JP5379719B2 - Computer, computer failure detection method, and program - Google Patents

Description

  The present invention relates to a computer, a computer fault detection method, and a program, and in particular, a computer capable of quickly detecting a fault that has occurred in a computer before reaching abnormal termination, etc., and a computer fault detection method, and Regarding the program.

  A computer called a High Availability System (hereinafter referred to as “HA System”) is a system designed so that it can be operated continuously without stopping even if a failure occurs, or a failure. This system is designed so that the time until recovery is shortened as much as possible even if it is forced to stop. For this reason, in the HA system, for example, by making the main part of the functional block redundant, the failure occurrence site is separated when a failure occurs, and the operation rate is increased by continuing operation using a normal site. In order to realize such a failure response, it is important to configure so that the failure occurrence site in the system can be detected quickly and appropriately.

  In order to detect computer failures, hardware has developed failure detection functions and redundancy in individual elements. On the other hand, the failure detection function in each program and redundancy by HA cluster software are applied to the software as well.

  In Patent Document 1, for the purpose of detecting an operation delay when a plurality of programs are linked by an event, an event to be monitored is specified, and standby and notification for the event notification specified as the monitoring target are specified. A configuration has been proposed in which the operation is recorded, the execution restart of the waiting thread is recorded, and this record is inspected to detect that there is a thread that does not resume running despite being notified of the event. .

JP 2007-72958 A

However, in Patent Document 1, it is considered that a failure that occurs when a single program is executed cannot be dealt with.
Furthermore, in open systems realized by combining a large number of softwares today, a large number of people and organizations are involved in development, and design standards may differ between them. It may be difficult to maintain a uniform thickness.

  Software faults can be broadly classified into three types: abnormal termination, error output, and slowdown / hangup. Abnormal termination means that processing of a program or application does not end in a normal state due to a bug in the program. Error output means that an error is detected during program execution and output to a user or the like through an output screen or the like. Slowdown / hangup is a phenomenon in which the processing speed of the computer decreases or the processing stops due to some cause. In the user interface, the response speed to the operation input by the mouse etc. decreases, the screen freezes, etc. It may be grasped as a phenomenon.

  For faults with abnormal termination or error output, even if the software design level is not uniform, as long as the developer can recognize the events that cause them as abnormal, the detection function for them will be implemented separately. , Not much problem.

  However, slowdown / hangup can occur in various parts of the system, so even if it is handled by individual hardware and software, it will still be a problem when multiple systems work together. Contains the potential to cause. In some cases, a design that does not detect a fault that unduly extends the processing time of the processor may be employed. Therefore, it cannot be said that the slowdown / hangup detection has been sufficiently performed. Accordingly, even when system service provision is stopped due to slowdown / hangup, this may not be automatically detected, and system downtime may occur.

  The present invention has been made in view of the above circumstances, and one object thereof is a computer capable of quickly detecting a failure that has occurred in a computer before abnormal termination or the like, and computer failure detection. A method and program are provided.

To achieve the above and other objects, one aspect of the present invention provides a plurality of processes comprising a processor and a memory, and constituting at least one software program stored in the memory by the processor. Is the time for which the processor is processing the process from the start of processing by the processor to the end of processing for each of the processes. The processor usage time and the processor non-use time, which is the time during which the processor has stopped processing, are obtained by sequentially measuring a plurality of times, and according to a predetermined statistical process, a statistical standard for each processor usage time Processor usage time reference value that is a value and each processor A processor nonuse time reference value is a statistical measure value of use time, and the processor processes the reference value acquisition unit that calculates and stores for each said process, any of the plurality of processes that make up the software program by the processor When the process is being processed, the processor use time and the processor non-use time are measured for the process, and the processor use time reference value and the processor non-use time reference value that are sequentially stored for the process If the comparison result is determined not to satisfy the predetermined criterion, it is determined that a failure has occurred during the process processing, and the measured processor non-use time is the predetermined first processor failure. It is longer than the usage time threshold, and the processor usage time at that time is a predetermined process. If it is determined shorter than the working time threshold, the computer, characterized in that a transition time and a said the determining reference value comparison processing unit prescribed shorter made fault by the processor using time threshold has occurred in the process is there.

  According to the present invention having the above-described configuration, a computer, a computer failure detection method, and a program capable of quickly detecting a failure that has occurred in a computer before reaching abnormal termination or the like are provided.

