV dump target area read processing unit 360. The HV dump target area read processing unit 360 reads the contents of the memory area used by the active hypervisor 351 and transmits it to the control domain. The HV dump collection processing unit 412-1 receives the memory contents read by the HV dump target area reading processing unit 360, and writes the received memory contents to a file to generate a hypervisor dump file.

  As described above, when collecting a memory dump of a running hypervisor, the data of the memory area used by the hypervisor is read out without stopping or restarting the hypervisor, and the data is filed as a hypervisor dump file. I am writing out.

(Second Embodiment)
In the second embodiment, an OS kernel memory dump is performed in addition to a hypervisor memory dump.

  9A, 9B, and 9C are flowcharts of the memory dump generation process according to the second embodiment.

  In the initial state, it is assumed that the domains 201-1 to 201-3 and the OSs 401-1 to 401-3 are activated and in an operating state, and the domain 201-4 and the OS 401-4 are not activated.

In step S601, a fatal error occurs in the hypervisor 351.
In step S602, the hypervisor 351 detects a fatal error.

  In step S603, the interrupt processing unit 370 notifies the operating OS, that is, the OS 401-i (i = 1 to 3) of the interrupt processing, and the OS panic instruction unit 353 instructs the OS 401-i to panic.

  In step S604, the panic processing unit 414-i receives the panic instruction and panics the OS 401-i.

  In step S605, the kernel dump target area notification processing unit 415-i notifies the hypervisor 351 of the kernel dump target area information. The kernel dump target area information is information indicating a memory area (dump target area) used by the OS 401-i kernel, and includes information on the start address (RA Base) and size of the memory area. The kernel dump target area information has a format as shown in FIG. 10, and is associated with a block number, a block memory start address (RA Base), and a block size.

  Steps S604 and S605 are executed for each logical domain that has received the panic instruction.

  In step S606, the PA-RA mapping processing unit 368 performs RA-PA conversion for converting the notified start address (RA Base) from the RA Base to the PA Base start address (PA Base).

  In step S <b> 607, the HV dump target area notification processing unit 354 notifies the firmware 311 of HV dump target area information indicating the memory area used by the hypervisor 351. Further, the HV dump target area notification processing unit 354 notifies the firmware 311 of the kernel dump target area information received from the OS 401-i. The notified kernel dump target area information includes the start address (PA Base) and size converted from RA Base to PA Base. In the embodiment, three pieces of kernel dump target area information corresponding to the stopped logical domain are notified.

  In step S608, the dump target area information / HV dump flag storage processing unit 312 uses the received HV dump target area information and the received kernel dump target area information as dump target area information as dump target area information / HV dump flag storage area. It stores in 313. Also, the HV dump flag setting unit 314 sets the HV dump flag to TURE and stores it in the dump target area information / HV dump flag storage area 313.

  In step S609, the HV restart processing unit 357 stops the hypervisor 351 (HV abort).

  In step S610, the memory initialization processing unit 315 initializes a memory area other than the area indicated by the dump target area information. That is, the memory initialization processing unit 315 initializes a memory area other than the area used by the hypervisor 351 and the area used by the kernel of the OS 401-i at the time of panic.

  In step S611, the HV usage area changing unit 316 changes the area used by the hypervisor 351 to an area other than the area indicated by the dump target area information. The HV restart command unit 317 instructs the hypervisor 351 to restart. The dump target area information / HV dump flag notification unit 318 reads the dump target area information and the HV dump flag from the dump target area information / HV dump flag storage area 313 and notifies the hypervisor 351 of the dump target area information / HV dump flag storage area 313. The dump target area information includes HV dump target area information and kernel dump target area information. The HV dump target area information / HV dump flag storage processing unit 355 receives the HV dump target area information and the HV dump flag in the dump target area information and stores them in the HV dump target area information / HV dump flag storage area 356. . The kernel dump target area information / kernel dump flag storage processing unit 366 receives the kernel dump target area information in the dump target area information and stores it in the kernel dump target area information / kernel dump flag storage area 367.

  In step S612, the HV restart processing unit 357 starts the hypervisor 351.

