WO2012111167A1

WO2012111167A1 - Trace information acquisition method, computer system, and program

Info

Publication number: WO2012111167A1
Application number: PCT/JP2011/053799
Authority: WO
Inventors: 健一井手上; 元樹小幡; 博泰西山
Original assignee: 株式会社日立製作所
Priority date: 2011-02-14
Filing date: 2011-02-22
Publication date: 2012-08-23

Abstract

In one embodiment of the present invention, a probe in an expansion program acquires trace information of a type that has been set for a procedure in each of a plurality of procedures included in an execution path. A level-of-importance calculation processor determines the level of importance of each of the plurality of procedures by using the trace information, and uses the level of importance of a procedure included in the plurality of procedures to determine a new level of importance for establishing the type of trace information to be acquired in another procedure. The level-of-importance calculation processor updates the type of trace information to be acquired in the other procedure in accordance with the new level of importance for establishing the type of trace information thus determined.

Description

Trace information acquisition method, computer system, and program

Import by reference

This application claims the priority of Japanese Patent Application No. 2011-28941, which was filed on February 14, 2011, and is incorporated herein by reference.

The present invention relates to a trace information acquisition method, a computer system, and a program, and more particularly to acquisition of trace information in program execution.

Business systems used in business operations of companies and other organizations are required to have high reliability including programs to be executed. Therefore, when a failure occurs in the execution of the program in the business system, the system maintenance manager is required to quickly recover the system from the failure.

In system recovery work, it takes a lot of time to identify the location of the failure, which may prolong the period until failure recovery. Therefore, there is a need for an analysis method that efficiently finds a fault location in program execution. At the same time, the system administrator needs to efficiently grasp the structure of the program and the behavior at the time of execution in order to efficiently find the fault location, devise the contents of correction, and actually correct it.

There is a tracer (trace program) as a support tool for grasping the program structure and execution behavior. The tracer outputs a trace information file including method call information in the execution program. The system administrator can grasp the failure occurrence process in the execution of the program by analyzing the output trace information.

The technology of the tracer is disclosed in, for example, Japanese Patent Laid-Open No. 2000-259449 (Patent Document 1). The technique described in Patent Literature 1 calculates the acquisition level of trace information from the sum of the usage rates of computer resources such as the CPU usage rate and the network usage rate. Further, by comparing the trace level constant set for the trace execution point with the calculated acquisition level, it is determined whether or not to output the trace information at the trace execution point to the trace file.

In addition, Japanese Patent Laid-Open No. 2008-257287 (Patent Document 2) selects a main method according to the processing time of a method, and executes a program for acquiring an execution path and the processing time of each method when the main method is executed. A trace acquisition control technique is disclosed.

JP 2000-259449 A JP 2008-257287 A

However, failures in business systems often occur when the execution path and program values are rare, and the failure generation process is often complex. For this reason, it takes a long time to investigate the cause. Therefore, the tracer acquires trace information that helps understand the structure and operation of the program to be traced, presents the trace information, presents the location of the failure and the failure occurrence process, and the system administrator indicates the location of the failure. In addition, it is required to be able to always obtain the failure occurrence process.

Furthermore, the tracer is mainly intended to analyze unexpected failures in the actual operating environment. As a requirement for realizing the purpose, the tracer is required to always execute in an actual operation environment and continuously acquire trace information including argument information.

In this respect, there is a problem that if the trace information is continuously acquired, the size of the log file becomes enormous and the I / O time for outputting the log also increases. The technique disclosed in Patent Document 1 changes the trace execution point by changing the acquisition level according to the resource usage status. Thereby, the trace information acquired according to a resource utilization condition can be reduced. The technique disclosed in Patent Document 2 selects a main method according to a processing time, and selects the main method as a target for performing detailed measurement. Thereby, the trace information to be acquired can be reduced.

However, the business system does not execute each procedure independently in executing the program, but executes each procedure in an execution path determined by the calling relationship between procedures. That is, the possibility of occurrence of a failure in program execution varies depending on the execution path, and even in the same procedure, a failure can occur with different probabilities in different execution paths. Therefore, it is important for the trace information acquired in each procedure to consider not only the nature of the procedure but also the nature of the execution path on which the procedure is executed.

The above conventional technology makes a determination on the trace information to be acquired individually for each trace execution point or method. Therefore, although trace information can be reduced, there is a possibility that important information in failure analysis cannot be obtained. Therefore, a technique capable of more appropriately acquiring trace information important for failure analysis in program execution is desired.

One aspect of the present invention is a method in which a computer system including a computer that executes a program acquires trace information in the execution of the program. In this method, in each of a plurality of procedures included in the execution path in the execution of the program, the type of trace information set for the procedure is acquired. The acquired trace information is stored in a storage area. The importance of each of the plurality of procedures is determined using trace information stored in the storage area. Using the importance level of the procedures included in the plurality of procedures, a new importance level that determines the type of trace information acquired in the other procedures included in the plurality of procedures is determined. The type of trace information acquired in the other procedure is updated according to a new importance level that determines the type of the determined trace information. In the execution of the other procedure, the updated type of trace information is acquired.

According to one aspect of the present invention, it is possible to more appropriately acquire trace information in program execution.

FIG. 2 is a diagram schematically illustrating a configuration of a computer having a trace function in the first embodiment. 6 is a flowchart illustrating an overall processing flow of the tracer in the first embodiment. 5 is a flowchart illustrating a flow of importance calculation processing in the first embodiment. In a 1st embodiment, it is a block diagram showing an example of composition of profile information. It is a figure which shows the example of the correspondence table of trace information acquisition level and importance reference value in 1st Embodiment. It is a figure which shows an example of the call tree constructed | assembled by the call relationship construction process in 1st Embodiment. 5 is a flowchart illustrating an execution path analysis process in the first embodiment. It is a figure which shows the example of the information provided to the method of a call tree in 1st Embodiment. 5 is a flowchart illustrating importance path reference value calculation processing for an execution path in the first embodiment. 5 is a flowchart illustrating a trace information acquisition level determination process in the first embodiment. 5 is a flowchart illustrating a program extension process in the first embodiment. 5 is a flowchart showing probe insertion processing in the first embodiment. 5 is a flowchart illustrating a trace information output process in the first embodiment. FIG. 3 is a block diagram illustrating an example of output trace information in the first embodiment. 6 is a diagram illustrating a display example of profile information in the first embodiment. FIG. 6 is a diagram illustrating a display example of trace information in the first embodiment. FIG. In the first embodiment, it is a diagram showing an example of trace execution path designation information. In a 1st embodiment, it is a figure showing an example of tracing object specification information. FIG. 3 is a diagram illustrating an example of a program to be traced in the first embodiment. It is a figure which shows the example of the argument and internal variable information of the method in profile information used in determination of the importance of the execution path of a program example in 1st Embodiment. It is a figure which shows the example of the call path information of the method in profile information used in determination of the importance of the execution path of a program example in 1st Embodiment. It is a figure which shows the example of the execution frequency information of the method in profile information used in determination of the importance of the execution path of a program example in 1st Embodiment. It is a figure which shows the example of the execution time information of the method in profile information used in determination of the importance of the execution path of a program example in 1st Embodiment. It is a figure which shows the example of the branch instruction number information of the method in profile information used in determination of the importance of the execution path of a program example in 1st Embodiment. It is a figure which shows the example of the calling relationship information of the method in profile information used in determination of the importance of the execution path of a program example in 1st Embodiment. It is a figure which shows the example of the line number information of the method in profile information used in determination of the importance of the execution path of a program example in 1st Embodiment. It is a figure which shows the example of the total instruction number information of the method in profile information used in determination of the importance of the execution path of a program example in 1st Embodiment. It is a figure which shows the example of the trace acquisition level information of the method in profile information used in determination of the importance of the execution path of a program example in 1st Embodiment. FIG. 20 is a diagram illustrating an example of a call tree generated from the method call relation information of FIG. 19F in the first embodiment. In the first embodiment, an example of a correspondence table of trace information acquisition levels / importance level reference values corresponding to the profile information of FIGS. 19A to 19I is shown. In the first embodiment, an example of a result of inserting a probe into a class of an execution program is shown. In the first embodiment, an example of a result of inserting a probe into a class of an execution program is shown. In the first embodiment, an example of a result of inserting a probe into a class of an execution program is shown. In the first embodiment, an example of a result of inserting a probe into a class of an execution program is shown. 5 is a flowchart illustrating a program execution process in the first embodiment. In 1st Embodiment, the example of the trace information acquired from the example program is shown. In 1st Embodiment, the example of the trace information acquired from the example program is shown. In the first embodiment, an example of output trace information of a program example is shown. In the first embodiment, an example of output trace information of a program example is shown. In the first embodiment, an example of output trace information of a program example is shown. In 1st Embodiment, the example of the trace information acquired in execution of the example program is shown. In 1st Embodiment, the example of the trace information acquired in execution of the example program is shown. In 2nd Embodiment, it is a block diagram which shows typically the structure of the computer which has a trace function containing a program monitoring process part. In 2nd Embodiment, it is a flowchart which shows the flow of the whole process of a tracer. 9 is a flowchart illustrating a program monitoring process in the second embodiment. It is an example of the display image of profile information in 2nd Embodiment. It is an example of the display image of profile information in 2nd Embodiment. It is an example of the display image of profile information in 2nd Embodiment. In 2nd Embodiment, it is an example of the display image of trace information. In 2nd Embodiment, it is an example of the display image of trace information. In 2nd Embodiment, it is an example of the display image of trace information.