FIG. 1 is a schematic diagram showing how to grasp a process execution state in a CPU for realizing a failure detection method according to the present invention. FIG. 2 is a schematic diagram showing an outline of a fault that recognizes an unusual state. FIG. 3 is a schematic diagram showing an outline of a failure that takes an abnormally long time for state transition. FIG. 4 is a schematic diagram showing an outline of a failure in which a certain state is abnormally long. FIG. 5 is an overall view of the hardware configuration of the computer 1 according to the embodiment of the present invention. FIG. 6 is a diagram illustrating an example of a software configuration of the computer 1. FIG. 7 is a diagram illustrating an example of the process management table 700. FIG. 8 is a diagram illustrating an example of the statistical information management table 800. FIG. 9 is a diagram illustrating an example of the temporary storage table 900. FIG. 10 is a diagram showing another example corresponding to the process management table 700 of FIG. FIG. 11 is a diagram illustrating an example of the statistical information management table 800 regarding the statistical information B. FIG. 12 is a diagram illustrating an example of a process flow of the state transition monitoring process. FIG. 13 is a diagram illustrating an example of a processing flow of statistical information collection processing. FIG. 14A is a diagram illustrating an example of a processing flow of statistical information comparison processing. FIG. 14B is a diagram illustrating an example of a processing flow of statistical information comparison processing.

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

<< Outline of Failure Detection Method in the Present Embodiment >>
The operation of general software (especially a service providing program) provides a service by executing a number of processes constituting the software while repeating certain state transitions in a processor that processes those processes. doing. A period during which the process has slept for a certain period (for example, 1 second) (processor non-use time) is defined as “state”, and a period during which the state changes (processor use time) is defined as “transition”. That is, a standby period from when a certain process sleeps until it starts to travel is referred to as “state”, and a period from when the process starts to sleep again is defined as “transition”. In the present specification, the term “process” is not limited to the case where UNIXOS is applied, but generally indicates an execution unit of a program executed by the OS.

  The concept of the failure detection method in the present embodiment will be described with reference to FIGS. FIG. 1 is a schematic diagram showing how to grasp a process execution state in a processor for realizing the failure detection method according to the present invention. Here, the state where the process is started and the processor is used (the process is running) is defined as “transition” 1001. A state that is paired with “transition” and the process has not used the processor for a certain period of time is defined as “state” 1002. A failure is detected by grasping a state transition formed by a pair of the transition 1001 and the state 1002 and capturing a behavior different from the normal state (for example, joining of the state 1003 in FIG. 1).

  FIG. 2 is a schematic diagram showing an outline of a fault that recognizes an unusual state. The case where the transition from the state 2001 to the state 2002 is unreasonably short unlike the normal state and a new transition 2003 to the state 2004 different from the normal state is recognized is illustrated. In the example of FIG. 2, it is assumed that the lock wait, which normally ends in a short time, is extended for some reason and the processing time of the entire process is delayed.

  FIG. 3 is a schematic diagram showing an outline of a failure that takes an abnormally long time for state transition. A transition 3003 from the state 3001 to the state 3002 takes a long time unlike the normal case, and does not transition to the state 3002. The example of FIG. 3 assumes a failure such as an infinite loop in which the processor is unreasonably occupied for a long time by another process.

  FIG. 4 is a schematic diagram showing an outline of a failure in which a certain state is abnormally long. In the state 1201, a case is shown in which the processor is not assigned for a longer time than in the normal state and the transition to the transition 1202 is not made. The example in FIG. 4 assumes a failure such as a hang-up or an illegal I / O wait.

  In the failure detection method according to the present embodiment, the time change of the above “state transition” is measured and measured inside an operating system (OS) operating on the processor, and the software failure is based on the statistical information. Is going to be detected. By monitoring and detecting faults with a fixed policy inside the OS, variations in design level on the installed software side are absorbed, and quick and reliable fault detection is realized. As a fixed policy, we focus on the state where the processor is not used (sleeping) by the process, and obtain statistical information about the time required for transition from one state to the next state to understand the operation of the process. The failure is detected by capturing the change.

  A specific failure detection mode will be described later with reference to a processing flow example of the failure detection method.

"System configuration"
Next, the computer 1 to which the failure detection method of this embodiment is applied will be described. FIG. 5 shows an overall view of the hardware configuration of the computer 1 according to an embodiment of the present invention.

  As shown in FIG. 5, the computer 1 can communicate with the central processing unit 110, the control unit 120, the main storage device 130, the auxiliary storage device 140, the input device 150, the output device 160, the communication control unit 170, and each other. And an internal bus 180 connected to the.

  The central processing unit 110 is a processor including, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). In the present specification, it is hereinafter referred to as “CPU” for simplicity. The control unit 120 is an interface that controls data transfer between the CPU 110 and other hardware blocks.

  The main storage device 130 is a memory having storage elements such as a RAM (Random Access Memory) and a ROM (Read Only Memory), for example, and, as will be described later, an OS, various programs, tables, and the like that control the entire computer 1 Data is read. The auxiliary storage device 140 is a storage device such as an HDD (Hard Disk Drive), a semiconductor storage device (Solid State Drive, “SSD”), and the like. Various programs executed by the computer 1 and tables referred to when the programs are executed. Kind is memorized.