  In step S613, the memory initialization processing unit 372 initializes a memory area other than the area indicated by the kernel dump target area information.

  In step S614, the PA-RA mapping processing unit 368, the OS restart command unit 358, and the dump dedicated domain startup processing unit 365 check the value of the kernel dump flag. Thereafter, processing corresponding to the value of the kernel dump flag is executed. For example, the PA-RA mapping processor 368 dumps the memory PA used by the OS 401-1 to 401-3 kernel when a panic occurs when the kernel dump flag is “1: Collect in the dump collection domain”. Assigned to the RA of the domain 204-4.

  Hereinafter, the processes of step S621, steps S622 to S626, and steps S632 to S635 are separately performed in parallel.

  However, when the kernel dump flag is “1: collection with a dump collection domain”, steps S626 and S635 are not executed, and when “2: collection using the memory dynamic reconfiguration function” is performed, step S621 is not executed.

  Here, Step S621 is processing related to the dump collection domain 204-4, Steps S622 to S626 are processing related to the control domain 204-1, and Steps S632 to S635 are logical domains 204-2 and 204-3. It is processing regarding.

  In step S621, kernel memory dump generation processing by the dump collection domain is performed. Details of the kernel memory dump generation process by the dump collection domain will be described later.

In step S622, the PA-RA mapping processing unit 368 changes the mapping between the physical address (PA) and the real address (RA) of the domain 201-1 as shown in 1) and 2) below. As a result, even if the OS 401-1 is restarted, the data in the memory area used by the panic kernel and the panic hypervisor 351 is not overwritten.
1) Do not assign the physical address of the memory used by the kernel and hypervisor when a panic occurs to the real address of the domain to be restarted. And,
2) Make sure that the memory size that can be used by the domain does not change before and after the restart.

  However, if the physical memory that can be allocated to the domain to be restarted is less than the predetermined value, priority is given to 1).

  In addition, it is determined by referring to the HV dump target area information and the kernel dump target area information which area is used by the kernel and the hypervisor 351 when the panic occurs.

  In step S623, the OS restart command unit 358 instructs the OS 401-1 to restart. Further, when the kernel dump flag is “2: Collected using the memory dynamic reconfiguration function”, the OS restart instruction unit 358 includes in the restart instruction that the kernel dump process using the memory DR function is performed. . The OS activation processing unit 409-1 that has received the instruction activates the OS 401-1.

In step S624, the OS 401-1 resumes the business.
In step S625, a memory dump generation process of the kernel using the memory dynamic reconfiguration (DR) function is performed. Details of the kernel memory dump generation process using the memory DR function will be described later. In step S626, a hypervisor memory dump generation process is performed. Step S626 is the same as the processing of steps S523 to S531 in FIG.

In step S632, the PA-RA mapping processing unit 368 changes the mapping between the physical address (PA) and the real address (RA) of the domains 201-2 and 201-3 as shown in 1) and 2) below. As a result, even if the OS 401-1 is restarted, the data in the memory area used by the panic kernel and the panic hypervisor 351 is not overwritten.
1) Do not assign the physical address of the memory used by the kernel and hypervisor when a panic occurs to the real address of the domain to be restarted. And,
2) Make sure that the memory size that can be used by the domain does not change before and after the restart.

  However, if the physical memory that can be allocated to the domain to be restarted is less than the predetermined value, priority is given to 1).

  In addition, it is determined by referring to the HV dump target area information and the kernel dump target area information which area is used by the kernel and the hypervisor 351 when the panic occurs.

  In step S633, the OS restart command unit 358 instructs the OSs 401-2 and 401-3 to restart. Further, when the kernel dump flag is “2: Collected using the memory dynamic reconfiguration function”, the OS restart instruction unit 358 includes in the restart instruction that the kernel dump process using the memory DR function is performed. . The OS activation processing units 409-2 and 409-3 that have received the instruction activate the OSs 401-2 and 401-3, respectively.

In step S634, the OSs 401-2 and 401-3 each resume business.
In step S635, a kernel memory dump generation process using the memory DR function is performed.