Hereinafter, modes for carrying out the present invention will be described. For clarity of explanation, the following description and drawings are omitted and simplified as appropriate. Moreover, in each drawing, the same code | symbol is attached | subjected to the same element and the duplication description is abbreviate | omitted as needed for clarification of description.

This embodiment describes a preferred method for efficiently collecting trace information effective for solving a problem in program execution. In the following, an example in which the present invention is applied to a Java (registered trademark) system will be described as a preferred example. The present invention is suitable for an object-oriented language, particularly a system for executing a program by Java, but is also applied to an object-oriented language different from Java or a system for executing a program written in a language different from an object-oriented language. be able to.

[First Embodiment]
In this embodiment, a probe for acquiring trace information is inserted into a Java program to be executed. In this embodiment, while acquiring trace information with a probe, information useful for solving a problem in program execution is acquired with low overhead by determining the acquisition information type of a procedure in each execution path.

The system of this embodiment identifies an execution path and determines the type of trace information to be acquired in a method in the execution path according to the importance of the execution path. In Java, a method is a procedure and a program to be executed. The execution path of the Java program includes one or more methods. The importance is an index representing the likelihood of failure.

The system according to the present embodiment increases the importance of an execution path or a method in which a failure is likely to occur, and acquires detailed trace information there. Similarly, the importance of an execution path or method in which a failure is unlikely to occur is reduced, and the trace information acquired there is simplified. As a result, trace information useful for failure analysis is acquired, and the amount of acquired trace information is reduced.

FIG. 1 is a block diagram schematically showing the configuration of an apparatus that executes the trace information acquisition method of this embodiment. The computer 101 includes a CPU 102 that is a processor and a memory 103 that is a storage device. The processor may include a plurality of one or more processor cores and may include a plurality of processor packages. The CPU 102 executes a Java virtual machine (Java VM) 114.

The Java virtual machine 114 includes a program expansion processing unit 104, an importance calculation processing unit 105, a trace information output processing unit 107, a screen display processing unit 115, a program reading processing unit 116, a profile information update processing unit 118, and a program execution processing unit 120. including. These are programs, and the CPU 102 can function as each processing unit by operating according to each program.

FIG. 1 shows each processing unit in the CPU 102 for convenience of explanation, but in an actual operation, each program is typically loaded into the memory 103 from the external storage device 131 or another nonvolatile storage device. Then, the CPU 102 acquires commands and data of each program from the memory 103 and operates according to the commands. The processing result of the CPU 102 is reflected in the data stored in the memory 103. A user can input information to the computer 101 using an input device 192 (for example, a mouse or a keyboard). The display device 191 (for example, a liquid crystal display device) displays a user-set image, an acquired trace information image, and other images.

In FIG. 1, the memory 103 includes a program 109, an extended program 110, acquisition trace information 124, call tree information 125, profile information 111 of the program 109, output trace information 112, trace information acquisition level / importance level reference value correspondence table 121, Trace target designation information 123 and trace execution path designation information 122 are stored. Each information is stored in a storage area corresponding to each memory 103.

The program 109 is a trace target, and the extended program 110 is a program in which a probe is inserted. The acquired trace information 124 is information acquired by the probe during the execution of the extension program 110, and is internal data of the tracer.

The call tree information 125 is information indicating a read relationship between methods of the extension program 110. This is internal data of the importance calculation processing unit 105 generated from the acquired trace information 124 by the importance calculation processing unit 105. The profile information 111 is profile information of the program 109 and is generated from the acquired trace information 124 and the analysis result of the program 109.

The output trace information 112 is output trace information converted from the acquired trace information 124. The trace information acquisition level / importance level reference value correspondence table 121 associates the trace information acquisition level of the method with the importance level reference value of the execution path. The execution path importance level reference value and the trace information acquisition level indicate the importance level of the execution path. The calculation method of the importance standard value will be described later.

The trace information acquisition level specifies the type of trace information acquired by the corresponding method. Therefore, the trace acquisition level and the importance reference value of the execution path associated therewith represent the importance of the method in acquiring trace information.

The trace target designation information 123 indicates a trace target designated (limited) by the user. The trace execution path designation information 122 indicates the designated trace acquisition level of the execution path designated by the user. Details of the information stored in the memory 103 in FIG. 1 will be described later.

These may be stored not in the memory 103 but in the external storage device 131. A storage device that stores data (including programs) necessary for the operation of the system can include a plurality of different types of components (eg, storage devices of different computers). In the present embodiment, the information stored in the memory 103 or the external storage device 131 does not depend on the data structure, and may be expressed in any data structure such as a table or a list. This is the same in the second embodiment.

Typically, the program 109, the profile information 111, the output trace information 112, the trace information acquisition level / importance reference value correspondence table 121, the trace target designation information 123, and the trace execution path designation information 122 are the external storage device 131 or others. And is loaded into the memory 103. Data (information) updated in the memory 103 is reflected in data (information) in the nonvolatile memory.

In the computer system of this embodiment, at least a part of the program that operates on the computer 101 in FIG. 1 can operate on another computer. For example, each processing unit of the Java virtual machine 114 can be executed by a computer different from the computer 101 that executes the extension program 110. Further, data (information) different from the expansion program 110 shown in the memory 103 may be stored in a computer different from the expansion program 110.

FIG. 2 shows the flow of processing in the overall configuration diagram of FIG. First, the program read processing unit 116 reads the program 109 stored on the memory 103 (S101). The program reading processing unit 116 performs processing for reading the program 109 to be traced into the Java virtual machine 114 from the memory 103 or the external storage device 131.

Next, the importance level calculation processing unit 105 refers to the profile information 111 to identify the execution path in the program 108 and calculate its importance level reference value. From the importance level reference value, a method included in the execution path is calculated. The trace information acquisition level (the type of information acquired by the method of the program 109) is determined (S102).

Next, the program extension processing unit 104 acquires trace information (data based on the call tree information 125, the profile information 111, and the output trace information 112) according to the trace information acquisition level determined by the importance calculation processing unit 105. A probe (program piece) is inserted into the program 109 (S103). The probe for acquiring the trace information is composed of one or a plurality of probes. Thereafter, the program extension processing unit 104 stores the program 109 in which the probe is inserted in the memory 103 as the extension program 110.

Next, the program execution processing unit 120 executes the extended program 110 in which a probe for acquiring trace information is inserted (S104). The probe stores the trace information acquired in the program execution in the memory 103. This information is shown as acquired trace information 124 in FIG. 1 and is internal data of the probe.

Next, the profile information update processing unit 118 determines whether the update conditions for the profile information 111 specified by the user in advance are satisfied (S105). An example of the update condition is the passage of a specified time. When the profile information 111 is updated (S105: YES), the process proceeds to a profile information update step (S106). When the profile information 111 is not updated (S105: NO), the process proceeds to step S108.

The profile information update processing unit 118 can determine the update condition every time the trace information acquisition probe in the extended program 110 is executed (S105), or perform it when designated in advance by the user. Also good.