  The input device 150 is, for example, a keyboard or a mouse, and accepts user operation input. The output device 160 is an output device such as a liquid crystal monitor and a printer, and includes a device that enables other output such as audio. The communication control unit 170 has a function of realizing communication with other devices including, for example, a network interface card (NIC), a host bus adapter (HBA), and the like.

  Next, the software configuration of the computer 1 will be described. FIG. 6 shows an example of the software configuration of the computer 1 in a state where each software is read into the main storage device 130 to be executed by the CPU 110. Note that these software are always stored in the auxiliary storage device 140, and are read into the main storage device 130 when the computer 1 is started or when a command is received from the OS based on a user operation input.

  In the example of FIG. 6, a program for realizing the OS 131, the program 132, and the failure detection unit 133 is installed in the computer 1 as software.

  The OS 131 controls the operation of each component constituting the computer 1. The OS 131 has a process scheduler 1311 (process allocation unit), reads a later-described program 132 stored in the auxiliary storage device 150 into the main storage device 130, and configures the loaded program 132 using the process scheduler 1311. The CPU 110 is assigned to the process being executed and the process is executed. As the OS 131, for example, a Windows (registered trademark), a UNIX (registered trademark) OS, or the like is preferably used, but the OS 131 is not particularly limited.

  The program 132 is an arbitrary program including an application program to be executed in the computer 1, and is read from the auxiliary storage device 140 into the main storage device 130 by the OS 131 and executed by the CPU 110 as described above. In the example of the present embodiment, the program 132 has two processes, process A (132A) and process B (132B). When the program 132 is executed by the CPU 110, the process scheduler 1311 assigns the CPU 110 to the processes A and B in a timely manner.

  The failure detection unit 133 is a functional block realized by the CPU 110 executing a program that realizes the function of the failure detection unit 133 (in FIG. 6, the failure detection unit 133 is read into the main storage device 130. Represented as a program). The failure detection unit 133 includes a state transition monitoring unit 1331, a statistical information collection processing unit (processor processing reference value acquisition unit) 1332, a statistical information storage unit 1333, a statistical information comparison processing unit (reference value comparison processing unit) 1334, and a failure notification. A portion 1335 is provided.

  The failure detection unit 133 is loaded from the auxiliary storage device 140 to the main storage device 130 by the OS 131 when the computer 1 is activated. After being loaded, the failure detection unit 133 is a control unit for the process to be detected by the process scheduler 1311 (process A (132A) and process B (132B) for executing the program 132 in the example of FIG. 6) by the process scheduler 1311. Information on the operation of the CPU 110 through 120 is acquired, and the statistical information collection processing unit 1332 stores the collected statistical information in the statistical information storage unit 1333.

  The failure detection unit 133 starts failure detection after the storage of the statistical information is completed. When executing failure detection, the statistical information comparison processing unit 1334 compares the information about the operation of the CPU 110 obtained from the state transition monitoring unit 1331 and the statistical information of the statistical information storage unit 1333, and a failure has occurred in any of the processes. Check if there is any. When the failure detection unit 133 confirms that a failure has occurred, the failure detection unit 133 notifies the OS 131 of the occurrence of the failure through the failure notification unit 1335. Upon receiving this notification, the OS 131 can present a warning message or the like to the administrator or the like through the output device 150.

  Next, tables generated and referred to by the failure detection unit 133 will be described. In this embodiment, the statistical information storage unit 1333 of the failure detection unit 133 holds a process management table 700, a statistical information management table 800, and a temporary storage table 900.

  FIG. 7 shows an example of the process management table 700. The process management table 700 is used to manage the status of statistical information related to the process 132 to be monitored by the failure detection unit 133, and each item of the monitoring target process 701, the statistical information collection completion flag 702, and the statistical information table 703. Is recorded. The process management table 700 is generated in the statistical information storage unit 1333 by the failure detection unit 133 when the computer 1 is started.

  In the monitoring target process 701, information for specifying the process 132 to be monitored by the failure detection unit 133 is recorded. In the example of FIG. 7, the monitoring target process A (132A) and the monitoring target process in the computer 1 are recorded. Since B (132B) is executed, both of them are recorded. In this case, when either process A or process B is finished executing, the record is deleted from the process management table 700. In addition to the processes A and B, if the CPU 110 is assigned to another process to be monitored (for example, the process C), a record related to the process C is added. In addition, in the monitoring target process 701, a process that an administrator or the like desires to be monitored can be registered from the input device 150.

  In the statistical information collection completion flag 702, information indicating whether statistical information used for determining whether or not a failure has been detected in the failure detection unit 133 is collected for each monitoring target process 701 is recorded. If the collection has been completed, “True” is recorded in the statistical information collection completion flag 702, and if the collection has not been completed, “False” is recorded.