Details of the kernel memory dump generation process will be described below.
The memory dump generation process of the kernel includes (1) a method of collecting a memory dump by the dump collection domain (step S621), or (2) a method of collecting a memory dump using the memory dynamic reconfiguration function (steps S626 and S635). Either of these is used.

(1) Method of collecting a memory dump in the dump collection domain The dump collection domain 201-4 does not need to be prepared for each logical domain even in a system in which a plurality of domains 201 exist. It ’s fine. When there is one dump collection domain 201-4, if a plurality of logical domains 201 panic simultaneously, a memory dump will be collected one domain at a time, regardless of whether dump collection is completed or not. There is no impact on operations because operations can be resumed promptly.

In the dump collection domain 201-4, it is not necessary to take over the business of the logical domain in which the panic has occurred, and therefore, the following hardware resources necessary for collecting the memory dump may be provided.
-The physical memory area used by the OS kernel of the logical domain where the panic occurred during the panic-One or more CPUs
-Disk to store the dump file and I / O resources required to use the disk

  FIG. 11 is a flowchart of a kernel memory dump generation process by the dump collection domain.

FIG. 11 corresponds to step S621 in FIG. 9B.
Here, the kernel memory dump generation process of the OS 401-i will be described.

  In step S651, the dump dedicated domain activation processing unit 365 activates the dump collection domain 201-4 in the firmware mode. The firmware mode is a mode that does not start the OS, that is, a mode that stops before starting the OS. Prevents the dump target area from being rewritten by not starting the OS.

  In step S652, the kernel dump target memory read processing unit 411-4 reads a memory area (kernel dump target area) used by the kernel of the operating system 401-i when a panic occurs. Information on the kernel dump target area (start address (RA Base), size, etc.) is obtained by notification from the firmware 311 or the hypervisor 351.

  In step S653, the kernel dump collection processing unit 413-4 writes the read memory contents to a file to generate a dump file.

  In step S654, the dump dedicated domain stop processing unit 419-4 stops the dump collection domain 201-4. Then, the dump dedicated domain stop processing unit 419-4 notifies the memory management unit 361 of the hypervisor 351 so that the kernel dump target area can be used as an unused memory, that is, a free memory. The kernel dump flag reset processing unit 420-4 instructs the hypervisor 351 to reset the kernel dump flag. The kernel dump flag reset processing unit 373 that has received the reset instruction resets the kernel dump flag.

  In step S655, the memory management unit 361 sets the kernel dump target area as a free memory that can also be used from the other logical domains 201-i.

FIG. 12 is a diagram illustrating collection of a memory dump by the dump collection domain.
The left side of FIG. 12 shows the operating state (and panic), the middle is when the system is restarted, and the right side is when the memory dump is collected by the dump collection domain.

  Here, the processing of the logical domain 201-1 is described. In addition, since the same process is performed also in the domains 201-2 and 201-3, the details are omitted.

  In the operating state on the left side of FIG. 12, an area with PA is mapped to an area with RA in logical domain 201-1.

  The area used by the OS 401-1 kernel when the OS 401-1 panics becomes the kernel dump target area.

  After the OS 401-1 panic, the PA-RA mapping is changed (step S <b> 622), and the logical domain 201-1 is an area used by the OS 401-1 kernel at the time of the panic (kernel dump target area). A different PA area is allocated, and the OS 401-1 restarts (middle of FIG. 12).

  At the time of dumping on the right side of FIG. 12, the PA area (kernel dump target area) used by the kernel of the OS 401-1 during the panic is allocated to the RA of the dump collection dedicated domain 201-4. The dump collection dedicated domain 201-4 reads the kernel dump target area and generates a dump file.

  FIG. 13 is a diagram showing PA-RA mapping information in memory dump collection by the dump collection domain.

  The left side of FIG. 13 shows the operating state (and during panic), the middle is during dumping, and the right side is after dumping.

  Here, the PA-RA mapping of the domain 201-1 (control domain # 0) and the dump collection dedicated domain 201-4 (dump collection dedicated domain # 3) is described.