In step S106, the profile information update processing unit 118 totals the trace information acquired by the probe and updates the profile information 111 on the memory 103. Further, the profile information update processing unit 118 updates the trace information acquisition level / importance level reference value correspondence table 121 (S107).

The process returns to the importance reference value calculation step (S102) after updating the profile information 111 and the trace information acquisition level / importance reference value correspondence table 121. Through the loop from step S102 to step S107, the profile information 111, the trace information acquisition level / importance level reference value correspondence table 121, the execution path (call tree), and the trace information acquisition level are updated during program execution.

In step S108, the trace information output processing unit 107 determines whether a failure has occurred during execution of the extended program 110 or whether the user has issued an output instruction. When a failure has occurred or an output command has been issued (S108: YES), the trace information output processing unit 107 performs a trace information output process (S109). The trace information output processing unit 107 converts the trace information 124 acquired in the program execution processing 106 into output trace information 112 and outputs (stores) it on the memory 103 (S109).

If the determination result in step S108 is NO, the Java program 110 continues to execute. Instead of the trace information output processing unit 107, the Java virtual machine command reception processing unit may determine whether the user has issued a trace data output command.

Next, in step S110, the screen display processing unit 115 determines whether to output the profile information 111 and the output trace information 112 on the screen. If the screen is not output (S110: NO), the process ends. In the case of screen output (S110: YES), in step S111, the screen display processing unit 115 displays the profile information 111 by the information related to the method and the calling relationship, and executes the output trace information 112 up to the location where the failure occurred. Display in the route.

The profile information 111 and the output trace information 112 are displayed on the display device 191 by an appropriate method according to the design. The profile information 111, the output trace information 112, and other display images in the present embodiment and other embodiments can be displayed on the display device 191 or other display devices. Alternatively, these may be printed by a printer.

(Importance calculation processing (S102))
Details of the steps (processing) in the processing described with reference to FIG. 2 will be described below. First, the importance calculation processing (S102) shown in the flowchart of FIG. 2 will be described with reference to FIG. The importance calculation processing unit 105 determines the importance of the execution path of the trace target program. The final numerical value indicating the importance of the execution path in this process is the trace information acquisition level.

The importance level calculation processing unit 105 calculates the importance level reference value according to the importance level reference, and from the importance level reference value and the trace information acquisition level / importance level reference value correspondence table 121, the execution path (the end point method) of the execution path. Determine the trace information acquisition level (type of trace information to be acquired).

Specifically, as shown in FIG. 3, the importance calculation processing unit 105 first determines whether or not the profile information 111 exists on the memory 103 (S301). If the profile information 111 does not exist (S301: NO), the importance calculation processing unit 105 proceeds to step S306. In step S306, the trace information acquisition levels of all methods are set to “standard” which is an initial value. The set information is stored in the memory 103. The trace information acquisition level will be described later with reference to FIG.

On the other hand, when the profile information 111 exists (S301: YES), the importance calculation processing unit 105 proceeds to the call relationship construction processing (S302). In the call relationship construction process (S302), the importance calculation processing unit 105 reads the profile information 111 and constructs a method call relationship (call tree) for the trace target program (see the example in FIG. 6). Here, the call tree is information indicating which method is called from which method and which method is called. The memory 103 stores this call tree information 125.

FIG. 4 shows a configuration example of the profile information 111. Profile information 111 includes method argument / internal variable information 1301, method call path information 1302, method execution count 1303, method execution time information 1304, method row count information 1305, method branch instruction count information 1306, method Call relation information 1307, method total instruction count information 1308, and method trace information acquisition level 1309 (information calculated in step S105). The profile information 111 is created and updated in the profile information update process (S106) based on the analysis result of the trace target program and the acquired trace information 124 which is data inside the tracer.

Next, the importance calculation processing unit 105 performs an execution path analysis process (S303). In this process, the importance calculation processing unit 105 adds the characteristic information and importance information of each method to the call tree created in the call relationship construction process (S302), and further extracts the execution path in the call tree.

Next, the importance calculation processing unit 105 calculates the importance reference value of each execution route extracted in the execution route analysis processing (S303) in the importance reference value calculation processing (S304) of the execution route. The importance reference value of the execution path is a value representing the importance of the execution path. The calculation method of the importance standard value will be described later.

Next, the importance calculation processing unit 105 acquires the trace information on the memory 103 from the importance reference value of the execution path calculated in the importance reference value calculation process (S304) in the trace information acquisition level determination process (S305). With reference to the level / importance reference value correspondence table 121, the trace information acquisition level of the method included in the execution path (in the preferred example described below, the end method of the execution path) is determined.

FIG. 5 shows an example of the trace information acquisition level / importance level reference value correspondence table 121. This table 121 associates importance level reference values with trace information acquisition levels. The trace information acquisition level is assigned to each method of the trace target program 109. The trace information acquisition level specifies the type of information acquired by the method. The trace information acquisition level represents the importance of the execution path and the importance of the method in the execution path.

This example has three levels of “detail”, “standard”, and “simple” as illustrated in FIG. As described above, the initial value of the trace information acquisition level is “standard”. At the “detail” level, the probe acquires the method call relationship, the method argument / return value, and the value of the internal variable. At the “standard” level, the probe acquires method call relationships and method arguments / return values, and omits acquiring internal variable values. At the “simple” level, the probe acquires only the method call relationship.

Generate method argument / internal variable information 1301, method call path information 1302, method execution count 1303, method execution time information 1304, and method call relation information 1307 in profile information 111 from “detail” level information can do. For a method for acquiring “standard” or “simple” level trace information, part or all of the method argument / internal variable information 1301 is omitted.

The example shown in FIG. 5 uses statistical information (average, variance, etc.) of the importance standard value of the execution path as a criterion for classifying the trace information acquisition level. As a method for determining the trace information acquisition level, an appropriate method is set by design. For example, the trace information acquisition level may be determined by comparing the importance reference value of the execution path with a specified value without using statistical information.

Next, the flow of the call relationship construction process (S302) shown in the flowchart of FIG. 3 will be described. As described above, the call relationship construction process (S302) constructs a method call relationship in the trace target program 109 (generation of call tree information 125). FIG. 6 shows an example of a call tree.

In FIG. 6,

methods

4503, 4504, and 4506 to 4508 are indicated by reference numerals among methods (nodes) surrounded by a rectangle. An arrow between methods is an edge, and one edge 4505 is indicated by a reference symbol. Further, the call tree of this example includes the number of calls at each edge, method characteristic information, and method importance information, in addition to the call relationship between methods.

In the call tree shown in FIG. 6, information on each method includes characteristic information and importance information such as an execution time, complexity (for example, the number of branch instructions), and the number of executions, in addition to an identifier for specifying the method. The importance level information is information related to the importance level of the execution path (method). The number given to the edge indicates the number of calls from the caller method to the callee method. The number of executions of one method is the total number of calls from all the read source methods.

In the call relation construction process (S302), the importance calculation processing unit 105 associates the method with the node and the call relation with the edge from the method call relation information (call relation information 1307 in FIG. 4) in the profile information 111. Build a call tree. This call tree does not include the number of calls, method characteristic information, and importance information. Such information is added in the subsequent execution path analysis process (S303). These may be added by the call relationship construction process (S302).

Next, the flow of the execution path analysis process (S303 in FIG. 3) will be described with reference to the flowchart in FIG. In the execution path analysis process (S 303), the importance level calculation processing unit 105 extracts an execution path from the call tree constructed in the call relation construction process (S 302) and stores it in the memory 103.

As shown in the flowchart of FIG. 7, first, in step S501, the importance calculation processing unit 105 performs method (node) characteristic information and importance on the call tree constructed in the call relationship construction processing (S302). Information and edge call count information (call tree characteristic information) are added. The call tree (call tree information 125) to which new information (call tree characteristic information) is added has the configuration shown in FIG.

FIG. 8 shows an example of information given to the method (node) of the call tree, and illustrates the node 4503 (method A). In this example, each node has, as method characteristic information, the execution time of the corresponding method, the number of branch instructions as complexity, and the number of executions.

The importance calculation processing unit 105 can acquire these pieces of information from the profile information 111. Specifically, the method characteristic information is acquired from the method execution time information 1304, the method branch instruction number information 1306, and the method execution number information 1303.