  In the statistical information table 703, a table associated with each monitoring target process 701 as a statistical information table (described later) to be referred to by the failure detection unit 133 is specified and recorded. In the example of FIG. 7, since the statistical information used for failure detection has already been collected for the process A (132A), the statistical information A table (specifically, the subsequent statistical information is recorded later). And a corresponding statistical information management table 800). On the other hand, since the collection of statistical information has not been completed for the process B, a temporary storage table 900 (“temporary storage 1”), which indicates that statistical information is being collected, is displayed in the corresponding statistical information table 703. , “Temporary storage 2”).

  Next, the statistical information management table 800 will be described. FIG. 8 shows an example of the statistical information management table 800 in the present embodiment. The statistical information management table 800 records statistical information to be compared and referred to by the failure detection unit 133 described later for the monitoring target process. In the statistical information management table 800, items of a monitoring item 801, a transition time 802, and a state waiting time 803 are recorded. In the monitoring item 801, the “state transition” after the CPU 110 is assigned to the monitoring target process (in this case, process A (132A)) and the execution is started, that is, the combination of “transition” and “state” is, for example, the example of FIG. Then, “state transition 1” and “state transition 2” are registered in order.

  In the transition time 802 (processor usage time reference value), a numerical value statistically collected as the time required for the transition in the associated state transition is recorded. In the state waiting time 803 (processor non-use time reference value), a numerical value statistically collected as the time during which the state in the associated state transition has continued is recorded. The example of FIG. 8 shows that the state transition 1 is statistically obtained with a transition time of 500 ms and a state standby time of 1000 ms. FIG. 8 shows assumed numerical values, and the unit of numerical values may be other than milliseconds. Note that the statistical processing for obtaining the transition time 802 and the state waiting time 803 employed here applies an appropriate method such as processing for obtaining an arithmetic average, a center value, or a mode value of a plurality of measurement values. Just do it.

  Next, the temporary storage table 900 will be described. FIG. 9 shows an example of the temporary storage table 900. The temporary storage table 900 is generated in the statistical information storage unit 1333 by the failure detection unit 133 when the process management table 700 is generated and the monitoring target process is registered when the system is started. As shown in FIG. 7, a plurality of temporary storage tables 900 are generated in association with monitoring target processes (completion flag = “False”) for which statistical information collection has not been completed. This is because the transition time and the state waiting time are collected and recorded a plurality of times in order to obtain a numerical value that can be statistically reliable for the state transition of the monitored process. When the predetermined number of times is collected, the plurality of temporary storage tables 900 generated thereby are merged by a predetermined procedure corresponding to the above-described statistical processing to generate the statistical information management table 800 of FIG. .

  The monitoring item 901, the transition time 902, and the state standby time 903 recorded in the temporary storage table 900 are the same as the corresponding items in the statistical information management table 800.

  FIG. 10 shows another example corresponding to the process management table 700 of FIG. In the process management table 700 of FIG. 10, since statistical information collection has been completed for the process B as well, “True” is associated with the statistical information collection completion flag 702, and the item of the statistical information table 703 is associated with “ Statistical information B "is recorded.

  FIG. 11 shows an example of the statistical information management table 800 related to the statistical information B. The statistical information management table 800 in FIG. 11 is referred to by the failure detection unit 133 when executing the failure detection process for the process B. The contents recorded in the statistical information management table 800 in FIG. 11 are the same as those in FIG.

<< Contents of failure detection processing >>
Next, based on the system configuration described above, the failure detection processing executed by the failure detection unit 133 in the computer 1 will be described with reference to a processing flow example.

State Transition Monitoring Process First, the state transition monitoring process executed by the state transition monitoring unit 1331 of the failure detection unit 133 will be described. FIG. 12 shows an example of the process flow of the state transition monitoring process. In this state transition monitoring processing, whether or not the state transition monitoring unit 1331 mainly performs processing for collecting statistical information to be used for failure detection processing for each process stored as a monitoring target in the process management table 700. A process for determining whether to actually execute the failure detection process using the statistical information already acquired is performed.

  First, the state transition monitoring unit 1331 monitors the process scheduler 1311 of the OS 131, and determines whether the process assigned to the CPU 110 by the OS 131 and started processing is a process stored in the process management table 700 (S1201). When it is determined that the process that has been started is not stored in the process management table 700 (S1201, No), the state transition monitoring unit 1331 starts the process of the process that is the monitoring target in the process management table 700. Until the process scheduler 1311 continues to monitor.

  When it is determined that the process that has been started is stored in the process management table 700 (S1201, Yes), the state transition monitoring unit 1331 has a statistical information collection completion flag 702 associated with the started process. Is confirmed (S1202). When it is determined that “True” is recorded in the statistical information collection completion flag 702 (S1202, Yes), the state transition monitoring unit 1331 causes the statistical information comparison processing unit 1334 to execute statistical information comparison processing (S1203). The contents of the statistical information comparison process will be described later.

  When it is determined that “False” is recorded in the statistical information collection completion flag 702 (S1202, No), the state transition monitoring unit 1331 causes the statistical information collection processing unit 1332 to execute statistical information collection processing (S1204). The contents of the statistical information collection process will be described later.