  In the PA-RA mapping information, a domain, a start address (PA Base), a size, and a start address (RA Base) are described in association with each other.

  At the time of the panic on the left side of FIG. 13, the area where the start address (PA Base) is xxxxx and the size is 8 GB is mapped to the area where the start address (RA Base) of the control domain # 0 is aaaaa (FIG. 12). Corresponding to the left side). This area is the kernel dump target area.

  After the OS 401-1 panic, the PA-RA mapping is changed (step S <b> 622), and the PA-RA mapping information is as shown in the middle of FIG. 13.

  At the time of dumping in the middle of FIG. 13, the area where the start address (PA Base) is xxxxx and the size is 8 GB is mapped to the area where the start address (RA Base) of the dump collection dedicated domain # 3 is aaaaa. That is, the PA area of the control domain # 0 at the time of the panic is mapped to the RA of the dump collection dedicated domain # 3. In addition, an area where the start address (PA Base) is yyyyy and the size is 8 GB is mapped to an area where the start address (PA Base) of the control domain # 0 is aaaaa. That is, a new PA area is assigned to the control domain # 0 after restart (corresponding to the right side of FIG. 12).

  After the dump file is generated, the kernel dump target area becomes free memory that can be used from other domains (step S655).

  That is, after dumping on the right side of FIG. 13, the mapping information of the dump collection dedicated domain # 3 is deleted.

  According to the method of collecting a memory dump in the dump collection domain, the dump is collected in another domain instead of the domain in which the abnormality was detected. The possibility of the next damage occurring is reduced.

  According to the method of collecting a memory dump in the dump collection domain, you are charged according to the amount and time of hardware resources (CPU, memory, disk, etc.) used by the user, such as Capacity on Demand (CoD). In the system, it is easy to avoid charging for the hardware resources used for dump collection, and it is possible to optimize the charge.

(2) Method of Collecting Memory Dump Using Memory Dynamic Reconfiguration Function Here, the process of the logical domain 201-1 (step S625) will be described. Note that the same processing is executed for the processing of the logical domains 201-2 and 201-3 (step S635), and the details are omitted.

  FIG. 14 is a flowchart of a kernel memory dump generation process using the memory dynamic reconfiguration function.

FIG. 14 corresponds to step S625 in FIG. 9C.
In step S641, the memory DR incorporation processing unit 416-1 uses the dynamic reconfiguration function of the memory to assign the memory area (kernel dump target area) used by the kernel of the OS 401-1 at the time of the panic to the domain 201-1. Incorporate into. Information on the kernel dump target area (start address (RA Base), size, etc.) is obtained by notification from the firmware 311 or the hypervisor 351.

  In step S642, the kernel dump target memory read processing unit 411-1 reads the incorporated memory area.

  In step S643, the kernel dump collection processing unit 413-1 writes the read memory contents to a file to generate a dump file.

  In step S644, the memory DR detachment processing unit 417-1 uses the dynamic reconfiguration function of the memory to detach the memory area used by the kernel of the OS 401-1 from the domain 201-1 when the panic occurs. The memory management unit 361 is notified so that the area is free memory. The kernel dump flag reset processing unit 420-1 instructs the hypervisor 351 to reset the kernel dump flag. The kernel dump flag reset processing unit 373 that has received the reset instruction resets the kernel dump flag.

  In step S645, the memory management unit 361 sets the separated area as a free memory that can be used from the other domains 201-2 and 201-3.

  Further, instead of steps S644 and S645, the free memory addition processing unit 418-1 uses an unused memory that can use a memory area (that is, a dumped area) used by the kernel of the OS 401-1 at the time of a panic, In other words, the memory management unit 402-1 may be notified so that the free memory is used, and the memory management unit 402-1 may perform processing for setting the dumped area as free memory.

  FIG. 15 is a diagram illustrating collection of a memory dump using the memory dynamic reconfiguration function.

  The left side of FIG. 15 shows the operating state (and panic), the middle shows when the system is restarted, and the right shows the time when the memory dump is collected by the dump collection domain.