The importance level information includes information on the importance level of the execution path (the importance level of the method associated with the trace information acquisition level). Specifically, an importance criterion, which is a criterion for determining the importance of an execution route, an importance criterion value of an execution route calculated from the execution criterion, and trace information in the method specified from the importance criterion value Includes acquisition level. The trace information acquisition level of the method is determined from this importance level reference value, and is also a value representing the importance level of the method. The importance standard value represents the importance of the execution path and the importance of the method that determines the trace information acquisition level. Details of the importance determination method will be described later.

In the execution path analysis process, the value of the importance standard value and the value of the trace information acquisition level are the previous value, the initial value, or not set. These new values (updated values) are determined in the importance calculation process (S304 in FIG. 3).

As can be understood from the loop of steps S101 to S107 in the flowchart of FIG. 2, the call relationship diagram (execution path), the execution path importance level reference value, and the method trace information acquisition level are updated during the execution of the program. (Changed dynamically) Accordingly, it is possible to set an appropriate type of trace information for each method in accordance with a change in a call tree (including characteristic information) during program execution. The updated level is the same as or different from the level before the update.

In the example of FIG. 8, the importance level is the number of executions. A smaller number of executions indicates a higher importance, and the importance decreases with an increase in the number of executions (increase in the importance level reference value). The importance criterion is composed of one or a plurality of elements selected from the characteristic information of the method. The characteristic element used as the importance criterion and the calculation method of the importance criterion value are preset in the importance calculation processing unit 105. In a preferred example, the user can change the importance information.

Specifically, the user designates the importance criterion and the importance calculation method using the input device 192 of the computer 101 or the input device of another computer connected via the network. The importance calculation processing unit 105 calculates the importance (reference value) by calculating the value of the characteristic element of the specified method by the specified method. Details of the importance level reference value calculation will be described later.

Next, the importance calculation processing unit 105 selects a method as a starting point of the call relationship, and registers the starting point as an execution path (S502). In the present embodiment, the starting point is the method that is executed first (without being called from another method) in the method calling relationship. The starting point may be any method specified by the user. In the example of FIG. 6, a node 4503 of methodA and a node 4507 of methodH correspond to the origin method. As described above, in the present embodiment, one execution path is configured by only the starting method.

Next, the importance calculation processing unit 105 selects the end point of the execution path from the methods adjacent to the start point (S503). For example, in the example of FIG. 6, the node 4504 of methodB adjacent to the node 4503 of methodA is the end point in the execution path of the node 4503-node 4504. A node 4508 of methodD adjacent to the node 4503 of methodA is an end point in the execution path of the node 4503 to the node 4508.

Next, in step S504, the importance calculation processing unit 105 determines whether there is an unsearched method in the method calling relationship created in the calling relationship construction processing (S302). When there is an unsearched method (S504: YES), the importance calculation processing unit 105 proceeds to step S505. If there is no unsearched method (S504: NO), the process proceeds to step S507.

In step S 505, the importance calculation processing unit 105 stores the execution path from the start point to the end point in the memory 103. As described above, for example, the node 4503-node 4504 is one execution path, and the node 4503-node 4508 is also another execution path.

Next, in step S506, the importance calculation processing unit 105 defines an unexamined method adjacent to the current end point as a new end point. For example, in the example of FIG. 6, a node 4506 of methodC adjacent to the node 4504 of methodB is defined as a new end point. After step S506 ends, the importance calculation processing unit 105 returns to step S504.

If it is determined in step S504 that there is no unsearched method (S504: NO), the importance calculation processing unit 105 selects in step S507 the starting point selected in step S502 and steps S506 and S508 (S508 will be described later). Determine if the end points are different.

If the starting point and the ending point are different (S507: YES), the importance calculation processing unit 105 proceeds to step S508. When the starting point and the ending point coincide with each other (S507: NO), the importance degree calculation processing unit 105 ends the execution route analysis process. In step S508, the importance calculation processing unit 105 returns the end point to the calling method. After step S508 ends, the importance calculation processing unit 105 returns to step S504.

As can be understood from the description with reference to FIG. 7, the end point method is a method at the end of the call tree (a method that does not call other methods, for example, the node 4506 of methodC in the example of FIG. 6), as well as the start method. All methods (including, for example, the node 4504 of methodB) included in the path that continues from to the end method by the call relationship (edge) are end point methods. The registered execution paths are all paths from an origin method to an arbitrary method having a call relationship (edge).

For example, in the path from the starting node 4503 to the node 4506 in FIG. 6, the

nodes

4504 and 4506 are end nodes. Accordingly, one execution path is from the origin node 4503 to the node 4504, and the other one execution path is from the origin node 4503 to the node 4506. Furthermore, in this embodiment, one execution path is configured by only the starting node.

In the execution of a program, a single process can be terminated in any method on the call tree. For example, a certain process is terminated by executing methodA and methodB, and another certain process is terminated by executing methodA, methodB, and methodC. As described above, by registering the path from the starting method to each method connected by the calling relationship as the execution path, it is possible to more appropriately determine the importance of the method included in the execution path and the execution path.

Next, the flow of the importance calculation processing of the execution path (S304 in FIG. 3) will be described with reference to the flowchart in FIG. In the execution path importance level reference value calculation process (S304), the importance level calculation processing unit 105 obtains the execution path importance level reference value analyzed in the execution path analysis process (S303 in FIG. 3).

First, the importance calculation processing unit 105 determines whether or not there is an execution path for which the calculation of the importance reference value of the execution path has not been completed (S701). When there is an execution path for which the calculation of the importance level reference value has not been completed (S701: YES), the importance level calculation processing unit 105 proceeds to step S702. When the calculation of the importance reference value of all execution paths is completed (S701: NO), the importance calculation processing unit 105 ends the importance reference value calculation process (S304) of the execution path.

Next, in step S702, the importance calculation processing unit 105 selects an execution path. Next, in step S703, the sum of importance reference values (values calculated from the characteristic values of the single method) of each method passing from the start method to the end method of the execution path selected in step S702 is obtained. For example, as shown in the example of FIG. 6, when the importance criterion is the number of executions of a method, the number of executions of each method is the importance criterion value of the method itself, and the importance calculation processing unit 105 The total number of execution times of each of the methods constituting the is calculated as the importance reference value of the execution path.

The importance level calculation processing unit 105 calculates the importance level reference value of the node (method) in the call tree information 125 (call tree shown in FIG. 6) based on the calculated importance level reference value (new importance level reference value) of the execution path. Update. In the present embodiment, the importance level calculation processing unit 105 updates the importance level reference value of the end point method in each execution path to the calculated value of the importance level reference value of the execution path. In this way, the trace information acquisition level (determining importance reference value) of the end point method is determined based on the importance of other methods included in the execution path. The importance calculation processing unit 105 stores the determined trace information acquisition level in the trace information acquisition level 1309 of the profile information 111.

In the example of FIG. 6, for example, the importance standard value of the execution path from methodA4503 to methodB4504 is given to methodB4504. The importance reference value of the execution path from methodA4503 to methodC4506 is given to methodC4506. The importance reference value of the execution path composed of method A4503 as the starting method is the same as the importance reference value of method A4503.

In the above example, the importance reference value of the execution path is obtained by the simple sum of the importance reference values of the individual methods included in it, but the importance reference value of the execution path is calculated by the importance reference value of the method alone. An average, a weighted sum (a sum obtained by applying a prescribed weight to the importance standard value of each method), or other statistical indicators may be used. In the calculation of the weighted sum (linear sum), for example, the weighting factor is set so as to decrease from the end point method of the execution path toward the start point method.

In the present embodiment, the elements used for calculating the importance reference value of the execution path are the number of branch instructions included in the method, the number of execution times of the method, the execution time of the method, the execution frequency of the method, and the length of the source code. One or more elements selected from the total number of instructions in the binary file can be used.

The importance increases as the method execution is rarer and the method is more complex. For example, the importance increases with an increase in complexity (for example, the number of branch instructions), execution time, code length, or number of instructions, and the importance decreases with an increase in the number of executions.

Alternatively, the number of arguments or variables that do not match a preset constraint may be used in the method argument, method internal variable value, or method internal variable type. In addition, this embodiment may use user related information for determining the importance of the execution path. As the user-related information, information related to the user who uses the system such as job position information, job information, and department information can be used.