  After the statistical information collection process is completed, the state transition monitoring unit 1331 determines whether the statistical information collection process has been executed a specified number of times (S1205). What is necessary is just to hold | maintain the frequency | count of execution about a statistical information collection process as a parameter in the state transition monitoring part 1331, for example. When it is determined that the statistical information collection process has been executed the specified number of times (S1205, Yes), the state transition monitoring unit 1331 uses a predetermined procedure for information regarding state transitions that have been collected and stored in the temporary storage table 900. Statistical processing is performed, and the obtained statistical information is stored in the statistical information management table 800 (S1207). As described above, the contents of the statistical processing are obtained by, for example, simply averaging the transition time and the state waiting time for the state transition obtained by measuring a plurality of times specified for a certain process, or by calculating the center value, the mode Appropriate statistical processing such as obtaining a value can be applied.

  The state transition monitoring unit 1331 then changes the statistical information collection completion flag 702 to “True” in the process management table 700 for the process being processed, and specifies the statistical information management table 800 corresponding to the statistical information table 703 ( S1208).

  When it is determined in S1205 that the statistical information collection process has not been executed the specified number of times (S1205, No), the state transition monitoring unit 1331 creates a new temporary storage table 900, and in the next statistical information collection process, The collected information is recorded in the newly created temporary storage table 900 (for example, “temporary storage 2”) (S1206).

  The state transition monitoring process described above is repeatedly executed unless the program for operating the computer 1 or the failure detection unit 133 is terminated after the system is started.

  According to the state transition monitoring process described above, when the failure detection unit 133 is required to execute failure detection of the process based on the execution state (state transition) of the process executed by the CPU 110, the normal time It is possible to automatically collect statistical information related to the process state transitions and monitor state transitions based on the statistical information.

Statistical Information Collection Processing Next, statistical information collection processing (S1204 in the state transition monitoring processing (FIG. 12)) executed by the statistical information collection processing unit 1332 of the failure detection unit 133 will be described. FIG. 13 shows an example of a processing flow of statistical information collection processing in the present embodiment. In the statistical information collection process, a process for collecting and storing statistical information in the statistical information storage unit 1333 is mainly executed by the statistical information collection processing unit 1332.

  First, the statistical information collection processing unit 1332 starts measuring the usage time of the CPU 110 by the process to be monitored (S1301). Next, the statistical information collection processing unit 1332 determines whether the monitoring target process has started using the CPU 110 (S1302). If it is determined that the monitoring target process is using (S1302, Yes), the CPU usage time started in S1301 is determined. Continue measuring.

  If it is determined in S1302 that the process to be monitored is not used (S1302, No), the statistical information collection processing unit 1332 suspends measurement of the usage time of the CPU 110 in order to measure only the time during which the CPU 110 is used. For example, the measured value of the usage time is stored in an appropriate storage location in the main storage device 130 (S1303). The state where the process does not use the CPU 110 includes a case where the process itself releases the CPU 110, a case where the process scheduler 1311 stops using the process, and the like.

  Next, the statistical information collection processing unit 1332 measures the time until the monitored process starts using the CPU 110 again, that is, the time when the CPU 110 is not used by the monitored process (S1304).

  Next, the statistical information collection processing unit 1332 determines whether the time during which the CPU 110 measured in S1304 is not used exceeds a value set as a threshold value recognized as a state (S1305). This threshold is a parameter for determining whether or not the time measured as not using the CPU 110 may be grasped as a “state”, and an arbitrary numerical value can be set.

  If it is determined that the time during which the measured CPU 110 is not used does not exceed the threshold (No in S1305), the statistical information collection processing unit 1332 does not grasp the measurement value as a state and grasps that it is still a transition. Therefore, the measurement of the usage time of the CPU 110 is resumed (S1311).

  When it is determined that the time when the measured CPU 110 is not used exceeds the threshold (S1305, Yes), the statistical information collection processing unit 1332 refers to the process management table 700 and measures the time when the CPU 110 is not used. The value is registered as the state waiting time 903 corresponding to “state transition 1” recorded in the monitoring item 901 of the temporary storage table 900 currently used for the monitoring target process.

  Next, the statistical information collection processing unit 1332 stores the measured usage time value of the CPU 110 stored in S1303 in the “state transition 1” recorded in the monitoring item 901 of the temporary storage table 900 that is also currently used. The corresponding transition time 902 is registered (S1307).

  As described above, since the measurement and recording of the transition time 902 and the state standby time 903 related to the first state transition related to the monitoring target process are completed, the statistical information collection processing unit 1332 stores the next state transition in the temporary storage table 900. In order to record the transition time 902 and the state waiting time 903 for (the state transition 2 in FIG. 9), a record relating to the next state transition 2 is added as the monitoring item 901.