  Here, the processing of the domain 201-1 is described. In addition, since the same process is performed also in the domains 201-2 and 201-3, the details are omitted.

  In the operation state on the left side of FIG. 15, an area with PA is mapped to an area with RA in domain 201-1.

  The area used by the OS 401-1 kernel when the OS 401-1 panics becomes the kernel dump target area.

  After the OS 401-1 panic, the PA-RA mapping is changed (step S 622), and the RA of the domain 201-1 includes an area (kernel dump target area) used by the OS 401-1 kernel during the panic. Are assigned different PA areas, and the OS 401-1 restarts (middle of FIG. 15).

  At the time of dumping after restart on the right side of FIG. 15, the area used by the kernel of the OS 401-1 (kernel dump target area) at the time of panic is incorporated in the RA of the domain 201-1. The domain 201-1 reads the kernel dump target area and generates a dump file.

  FIG. 16 is a diagram showing PA-RA mapping information in collecting a memory dump using the memory dynamic reconfiguration function.

  The left side of FIG. 16 shows the operating state (and during panic), the middle is during dumping, and the right side is after dumping.

  Here, the PA-RA mapping of the domain 201-1 (control domain # 0) is described.

  In the PA-RA mapping information, a domain, a start address (PA Base), a size, and a start address (RA Base) are described in association with each other.

  At the time of panic on the left side of FIG. 16, the area where the start address (PA Base) is xxxxx and the size is 8 GB is mapped to the area where the start address (RA Base) of the control domain # 0 is aaaaa (FIG. 15). Corresponding to the left side). This area is the kernel dump target area.

  After the OS 401-1 panic, the PA-RA mapping is changed, and the kernel dump target area is incorporated in the control domain # 0, and the PA-RA mapping information is as shown in the middle of FIG. 16.

  At the time of dumping in the middle of FIG. 16, the area where the start address (PA Base) is yyyyy and the size is 8 GB is mapped to the area where the RA start address (RA Base) of the control domain # 0 is aaaaa. Further, an area having a start address (PA Base) of xxxxx and a size of 8 GB is mapped to an area having a start address (RA Base) of control domain # 0 of bbbbb.

  That is, a new PA area is allocated to the control domain # 0 after the reboot, and further, after the reboot of the control domain # 0, the kernel dump target area is incorporated into the control domain # 0 (corresponding to the right side in FIG. 15). .

  After the dump file is generated, the kernel dump target area becomes free memory that can be used from other domains (step S645).

  That is, the mapping information of the kernel dump target area is deleted after dumping on the right side of FIG.

  According to the method of collecting a memory dump using the memory dynamic reconfiguration function, not the operating system that detected the abnormality but the new operating system after the restart collects the dump, so the abnormality is detected again during the dump collection process. The possibility of secondary damage such as hang-up is reduced.

  According to the memory dump generation processing according to the second embodiment, when an error is detected and the hypervisor and the operating system are restarted, copying is not performed to another memory or the like even if the size of the memory dump is large So you can quickly reboot the hypervisor and operating system. Thereby, business stop time can be shortened.

  According to the memory dump generation processing according to the second embodiment, by collecting the memory dump of the hypervisor and the kernel, even if the error is caused by both the hypervisor and the domain, the error analysis is effectively performed. It can be performed.

(Third embodiment)
In the third embodiment, an error is detected by the OS and a kernel memory dump is performed.

Here, a case where a kernel memory dump of the OS 401-1 is generated will be described.
FIG. 17 is a flowchart of a memory dump generation process according to the third embodiment.

  First, the memory management unit 402-1 allocates the memory used by the kernel from the smallest (or largest) real address (RA) of the memory 203-1, when the OS 401-1 is activated. In this way, the size of the memory area (dump target area) used by the kernel is made as small as possible. Further, the mapping information extraction / storage processing unit 406-1 is information necessary for collecting / analyzing a dump of the memory used by the kernel (for example, a text area of the kernel, a data area, a heap area, a stack area). For example, mapping information (logical address, physical address, size, etc.), address conversion table, page table, and various control table mapping information) of each segment is written in the mapping information storage area 407-1. Further, it is assumed that the domains 201-1 to 201-3 and the OSs 401-1 to 401-3 are activated and in an operating state, and the domain 201-4 and the OS 401-4 are not activated.