The importance calculation processing unit 105 may calculate a weighted sum of arbitrary elements in the method characteristic information as the importance of the execution path. For example, the weighting coefficient is set so as to decrease from the end point method of the execution path toward the start point method. In addition, an arbitrary value in the method characteristic information may be a vector, and an output value when the vector is input to a discriminator created by machine learning may be used as the importance of the execution path.

Next, the trace information acquisition level determination process (S305 in FIG. 3) will be described with reference to the flowchart in FIG. The trace information acquisition level determination process (S305) is a process for obtaining the trace information acquisition level of the method included in the execution path from the importance reference value of the execution path calculated in the execution path importance calculation process (S303). .

In this configuration, the trace information of the end method of the execution path is determined from the importance reference value of the execution path (the importance reference value of the method in determining the trace acquisition level). The probe inserted into the method acquires the type of trace information indicated by the set trace information acquisition level when executing the method.

First, in step S801, the importance calculation processing unit 105 determines whether there is an execution path whose trace information acquisition level has not been determined. If there is an undetermined execution path (S801: YES), the importance calculation processing unit 105 proceeds to step S802. When the determination of the trace information acquisition level is completed in all execution paths (S801: NO), the importance calculation processing unit 105 ends the trace information acquisition level determination process (S305).

Next, in step S802, the importance calculation processing unit 105 selects an execution path whose trace information acquisition level is not determined. Next, in step S803, referring to the trace information acquisition level / importance level reference value correspondence table 121, the type of information (trace information acquisition level) acquired at the time of trace execution is determined from the importance level reference value of the execution path.

In the trace information acquisition level / importance level reference value correspondence table 121 shown in FIG. 5, the threshold value of the importance level range 2001 may be a value other than a statistic, or may be a value obtained by machine learning or a user-specified value. . The threshold value of the importance level range 2001 is updated after the profile information 111 is updated (S106 in FIG. 2) (S107 in FIG. 2). The number of levels of the trace information acquisition level 2002 and the type of information 2003 to be acquired in FIG. 5 can be changed by the user.

Next, in step S804, the importance calculation processing unit 105 updates the value of the trace information acquisition level of the end point method in the execution path selected in step S802 to the value of the trace information acquisition level determined in step S803. The importance calculation processing unit 105 returns to step S801 after executing step S804. Through the above processing, the type of trace information acquired in the method is updated.

In a preferred example, when the value of the trace information acquisition level is different from before and after the update in step S804, the importance calculation processing unit 105 prioritizes the level value with which the trace acquisition information is simpler. For example, in the example illustrated in FIG. 5, the trace information to be acquired is simplified in the order of “simple”, “standard”, and “detail”.

When a single method is called from multiple methods, that method becomes the end point method of multiple different execution paths. The importance calculation processing unit 105 selects an importance reference value to be applied to the end point method from the importance reference values of the plurality of execution paths according to a preset rule.

In a preferred example, the importance calculation processing unit 105 selects a value having the lowest importance reference value. That is, the race information acquisition level that is the simplest (the smallest type of acquired trace information) is selected. This makes it possible to reduce the amount of trace information to be acquired while acquiring necessary trace information.

(Program extension processing (S104))
Next, the program extension process (S103) shown in the flowchart of FIG. 2 will be described with reference to the flowchart of FIG. In this process, the program extension processing unit 104 embeds a probe for acquiring trace information in the trace target program 109.

As shown in FIG. 11, first, in step S901, the program extension processing unit 104 determines whether there is an unexamined trace target class. When there is an uninvestigated class (S901: YES), the program extension processing unit 104 proceeds to step S902. If all classes have been checked (S901: NO), the program extension processing unit 104 proceeds to step S905.

Next, in step S902, the program extension processing unit 104 reads the trace target class from the program 109 on the memory 103. Next, in step S903, it is determined whether or not there is a method in which a probe for acquiring trace information is not inserted in the class selected in step S902.

If there is a method that has not been inserted (S903: YES), the program extension processing unit 104 proceeds to probe insertion step S904. When the insertion processing for all the methods in the class has been completed (S903: NO), the program extension processing unit 104 returns to step S901.

Next, in step S904, the program extension processing unit 104 inserts a probe based on the trace information acquisition level obtained in the importance calculation process (S102 in FIG. 2) into the method belonging to the class read in step S902. After the probe insertion (S902), the program extension processing unit 104 returns to step S903.

Next, in step S905, the program expansion processing unit 104 outputs the program in which the probe is inserted in step S904 to the memory 103 as the expansion program 110. Finally, in step S906, the program extension processing unit 104 loads the class into which the probe has been inserted in the extension program 110. After step S906 ends, the program expansion process (S103) ends.

The probe insertion process (S904 in FIG. 11) will be described with reference to FIG. The probe insertion process (S904) inserts a probe based on the trace information acquisition level obtained in the importance calculation process (S102 in FIG. 2). First, in step S1001, the program extension processing unit 104 reads a method for inserting a probe.

Next, in step S1003, the program extension processing unit 104 inserts a method start information acquisition probe at a location corresponding to the first line of the method in the bytecode storage unit of the method read in step S1001. Next, in step S1004, the program extension processing unit 104 determines whether the trace information acquisition level obtained in the importance level calculation process (S102) is “standard” or “detailed” (hereinafter referred to as a normal state). To do.

When the trace information acquisition level is in the normal state (S1004: YES), the program extension processing unit 104 proceeds to step S1005. If not in the normal state (S1004: NO), the program extension processing unit 104 proceeds to step S1014. In step S1005, the program extension processing unit 104 inserts an argument information acquisition probe into the bytecode storage unit of the method read in step S1001. The probe insertion point is, for example, immediately after the method start information acquisition probe.

Next, in step S1014, the program extension processing unit 104 reads one instruction of the byte code of the method read in step S1001. Next, in step S1006, the program extension processing unit 104 determines whether or not to access the internal variable with the byte code read in step S1014. Here, the access to the internal variable in step S1006 refers to both processing of reading the internal variable and rewriting the internal variable.

When accessing the internal variable (S1006: YES), the program extension processing unit 104 proceeds to step S1007. If not accessed (S1006: NO), the program extension processing unit 104 proceeds to step S1009.

Next, in step S1007, the program extension processing unit 104 determines whether or not the trace information acquisition level obtained in the importance calculation processing (S102) is a normal level. When the trace information acquisition level is the normal state (S1007: YES), the program extension processing unit 104 proceeds to step S1008. If it is not the normal state (S1007: NO), the program extension processing unit 104 proceeds to step S1009. Next, in step S1008, the program extension processing unit 104 inserts an internal variable information acquisition probe into the bytecode of the method read in step S1001.

Next, in step S1009, the program extension processing unit 104 determines whether or not the byte code read in step S1014 is a return instruction. In the case of a return instruction (S1009: YES), the process proceeds to step S1010. If the instruction is not a return instruction (S1009: NO), the process returns to step S1014.

Next, in step S1010, the program extension processing unit 104 determines whether or not the trace information acquisition level obtained in the importance calculation processing (S102) is in a normal state. When the trace information acquisition level is the normal state (S1010: YES), the program extension processing unit 104 proceeds to step S1011. If it is not the normal state (S1010: NO), the program extension processing unit 104 proceeds to step S1012.

Next, in step S1011, the program extension processing unit 104 inserts a return value information acquisition probe into the bytecode of the method read in step S1001. In step S1012, the program extension processing unit 104 inserts a method end information acquisition probe into the bytecode of the method read in step S1001.

In step S1013, the program extension processing unit 104 determines whether it is the end of the method. When it is the end of the method (S1013: YES), the program extension processing unit 104 ends the probe insertion process 204. If it is not the last (S1013: NO), the program extension processing unit 104 returns to step S1014.

When a method belongs to a plurality of execution paths, the probe inserted by the above process acquires trace information in the same manner in all of the plurality of execution paths to which the method belongs. Unlike this, the probe may acquire the trace information according to the trace information acquisition level determined for each execution path by specifying the execution path from the call source method and the call destination method. In the same method, the probe acquires detailed trace information in an execution path with high importance, and acquires simplified trace information in an execution path with low importance.