  Next, the statistical information collection processing unit 1332 determines whether or not the process to be monitored has ended (S1309), and when determining that it has not ended (S1309, No), measures the transition time 902 in the next state transition. Therefore, measurement of the usage time of the CPU 110 is started again (S1310), the process is returned to S1302, and the state transition is recorded until the process to be monitored is completed. If it is determined in S1309 that the process to be monitored has ended (S1309, Yes), the statistical information collection processing unit 1332 ends the process.

  According to the statistical information collection processing described above, each of the monitoring target processes is calculated from the start to the end (specifically, from the start to the end in FIG. 1) to calculate statistical information used for the failure detection processing. The transition time of the state transition and the state waiting time can be obtained.

Statistical Information Comparison Processing Next, statistical information comparison processing will be described. 14A and 14B show an example of a processing flow of statistical information comparison processing in the present embodiment. In this statistical information comparison processing, failure detection is performed by comparing the information obtained from the state transition monitoring unit 1331 with the information stored in the statistical information storage unit 1334 by the statistical information comparison processing unit 1334 of the failure detection unit 133. Processing is executed. The time related to the state transition of the monitoring target process is measured by monitoring the state of the process scheduler 1311 of the OS 131 as in the case of the statistical information collection process illustrated in FIG.

  First, the statistical information comparison processing unit 1334 starts measuring the usage time of the CPU 110 by the monitoring target process (S1401). Next, the statistical information comparison processing unit 1334 determines whether the process is using the CPU 110 (S1402). If it is determined that the process is using (S1402, Yes), the status transition recorded for the monitored process is displayed. Transition time 802 (for example, when process A in process management table 700 is the monitoring target, transition time 802 recorded for “state transition 1” in statistical information management table 800 in FIG. 8) and measured usage time of CPU 110 Comparison with (processor usage time measurement value) is executed (S1403).

  The statistical information comparison processing unit 1334 determines whether the measured value of the CPU usage time exceeds the threshold value defined in advance for the transition time 802 in the comparison processing in S1403 (S1404). The threshold related to the transition time 802 can be specified as a parameter by the administrator or the like through the input device 150 and can be stored in the statistical information comparison processing unit 1334. As a specific example of the threshold value to be specified, it is conceivable to specify a value that is twice or three times the transition time 802 of the state transition.

  When it is determined that the CPU usage time measurement value does not exceed the specified threshold (S1404, No), the statistical information comparison processing unit 1334 determines that the transition time of the process to be monitored is normal, and the CPU 110 uses During this time, the processing of S1402 to S1404 is repeated. On the other hand, if it is determined that the CPU usage time measurement value exceeds the specified threshold value (S1404, Yes), the statistical information comparison processing unit 1334 has a fault with an abnormally long transition time in the monitored process. The failure notification unit 1335 notifies the occurrence of a failure through the OS 131 and the output device 160. The failure having an abnormally long transition time is a failure state in which the monitored process occupies the CPU 110 for a longer time than usual, such as an infinite loop.

  If the statistical information comparison processing unit 1334 determines in S1402 that the monitoring target process is not using the CPU 110 (No in S1402), the statistical information comparison processing unit 1334 interrupts the usage time measurement of the CPU 110 (S1406). Then, measurement of the time during which the monitored process is not using the CPU 110 (processor non-use time) is started (S1407).

Next, the statistical information comparison processing unit 1334 determines whether the monitoring target process uses the CPU 110 again (S1408). The determination process in S1408 is executed to determine whether a failure has occurred that causes the CPU 110 to stop abnormally long. When it is determined that the CPU 110 is not used (S1408, Yes), the statistical information comparison processing unit 1334 further determines whether the stop time of the CPU 110 exceeds an allowable threshold (S1409). A threshold that is allowed as the stop time (hereinafter referred to as “stop allowable threshold”) can be held in the statistical information comparison processing unit 1334 as a parameter when the administrator or the like inputs it through the input device 150. An arbitrary numerical value such as a value twice or three times the numerical value stored in the status waiting time 803 of the statistical information management table 800 can be designated as the stop allowable threshold value.

  In S1409, when it is determined that the measured value of the CPU non-use time (processor non-use time measurement value) exceeds the stop allowable threshold (S1409, Yes), the statistical information comparison processing unit 1334 has an abnormally long standby time. It is determined that a failure has occurred, and the failure notification unit 109 notifies the failure occurrence through the OS 131 and the output device 160 (S1410).

  In S1409, when it is determined that the measured value of the CPU non-use time does not exceed the stop allowable threshold (No in S1409), the statistical information comparison processing unit 1334 returns the process to S1408 and determines whether the CPU 110 is used again. To do. On the other hand, if it is determined in S1409 that the CPU non-use time has not exceeded the stop allowable threshold, and further in S1408 it is determined that the CPU 110 is being used again (No in S1408), the statistical information comparison processing unit 1334 In order to confirm whether the non-use time can be determined as the state, the processing is shifted to S1411 in FIG. 14B.