In step S701, a fatal error occurs in the OS 401-1.
In step S702, the OS 401-1 detects a fatal error.

  In step S703, the panic processing unit 414-1 panics (emergency stop) the OS 401-1.

  In step S704, the kernel dump target area notification processing unit 415-1 obtains information (kernel dump target area information) on the memory area (kernel dump target area) used by the kernel of the OS 401-1 at the time of emergency stop (panic). Notify the hypervisor 351. In addition, the kernel dump flag notification unit 421-1 notifies the hypervisor 351 of the kernel dump flag.

  In step S705, the kernel dump target area information / kernel dump flag storage processing unit 366 stores the received kernel dump target area information and kernel dump flag in the kernel dump target area information / kernel dump flag storage area 367.

  In step S706, the memory initialization unit 372 initializes a memory area other than the area indicated by the kernel dump target area information. That is, the memory initialization unit 372 does not perform initialization processing of the memory area used by the kernel of the OS 401-1 at the time of panic (that is, does not update data). As a result, the data in the memory area used by the OS 401-1 kernel at the time of panic remains as it is.

  In step S707, the PA-RA mapping processing unit 368, the OS restart command unit 358, and the dump dedicated domain startup processing unit 365 check the value of the kernel dump flag. Thereafter, processing corresponding to the value of the kernel dump flag is executed. For example, the PA-RA mapping processor 368 dumps the memory PA used by the OS 401-1 to 401-3 kernel when a panic occurs when the kernel dump flag is “1: Collect in the dump collection domain”. Assigned to the RA of the domain 204-4.

  Thereafter, the processes of step S708 and steps S709 to S712 are executed separately in parallel.

  However, step S712 is not executed when the kernel dump flag is “1: collection with dump collection domain”, and step S708 is not executed when “2: collection using the memory dynamic reconfiguration function”.

  In step S708, kernel memory dump generation processing by the dump collection domain is performed. Step S708 is similar to the process of step S621 in FIG.

In step S709, the PA-RA mapping processing unit 368 changes the mapping between the physical address (PA) and the real address (RA) of the panicked domain 201-1 as shown in 1) and 2) below. Thereby, even if the OS 401-1 is restarted, the data in the memory area used by the kernel of the OS 401-1 at the time of the panic is not overwritten.
1) Do not assign the physical address of the memory used by the kernel when a panic occurs to the real address of the domain to be restarted. And,
2) Make sure that the memory size that can be used by the domain does not change before and after the restart.

  However, if the physical memory that can be allocated to the domain to be restarted is less than the predetermined value, priority is given to 1).

  In step S710, the OS restart command unit 358 instructs the domain 201-1 to restart the OS 401-1. Further, when the kernel dump flag is “2: Collected using the memory dynamic reconfiguration function”, the OS restart instruction unit 358 includes in the restart instruction that the kernel dump process using the memory DR function is performed. . The OS startup processing unit 409-1 restarts the OS 401-1, without writing the memory dump used by the kernel to a disk or the like.

In step S711, the OS 401-1 resumes the business.
In step S712, kernel memory dump generation processing is performed. Note that the processing in step S712 is the same as the processing in step S625 described above, and a description thereof will be omitted.

  According to the memory dump generation processing according to the third embodiment, when an error is detected and the operating system is urgently stopped (panic), even if the size of the memory dump is large, copying is not performed to another memory or the like So you can quickly restart the operating system. Thereby, business stop time can be shortened.

(Fourth embodiment)
In the fourth embodiment, an error is detected by the OS, a kernel memory dump is performed, and a memory dump of a running hypervisor is collected (referred to as a hypervisor live dump).

Here, a case where a kernel memory dump of the OS 401-1 is generated will be described.
18A and 18B are flowcharts of a memory dump generation process according to the fourth embodiment.

  Steps S801 to S811 are the same processes as steps S701 to S711 in FIG.