(Profile update (S106) / Correspondence table update process (S107))
Next, the profile update process (S106) and the trace information acquisition level / importance reference value correspondence table update process (S107) in the flowchart of FIG. 2 will be described. The profile update processing unit 118 creates profile information 111 (see FIG. 4) for calculating the importance reference value of the execution path from the analysis result of the trace target program and the acquired trace information 124. Further, the profile information 111 is updated with the value of the new acquired trace information 124.

As shown in FIG. 5, the trace information acquisition level / importance level reference value correspondence table 121 uses statistical data of the importance level reference value, and the profile update processing unit 118 uses the new acquired trace information 124 value. The trace information acquisition level / importance level reference value correspondence table 121 is updated. The trace information acquisition level of the method is updated by the importance calculation process (S102 in FIG. 2) using the updated profile information 111.

(Trace information output process (S109))
Next, the trace information output process (S109) in FIG. 2 will be described with reference to the flowchart in FIG. In this processing, the trace information output processing unit 107 converts the trace information 124 acquired by the probe into output trace information 112 and outputs it to the memory 103.

As shown in FIG. 13, first, in step S1101, the trace information output processing unit 107 takes in the trace information 124 acquired in the trace information acquisition step (see S1403 in FIG. 23). Next, in the trace information conversion step (S1102), the trace information acquired in step S1101 is converted into an output trace information format. In step S1103, the output trace information 112 converted into the output trace information format in the trace data conversion step (S1102) is stored in the memory 103 or the external storage 113.

The operation of the trace information conversion step (S1102) will be described. The trace information conversion process 1104 converts the acquired trace information 124 into output trace information 112. First, one record is read from the acquired trace information 124. Next, when the read information is Call information (information indicating that the method has been called), the information is converted into the Call information format of the output trace information 112. Similarly, when the read information is Return information (information indicating that the method has been executed), the information is converted into the Return information format of the output trace information 112. This process is repeated for all trace information.

FIG. 14 shows a configuration example of the output trace information 112, which is the output result of the trace information. In this example, the output trace information 112 includes a trace information management table 1801, a trace information log 1802, a method ID table 1803, and an exception ID table 1804. Details will be described later with reference to specific examples.

(Screen display process (S111))
The screen display processing unit 115 displays the profile information 111 and the output trace information 112. FIG. 15 shows a display example of the profile information 111. The example of FIG. 15 displays profile information as a call tree. This call tree is the same as the call tree shown in FIG. 6, and detailed description thereof is omitted. An example of detailed information of each node is the same as in FIG. FIG. 16 shows an example of the display 1603 of the trace information 112. Since the trace information 112 is data to be displayed in the order of method execution, it is displayed as a call trace (execution history).

(User specified for trace information acquisition)
As described with reference to FIG. 1, the user can specify the trace acquisition level of a specific execution path (its end point method) by using the trace execution path specification information 122. Further, the trace target range (class, method, package (library), etc.) can be specified (limited) by the trace target specifying information 123. 17A and 17B show an example of the trace execution path designation information 122 and an example of the trace target designation information 123, respectively. As a result, it is possible to realize trace information acquisition and error analysis processing adapted to the user's request.

The trace execution path designation information 122 created and input by the user designates the execution path to be traced and the trace information acquisition level of the execution path. In the example of FIG. 17A, three execution paths and respective trace acquisition levels are specified. For example, the route of method_a-method_b is designated, and “normal” is designated as the trace information acquisition level.

The tracer acquires the type of trace information indicated by the trace information acquisition level set by the user in the execution path to be traced set by the user. Specifically, the Java VM 114 (tracer) reads the trace execution path designation file 122 before the program reading process (S101 in FIG. 2). If the trace execution path designation information 122 does not exist, the tracer does not set the trace acquisition information level of the execution path (end point method) specified by the user.

If the trace execution path designation information 122 exists, the importance calculation processing unit 105 determines that the trace execution path designation information 122 is between step S507 and “end of the execution path analysis process” in the execution path analysis (FIG. 4). The execution path designated by the trace execution path designation information 122 is registered. The importance calculation processing unit 105 assigns information indicating that the execution path designated by the trace execution path designation information 122 to the execution path designated by the user is an execution path designated by the user. May be.

Furthermore, the importance calculation processing unit 105 reads the trace execution path designation file 122 read by the undetermined execution path selected in step S802 between step S802 and step S803 in the trace acquisition level determination process (FIG. 10). Whether or not the execution path is described in the above is determined with reference to the registration information of the execution path.

When the execution path is designated by the user, the importance calculation processing unit 105 refers to the trace execution path designation information 122, sets the trace information acquisition level of the execution path to a corresponding value, and returns to step S801. If the execution path is not designated by the user, the importance calculation processing unit 105 proceeds to step S803. Through the above processing, the trace information acquisition level of the designated execution path can be set to the designated value.

Next, the trace target specifying information 123 will be described. The user can specify the trace target by the trace target specifying information 123 and limit the acquisition target range of the trace information. Specifically, the Java VM 114 (tracer) reads the trace target designation information 123 before the program reading process (S101 in FIG. 2). If the trace target designation information 123 does not exist, the trace acquisition range is not limited.

The program extension processing unit 104 determines whether the method read in the processing step S1001 between the step S1001 and the step S1003 in the probe insertion process (FIG. 12) is the package, class, or class described in the read trace target designation information 123. Determine whether it matches the method. If they match, the program extension processing unit 104 proceeds to step S1003. If they do not match, the probe insertion process is terminated. With the above processing, it is possible to limit the acquisition target range of the trace information.

(Example program processing)
In the following, each of the above processes will be described using the example of the program shown in FIG. 19A to 19I show examples of profile information 111 (see also FIG. 4) used in determining the importance of the execution path of this program example. The importance calculation processing unit 105 refers to the profile information 111 of FIGS. 19A to 19I to determine the trace information acquisition level of the method. As described above, the profile information update processing unit 118 creates and updates the profile information 111 of FIGS. 19A to 19I.

The program example shown in FIG. 18 includes each program (module) of class SampleClassA1401, class SampleClassB1402, class SampleClassC1403, and class ExceptionHandler 1404.

18 is normally executed in the order of method_a (), method_b (), and method_c in the program 109 of FIG. However, in the program of FIG. 18, as an exception, method_a (), method_b (), and throwException () may be executed in this order. The specific contents of each class are as shown in FIG. 18, and a detailed description thereof is omitted here.

19A to 19I show method argument / internal variable information 1301, method call path information 1302, method execution count information 1303, method execution time information 1304, obtained in the execution of the program 109 in FIG. Specific examples of method branch instruction number information 1306, method call relation information 1307, method line number information 1305, method total instruction number information 1308, and trace information acquisition level 1309 are shown. The contents of each information are as shown in each figure, and some information will be supplementarily described below.

The call route information 1302 shown in FIG. 19B includes information on the execution route of the trace target program 109. The description indicated by 1502 in FIG. 19B is executed in the order of method_a (), method_b (int), and method_c (int). The description of reference numeral 1508 in FIG. 19B indicates that the importance level reference value of the execution path is 115.

The method execution time information 1304 shown in FIG. 19D indicates the time taken for one execution of each method, but may be the total execution time of each method. For example, the execution time of one method is an average value of the times of a plurality of executions of the method.

The method call relationship information 1307 shown in FIG. 19F indicates each call relationship between methods and the number of calls in each call relationship. For example, there is a call relationship from method_b (int) to method_c (int), and the number of calls is 39 times.

The importance calculation processing unit 105 constructs a call relationship (call tree) from the method call relationship information 1307 shown in FIG. 19F in the call relationship construction processing (S302 in FIG. 3). FIG. 20 shows an example of a call tree generated from the method call relation information 1307 in FIG. 19F. The importance calculation processing unit 105 can know which method is likely to be called from which method from the call tree of FIG.

Next, in the execution path analysis process (S303 in FIG. 3 and FIG. 7), the importance calculation processing unit 105 extracts the execution path from the constructed call tree and outputs the method call path information 1302. In the case of the call tree of FIG. 20, the extracted execution paths are the four execution paths of the method call path information 1302 of FIG. 19B.

Specifically, they are four execution paths: method_a (), method_a () → method_b (), method_a () → method_b () → method_c (), method_a () → method_b () → throwException (). .

The execution path here is a path obtained by fixing the starting point (the starting point method of the execution path) in the configured call tree and sequentially searching for the end point adjacent to the starting point (the end method of the execution path). .