  In S1411, the statistical information comparison processing unit 1334 determines whether the CPU non-use time exceeds a predetermined threshold and can be recognized as a state. When it is determined that the threshold value is not exceeded (S1411, No), the statistical information comparison processing unit 1334 restarts the measurement of the usage time of the CPU 110 (S1417).

  On the other hand, if it is determined that the threshold value is exceeded (S1411, Yes), the statistical information comparison processing unit 1334 compares the CPU usage time with the transition time 802 recorded in the statistical information management table 800 of the monitored process. (S1412), it is determined whether the comparison result between the two is below a predetermined threshold (S1413). This threshold value can be set in the statistical information comparison processing unit 1334 as a parameter, for example. As an example of the threshold value, a value that is ½ times or １ ／ of the value measured at the transition time 802 can be specified. Thus, it is determined whether the time required for the state transition is unduly short. If it is determined that the threshold value is below the threshold (S1413, Yes), the statistical information comparison processing unit 1334 determines that the transition time 802 is unduly short and causes a transition different from normal, and the failure notification unit 108, The occurrence of a failure is notified through the OS 131 and the output device 160 (S1414). As an example of a transition that is different from normal, it may be assumed that a process that is completed in a short time that is not normally recognized as a state, such as waiting for a lock, is delayed for some reason and is newly recognized as a state. .

  On the other hand, when it is determined in S1413 that the threshold value is not below (S1413, No), the statistical information comparison processing unit 1334 records the transition time 802 and the state standby time 803 for the next state transition of the monitored process. Therefore, the monitoring item 801 is shifted to the next state transition (S1415). Then, it is determined whether the process to be monitored has ended (S1416). If it is determined that the process has ended (S1416, Yes), the statistical information comparison process for the process to be monitored is ended.

  If it is determined that the process has not been completed (S1416, No), the statistical information comparison processing unit 1334 measures the usage time of the CPU 110 again (S1418), shifts the process to S1402, and fails until the monitored process ends. Continue the statistical information comparison process to check whether or not the error occurred.

  As described above, according to the computer 1 according to the present embodiment, for each process executed by the CPU 110, the time measurement value regarding the state transition is sequentially compared with the reference value obtained statistically. Since the failure that occurred in the computer 1 can be reliably detected at an early stage that does not lead to an event such as a hang-up or an error detection by the OS 131, the downtime of the computer 1 is reduced as much as possible and the availability is improved. There is an effect.

  In the present specification, the present invention has been described with reference to the accompanying drawings according to the embodiment, but the present invention is not limited to such an embodiment. The present invention can be implemented in various forms within the scope of the invention described in the claims, regardless of the above-described embodiment, and is equivalent to the invention described in the claims. Products are also included in the present invention.

1 Computer 110 CPU 120 Control Unit 130 Main Storage Device 131 OS 1311 Process Scheduler 132 Program 132A, 132B Process 133 Failure Detection Unit 1331 State Transition Monitoring Unit 1332 Statistical Information Collection Processing Unit 1333 Statistical Information Storage Processing Unit 1334 Statistical Information Comparison Processing Unit 1335 Failure Notification unit 140 Auxiliary storage device 150 Input device 160 Output device 170 Communication control unit 180 Internal bus 700 Process management table 701 Monitored process 702 Statistical information collection completion flag 703 Statistical information table 800 Statistical information management table 801 Monitoring item 802 Transition time 803 State Standby time 900 Temporary storage table 901 Monitoring item 902 Transition time 903 State standby time

Claims (12)