Hereinafter, step S812 and step S813 are executed in parallel.
In step S812, kernel memory dump generation processing is performed. Note that the processing in step S812 is the same as the processing in step S625 described above, and a description thereof will be omitted.

  In step S813, memory dump generation processing of the active hypervisor 351 is performed in the control domain 204-1.

  Hereinafter, details of the memory dump generation processing of the operating hypervisor 351 will be described.

FIG. 19 is a flowchart of the memory dump generation process of the hypervisor that is operating.
FIG. 19 corresponds to step S813 in FIG. 18B.

  In the fourth embodiment, for example, the data structure of the HV dump flag can be changed to 0: not collected, 1: HV dump at the time of abnormality, 2: HV live dump. When the HV dump flag is 0 or 1, the HV memory dump determination unit 408-1 determines not to collect the HV live dump, and when the HV dump flag is 2, the HV memory dump determination unit 408-1 determines to collect the HV live dump.

  In step S831, the HV memory dump determination unit 408-1 requests the hypervisor 351 to transmit an HV dump flag.

  In step S832, when receiving the request, the HV memory dump flag reading / transmitting unit 359 reads out the HV dump flag from the HV dump target area information / HV dump flag storage area 356 and transmits it to the OS 401-1.

  In step S833, the HV memory dump determination unit 408-1 determines whether to collect a live dump of the active hypervisor 351 based on the received HV dump flag. If it is determined that a live dump of the active hypervisor 351 is to be collected, the control proceeds to step S834, and if it is determined not to be collected, the process ends.

  In step S834, the calling unit 410-1 of the HV dump target area reading processing unit calls the HV dump target area reading processing unit 360.

  In step S835, the HV dump target area read processing unit 360 reads the memory area currently used by the hypervisor 351, and transmits the read memory contents to the control domain 204-1.

  In step S836, the HV dump collection processing unit 412-1 receives the memory contents, and writes the received memory contents to a file to generate a hypervisor dump file.

  As described above, in the process of generating a memory dump of a running hypervisor, data in the memory area used by the hypervisor is read without stopping or restarting the hypervisor, and the data is used as a dump file for the hypervisor. Exporting to a file.

  According to the memory dump generation processing according to the fourth embodiment, when an error is detected and the operating system is restarted, even if the memory dump size is large, copying is not performed to another memory or the like. You can restart the operating system. Thereby, business stop time can be shortened.

  According to the memory dump generation process according to the fourth embodiment, by collecting the memory dump of the hypervisor and the kernel, even if the error is caused by both the hypervisor and the domain, the error analysis is effectively performed. It can be performed.

  Although a plurality of embodiments have been described above, the embodiments are not limited to apparatuses and methods, and may be configured as programs, or may be configured as a computer-readable recording medium storing the programs. I can do it. As the recording medium, for example, a flexible disk (FD), a hard disk drive, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a DVD-ROM, a DVD-RAM, a magnetic tape, a nonvolatile memory card, and the like are used. .

  For example, the program according to the embodiment is read from a recording medium storing the program and stored in the memories 13 and 23 and the nonvolatile memory 14. The CPUs 12 and 22 read out and execute programs from the memories 13 and 23 and the nonvolatile memory 14 to execute the various processes of the above-described embodiment.

  The present invention is not limited to the embodiments described above, and various configurations can be made without departing from the gist of the present invention. For example, the number of logical domains is not limited to four, and can be any number.