In the execution path importance level reference value calculation process (S304 in FIG. 3 and FIG. 9), the importance level calculation processing unit 105 executes the execution path importance level reference value for each execution path extracted in the execution path analysis process (S303). Calculate In this example, the importance calculation processing unit 105 obtains the importance reference value of the execution path from the number of executions of the method included therein. The number of execution times of each method is included in the method execution number information 1303 shown in FIG. 19C.

In this example, the importance calculation processing unit 105 calculates a simple sum of the number of execution times of the methods included in the execution path, and uses the value as the importance reference value of the execution path. The calculation result of the importance reference value of each execution path is a numerical value corresponding to each execution path in the method call path information 1302 of FIG. 19B. For example, in the case of the execution path 1502 (execution path executed in the order of method_a (), method_b (), method_c ()), the importance reference value of the execution path is 5 + 45 + 65 = 115 (reference numeral 1508 in FIG. 19B indicates value).

In the trace information acquisition level determination process (S305 in FIG. 3 and FIG. 10), the importance degree calculation processing unit 105 calculates the importance reference value of the execution path and the importance of the trace information acquisition level and the execution path illustrated in FIG. The trace information acquisition level of each execution path is obtained from the degree reference value correspondence table 121.

FIG. 21 shows a correspondence table 121 of trace information acquisition levels and execution path importance reference values corresponding to the profile information 111 of FIGS. 19A to 19I. The average value and the standard deviation of the importance reference values of the execution path are 55.3 and 45.2, respectively. Therefore, the range of the importance reference value of the execution path corresponding to the trace information acquisition level is as shown in FIG. The importance level calculation processing unit 105 determines the trace information acquisition level of each execution path in the method call path information 1302 shown in FIG. 19B using the correspondence table 121 of FIG.

Specifically, the trace information acquisition level of the execution path method_a () is “detailed” because its importance level reference value is 5. The trace information acquisition level of the execution path method_a () → method_b () is “standard” because the importance reference value of the execution path is 50. Similarly, the trace information acquisition level of the execution path method_a () → method_b () → method_c () is “simple”, and the trace information acquisition level of method_a () → method_b () → throwException () is “standard”.

Next, based on the trace information acquisition level of each method obtained by the trace information acquisition level determination process (S305 in FIG. 3) in the program extension process (S103 in FIG. 2 and FIG. 11), the program extension processing unit 104 A probe for acquiring trace information is inserted into each method of 18 example programs. 22A to 22D show examples of results of inserting probes into classes 2801 to 2804 of the program 109 (examples of the extended program 110).

Since the trace information acquisition level of method_a () is “detailed”, it is necessary to acquire the method call relationship, method argument / return value, and internal variable value for method_a (). Probes indicated by

reference numerals

2805, 2806, 2807, and 2808 in FIG. 22A are probes for acquiring trace information inserted into the class 2801. Since these probes are inserted directly into the Java class file, they cannot be viewed from the outside.

Here, in order to simplify the explanation, the probe is described as an API (Application Programming Interface). In FIG. 22A, methodEntry () 2805 is a probe for acquiring method start information. getArgent () 2806 is a probe for acquiring a method argument.

GetLocalValue () 2807 is a probe for acquiring internal variable information. getReturnValue () 2808 is a probe for acquiring return value information. methodExit () 2809 is a probe for acquiring method end information.

Since the trace information acquisition level of method_b () in FIG. 22B is “standard”, a probe for acquiring method call relation information and method argument / return value information is inserted. Specifically, methodEntry (), getArgent (), getReturnValue (), and methodExit () are inserted into class 2802.

Since the trace information acquisition level of method_c () in FIG. 22C is “simple”, a probe for acquiring method call relation information is inserted. Specifically, methodEntry () and methodExit () are inserted into class 2803.

Since the trace information acquisition level of throwException () in FIG. 22D is “standard”, a probe () for acquiring method call relation information and method argument / return value information is inserted. Specifically, methodEntry (), getArgent (), getReturnValue (), and methodExit () are inserted into class 2804.

Next, the program execution processing (S104 in FIG. 2) of the example of the extended program shown in FIGS. 22A to 22D by the program execution processing unit 120 will be described with reference to the flowchart in FIG. As shown in FIG. 23, the program execution processing unit 120 reads one instruction from the extension program 110 in step S1401.

Next, in step S1402, the program execution processing unit 120 determines whether or not the instruction read in step S1401 is a trace information acquisition probe. In the case of the trace information acquisition probe (S1402: YES), the program execution processing unit 120 proceeds to step S1403. If the probe is not a trace information acquisition probe (S1402: NO), the program execution processing unit 120 ends the program execution processing 121. Next, the program execution processing unit 120 acquires trace information in the trace information acquisition process (S1403).

Trace information acquired by the trace information acquisition process (S1403) includes method start information (methodEntry ()), method end information (methodExit ()), method argument information (getArgent ()), and method return value information ( getReturnValue ()) or internal variable information (getLocalValue ()).

The method start information probe is executed at the beginning of the method, and records the method ID (a positive integer value uniquely assigned to each method), the line number of the calling method, and the information acquisition time. The probe of the method end information is called immediately before the return process, and records the method ID, the line number of the calling method, and the information acquisition time. From the probe information, the calling relationship between the method and the caller method can be specified. The information acquisition time recorded by the probe is used to calculate the method execution time and the method execution frequency.

The method argument information probe is inserted and executed immediately after the method start information acquisition probe. The information to be acquired and recorded includes the method ID, argument name, variable type, argument value, and information acquisition time. The probe of the return value information of the method is executed immediately before the method return process. The information to be acquired and recorded is the method ID, return value name, variable type, and return value information acquisition time. As a result, the method argument and return value can be specified.

The probe that acquires internal variable information is inserted immediately before accessing the variable. The information to be acquired and recorded includes a method ID, an internal variable name, an internal variable type, an internal variable value, a caller method line number, an access to the internal variable, and an information acquisition time. Thereby, the value of the internal variable of the method can be specified.

24A and 24B show examples of trace information acquired by the trace information acquisition process (S1403 in FIG. 23) from the program example in FIG. A table 2901 in FIG. 24A is a table of trace information. A table 2902 in FIG. 24B is a method ID table.

Numerals 1 to 8 in the first column of the table 2901 in FIG. 24A are numbers given for explanation. The method ID of the table 2902 in FIG. 24B is uniquely assigned to each method.

The first line (2904 in FIG. 24A) in the table 2901 indicates that a method (method_a ()) having a method ID of 1 has been called. Since the line number of the calling method is unknown, “none” is recorded. The second line in the table 2901 (2905 in FIG. 24A) represents the argument information of the method (method_a ()) whose method ID is 1. Since the method (method_a ()) takes no argument, the name, type, and value of the argument are recorded as “none”.

The third line (2906 in FIG. 24A) in the table 2901 is information when an internal variable is accessed in a method (method_a ()) having a method ID of 1. Since the line number of the caller is unknown, “none” is recorded. “-11” is stored in the variable “id”, which indicates that the variable is accessed in the third line of (method_a ()).

The exception on the eighth line (2907 in FIG. 24A) in the table 2901 is trace information acquired in the exception information acquisition process when an exception occurs. This information indicates that an exception has occurred in the method whose method ID is 4. Also, the exception catch information is “none”, indicating that no exception was caught by any method.

25A to 25C show examples of the output trace information 112 when the execution of the program example of FIG. 18 is traced. In this example, the trace information management table 1801 in FIG. 25A stores the version number of the tracer, the process ID of the Java virtual machine, the trace execution ID (ID issued by the tracer when the trace is executed), the number of threads, and the thread ID. To do. In this example, the trace information management table 1801 is generated at the start of tracing.

The method ID table 1803 in FIG. 25B stores a method name and a method ID corresponding to the method name. In this example, the method ID table 1803 is generated and appended when a class is read. The trace information log 1802 in FIG. 25C shows the type of trace information (method start / end, exception), the caller's method ID (the method ID in which an exception occurred in the case of an exception), and the callee's method ID (in the case of an exception). Stores the ID of the method that caught the exception.