A computer comprising a processor and a memory, wherein the processor executes a plurality of processes constituting at least one software program stored in the memory to execute the software program,
For each process, the processor usage time, which is the time during which the processor is processing the process from when the process is started by the processor to the end of the process, and the processor stops processing the process. The processor non-use time, which is a certain time, is measured and acquired sequentially several times, and according to a predetermined statistical process, a processor use time reference value that is a statistical reference value of each processor use time and each processor non-use A processor processing reference value acquisition unit that calculates and stores a processor non-use time reference value that is a statistical reference value of time for each of the processes;
When one of the plurality of processes constituting the software program is processed by the processor, the processor use time and the processor non-use time are measured for the process, and the process is sequentially stored for the process. A criterion for determining that a failure has occurred during the process processing when it is determined that the comparison result does not satisfy a predetermined determination criterion. A value comparison processing unit ,
When the reference value comparison processing unit determines that the measured processor non-use time is longer than a predetermined first processor non-use time threshold and the processor use time at that time is shorter than a predetermined processor use time threshold, Determining that a failure has occurred in which the transition time is shorter than the predetermined processor usage time threshold in the process ;
A computer characterized by that.   The computer according to claim 1, further comprising a process allocating unit that performs a process of allocating the processor to each of the processes, wherein the processor processing reference value acquisition unit and the reference value comparison processing unit include: A computer characterized by measuring the processor use time and the processor non-use time by monitoring an operation. 2. The computer according to claim 1, wherein the reference value comparison processing unit is configured to measure the processor usage time measurement value and the corresponding processor usage time reference value during processing of any of the plurality of processes by the processor. And when it is determined that the difference between the measured processor usage time value and the corresponding processor usage time reference value exceeds a predetermined value, it is determined that a failure has occurred during the process processing. A computer characterized by The computer according to claim 1, wherein the reference value comparison processing unit is configured to use the processor non-use time measurement value and a predetermined second processor non-use during processing of any of the plurality of processes by the processor. A comparison with a time threshold value, and when it is determined that the processor non-use time measurement value exceeds the predetermined second processor non-use time threshold value, it is determined that a failure has occurred during the process processing. A featured computer. A computer fault detection method comprising a processor and a memory, wherein the processor processes a plurality of processes constituting at least one software program stored in the memory and executes the software program. The processor is
For each process, the processor usage time, which is the time during which the processor is processing the process from when the process is started by the processor to the end of the process, and the processor stops processing the process. The processor non-use time, which is a certain time, is measured and acquired sequentially several times, and according to a predetermined statistical process, a processor use time reference value that is a statistical reference value of each processor use time and each processor non-use A processor non-use time reference value, which is a statistical reference value of time, is calculated and stored for each of the processes;
When one of the plurality of processes constituting the software program is processed by the processor, the processor use time and the processor non-use time are measured for the process, and the process is sequentially stored for the process. Is compared with the processor usage time reference value and the processor usage time reference value, and when it is determined that the comparison result does not satisfy a predetermined determination criterion, it is determined that a failure has occurred during the process processing ,
When it is determined that the measured processor non-use time is longer than a predetermined first processor non-use time threshold and the processor use time at that time is shorter than a predetermined processor use time threshold, the transition time in the process is the predetermined A failure detection method for a computer, characterized in that it is determined that a failure has occurred that is shorter than a processor usage time threshold . 6. The failure detection method for a computer according to claim 5 , wherein the processor measures the processor use time and the processor non-use time by monitoring an assignment status of the processor to each of the processes. A computer fault detection method characterized by the above. 6. The failure detection method for a computer according to claim 5 , wherein the processor uses the processor usage time measurement value and the corresponding processor usage time reference value during processing of any of the plurality of processes by the processor. And when it is determined that the difference between the measured processor usage time value and the corresponding processor usage time reference value exceeds a predetermined value, it is determined that a failure has occurred during the process processing. A computer fault detection method characterized by the above. 6. The failure detection method for a computer according to claim 5 , wherein the processor uses the processor non-use time measurement value and a predetermined second processor non-use during processing of any of the plurality of processes by the processor. A comparison with a time threshold value, and when it is determined that the processor non-use time measurement value exceeds the predetermined second processor non-use time threshold value, it is determined that a failure has occurred during the process processing. A feature of a computer fault detection method. A computer comprising a processor and a memory, wherein the processor executes a plurality of processes constituting at least one software program stored in the memory and executes the software program.
For each process, the processor usage time, which is the time during which the processor is processing the process from when the process is started by the processor to the end of the process, and the processor stops processing the process. The processor non-use time, which is a certain time, is measured and acquired sequentially several times, and according to a predetermined statistical process, a processor use time reference value that is a statistical reference value of each processor use time and each processor non-use Calculating and storing a processor non-use time reference value, which is a statistical reference value of time, for each of the processes;
When one of the plurality of processes constituting the software program is processed by the processor, the processor use time and the processor non-use time are measured for the process, and the process is sequentially stored for the process. Comparing with the processor usage time reference value and the processor usage time reference value, and determining that a failure has occurred during the process processing when it is determined that the comparison result does not satisfy a predetermined determination criterion; ,
When it is determined that the measured processor non-use time is longer than a predetermined first processor non-use time threshold and the processor use time at that time is shorter than a predetermined processor use time threshold, the transition time in the process is the predetermined Determining that a failure has occurred that is shorter than the processor usage time threshold of
A program characterized by that. The program according to claim 9 , wherein the processor is caused to execute a step of measuring the processor use time and the processor non-use time by monitoring a state of assignment of the processor to each process. A program characterized by The program according to claim 9 , wherein the processor usage time measurement value is compared with the corresponding processor usage time reference value during processing of any of the plurality of processes by the processor. If it is determined that the difference between the processor usage time measurement value and the corresponding processor usage time reference value exceeds a predetermined value, a step of determining that a failure has occurred during the process processing is executed. A program characterized by 10. The program according to claim 9 , wherein the processor non-use time measurement value and a predetermined second processor non-use time threshold value are set in the processor during processing of any of the plurality of processes by the processor. In comparison, when it is determined that the processor non-use time measurement value exceeds the predetermined second processor non-use time threshold value, a step of determining that a failure has occurred during the process processing is executed. A featured program.

Images