10 Server 11 System board 12 CPU
13 Memory 14 Non-volatile Memory 21 Service Processor 22 CPU
23 Memory 31 Disk Unit 32 Hard Disk Drive 41 Communication Interface 51 Bus 61 Physical Partition 201 Logical Domain 202 CPU
203 Memory 204 Disk 311 Firmware 312 Dump target area information / HV dump flag storage processing section 313 Dump target area information / HV dump flag storage area 314 HV dump flag setting section 315 Memory initialization processing section 316 HV use area change section 317 HV re- Start command section 318 Dump target area information / HV dump flag notification section 319 PA-RA mapping notification section 320 Memory release processing section 321 HV dump flag reset processing section 351 Hypervisor 352 Domain emergency stop instruction section 353 OS panic instruction section 354 HV dump Target area notification processing section 355 HV dump target area information / HV dump flag storage processing section 356 HV dump target area information / HV dump flag storage area 357 HV restart processing section 358 OS restart command section 359 HV memory dump read / transmit section 360 HV dump Memory read processing unit 361 Memory management unit 362 Memory release processing unit 363 HV dump flag reset processing unit 364 HV dump flag notification unit 365 Dump dedicated domain activation processing unit 366 Kernel dump target area information / kernel dump flag storage processing unit 367 Kernel dump Target area information / kernel dump flag storage area 368 PA-RA mapping processing section 369 PA-RA mapping information storage area 370 Interrupt processing section 371 Memory dump processing activation section 372 Memory initialization processing section 373 Kernel dump flag reset processing section 401 Operating system 402 Memory Management Unit 403 File Management Unit 404 Process Management Unit 405 Interrupt Processing Unit 406 Mapping Information Extraction / Storage Processing Unit 407 Mapping Information Storage Area 408 HV Memory Dump Judgment Unit 409 OS activation processing unit 410 Calling unit of HV dump target area reading processing unit 411 Kernel dump target memory reading processing unit 412 HV dump collection processing unit 413 Kernel dump collection processing unit 414 Panic processing unit 415 Kernel dump target area notification processing unit 416 Memory DR incorporation processing unit 417 Memory DR separation processing unit 418 Free memory addition processing unit 419 Dump dedicated domain stop processing unit 420 Kernel dump flag reset processing unit 421 Kernel dump flag notification unit

Claims (8)

A firmware having a storage device and an arithmetic processing device, and performing processing of a hypervisor that controls a virtual machine on which an operating system operates, and control of a system including the memory and the processor by the arithmetic processing device and the storage device In the information processing apparatus in which each process is executed,
The hypervisor is
If it detects an error in the running operating system, stop the operating system in which the error was detected,
The firmware is
Stop the hypervisor that controls the virtual machine running the operating system where the error is detected,
Changing the memory area used by the stopped operating system to a second memory area different from the first memory area used by the kernel of the stopped operating system;
Start the stopped hypervisor,
The activated hypervisor is
Activating the stopped operating system using the second memory area as a use area;
The booted operating system is
Resume the stopped program in response to the operating system shutdown that detected the error,
An information processing apparatus that executes a process of writing data read from a memory area used by the hypervisor to a file as a dump file of the hypervisor while the hypervisor that controls the virtual machine is operating. In the information processing apparatus,
The booted operating system is further
The information processing apparatus according to claim 1, wherein a process of writing data read from the first memory area to a file as a dump file of an operating system is executed. The information processing apparatus further includes:
Divided into a plurality of physical partitions each having one or a plurality of system boards each having the storage device and the processing unit;
The information processing apparatus according to claim 1, wherein each of the plurality of physical partitions includes a logical domain corresponding to a virtual machine on which an operating system operates, a hypervisor that controls the logical domain, and the firmware. The firmware further includes
After stopping the hypervisor, the memory area used by the hypervisor is changed to the second memory area while retaining the contents of the first memory area, and at least the changed second memory area The information processing apparatus according to any one of claims 1 to 3, wherein the hypervisor is activated after initialization. The firmware further includes
Stores flag information indicating whether to dump the area used by the hypervisor,
5. If the flag information indicates that the area used by the hypervisor is to be dumped, the data called from the first memory area is written to a file as a hypervisor dump file. The information processing apparatus according to claim 1. The hypervisor further includes
The information processing apparatus according to claim 5, wherein another logical domain that writes data read from the first memory area to a file as a hypervisor dump file is executed based on the flag information. The firmware further includes
The information processing apparatus according to claim 6, wherein the flag information is initialized after the data is written in a file as a hypervisor dump file. The firmware further includes
The information processing apparatus according to any one of claims 1 to 7, wherein the first memory area is released as a resource after the data is written in a file as a hypervisor dump file.