FIGS. 26A and 26B show examples of trace information acquired in the execution of the program of FIG. However, the trace information in FIG. 26A and FIG. 26B shows the result of changing the argument of the program 109 in FIG. FIG. 26A is a call trace 1901 generated by the method of this embodiment. FIG. 26B shows a trace 1902 generated by a conventional method (a method for acquiring information in detail by all methods).

26A and 26B are compared, the trace information in FIG. 26B displays data indicated by

reference numerals

1903, 1904, and 1905. Since data (information) 1903, 1904, and 1905 are data that are not used for specifying an error, they are not necessary for failure analysis. On the other hand, the trace information shown in FIG. 26A does not include the unnecessary data, and the technique according to the present embodiment shows that the trace information can be efficiently acquired as compared with the trace information of the conventional technique. .

In an actual application, method calls are deep and the number of executions is very large. Therefore, if a trace is acquired by a conventional method, there is a high possibility that the amount of trace information will be enormous.

In contrast, in the present embodiment, the importance of the execution path is determined from the importance of a plurality of procedures included in the execution path, and the type of trace acquisition information in the method included in the execution path is determined according to the importance of the execution path. To decide.

As described above, this embodiment obtains detailed information on a method with high importance from the viewpoint of the importance of an execution path including other methods, not the importance of a single method, and information on methods that are not important. To reduce the amount of trace information data. For this reason, the present embodiment can reduce the data amount of the trace information while appropriately keeping important trace information.

The above configuration applies the importance of the execution path only to the end point method in the specified execution path. In the above configuration, all methods are end point methods of any execution path, and importance (acquired information type) of all methods can be determined and updated by the above application. Depending on the design, the importance of the execution path may be applied to a plurality of methods included in one execution path.

As described above, the importance of the execution path is preferably determined using the importance (trace information) of all the methods included therein, but depending on the design, it is selected according to a preset rule. The importance of the execution path may be determined from the importance of some methods. When there is no user designation, it is preferable to repeat the importance calculation for all the execution paths of the trace target program, but it may be performed only in a part of the range by design.

[Second Embodiment]
In the following, a second embodiment of the present invention will be described. In the present embodiment, the importance (trace acquisition level) of the call source method is determined using the importance of the call destination method (trace acquisition level in the following example). As a result, when the method of the caller adjacent to the method with high importance of the method is low in importance and the method with high importance of the method is isolated on the call tree, the execution path including them is changed. It can be prevented that it is not extracted as an important execution path. In the following example, a trace acquisition level is used as a value representing importance, but other values, for example, importance reference values may be used as values representing importance.

FIG. 27 schematically shows the configuration of the computer according to the present embodiment. The Java VM 114 of this embodiment includes a program monitoring processing unit 119 in addition to the configuration of the first embodiment. Further, the program monitoring processor 119 stores the monitoring information 126 in the memory 103. In the following, differences from the first embodiment will be mainly described.

In this example, the program monitoring processing unit 119 has a method importance (trace information acquisition level in the following example) set in the execution path analysis process (see S303 in FIG. 3) greater than or equal to a specified value (eg, “details”). Monitor the method.

Specifically, it is monitored whether or not a method whose importance is equal to or higher than a specified value (threshold) is called. The program monitoring processing unit 119 changes the trace information acquisition level of a method (caller method) that calls a monitored method when a predetermined condition is satisfied, and stores information indicating the change in the monitoring information 126. .

The profile information update processing unit 118 refers to the monitoring information 126 in the update of the profile information 111, acquires information that identifies the method whose trace information acquisition level has been changed and the trace information acquisition level. Is stored in the trace information acquisition level 1309 (see FIG. 19I). The program extension processing unit 104 inserts a probe corresponding to the trace information acquisition level for a method whose trace information acquisition level indicated by the profile information 111 is changed.

In this way, by increasing the trace information acquisition level of the caller method according to the trace information acquisition level of the callee method (by increasing the level of detail of the trace information to be acquired), When a problem occurs, the cause can be estimated from the trace information in the execution path.

The operation of the program monitoring processing unit 119 (program monitoring processing (S112)) will be described with reference to the flowchart of FIG. In the present embodiment, after executing the program execution process (S104) in the overall process flow of FIG. 28, the program monitoring processing unit 119 executes the program monitoring process (S112).

As shown in the flowchart of FIG. 29, the program monitoring processing unit 119 determines in step S1601 whether or not a method having an importance level equal to or higher than a specified value has been called. When a method having an importance level equal to or higher than a specified value is called (S1601: YES), the program monitoring processing unit 119 proceeds to step S1602. If not called (S1601: NO), the program monitoring processor 119 ends the program monitoring process.

Next, in step S1602, the program monitoring processing unit 119 determines whether or not a method whose importance is equal to or higher than a specified value is a starting method. When the method is the starting method (S1602: NO), the program monitoring processing unit 119 ends the program monitoring processing. If it is not the starting method (S1602: YES), the program monitoring processor 119 proceeds to step S1603.

Next, in step S1603, the program monitoring processor 119 determines whether or not a performance failure has occurred in the Java virtual machine 114. In the present embodiment, the performance failure is assumed to be when a performance index such as a CPU usage rate, a memory usage amount, or a heap memory usage amount of the Java virtual machine 114 exceeds a preset threshold.

If it is determined in step S1603 that a performance failure has occurred (S1603: YES), the program monitoring processor 119 initializes the monitoring information 126 and ends the program monitoring process. If it is determined that no performance failure has occurred (S1603: NO), the program monitoring processor 119 proceeds to step S1604.

Next, in step S 1604, the program monitoring processing unit 119 increases the trace information acquisition level of the caller of the method having high importance, and stores information indicating this in the monitoring information 126. After the end of step S1604, the program monitoring processor 119 ends the program monitoring process.

Thus, it is possible to more effectively support the discovery of the cause of the failure by efficiently collecting the trace information and refining the trace information in the execution path having a high importance from the trace information.

The above example raises the trace information acquisition level of the calling method each time a method whose trace information acquisition level is higher than the specified value is called. When the importance change condition includes a method call, it is possible to acquire trace information useful for failure analysis while suppressing an increase in acquired trace information.

In the above example, the specified number of calls, which is a condition for raising the caller method, is one, but the specified number may be multiple. The condition for changing the trace information acquisition level of the calling method may not include the method call and may include an event different from this. The program monitoring processor 119 may raise the trace information acquisition level of the caller method (the importance of the caller method) to match the trace information acquisition level of the callee method (the importance of the callee method), The trace information acquisition level of the caller method may be raised to a value lower than the trace information acquisition level of the callee method.

Hereinafter, the trace information of the program example of FIG. 18 will be described with reference to FIGS. 30A to 30C and FIGS. 31A to 31C. It is assumed that the profile information 111 for the program example in FIG. 18 holds the information in FIGS. 30A to 30C. The importance calculation standard is the number of executions, and the importance standard value of each method shown in FIGS. 30A to 30C is the number of executions of the method itself. A smaller importance number (less execution times) indicates a more important method (for failure analysis).

30A to 30C are examples of display screens of the profile information 111. FIG. The trace information acquisition level of each method is determined based on the importance of the method. The determined trace information acquisition level is displayed as the trace information acquisition level in each method of FIGS. 30A to 30C. In order to make the process of changing the trace information acquisition level easier to understand, methods whose trace information acquisition level is “detailed” are displayed with a thick frame.

ExceptionExceptionHandler. Is a method with high importance. Once the throwException () 2103 is executed, the ExecutionHandler. SampleClassB., which is the method of the caller of throwException (). The trace information acquisition level of method_b (int) 2102 is raised.

In this embodiment, in order to maximize the trace information acquisition level when a method with high importance is executed, SampleClassB. The trace information acquisition level of method_b (int) 2103 is set to “detail”. FIG. 31A shows an example 2201 of trace information at this time. In the information 2201 in FIG. 31A, it is difficult to specify the cause of the failure.

ExceptionException. If throwException () is executed twice, SampleClassB. SampleClassA., which is a method of calling method_b (int). The trace information acquisition level of method_a () is also set to “detail”.

FIG. 31B shows a trace information example 2202 at this time. The information 2202 in FIG. 31B is more detailed than the information in FIG. 31A, but it cannot be determined whether the cause of the failure is in method_a () or method_b (int).

ExceptionException. If throwException () is executed three times, SampleClassA. Since there is no method that calls method_a (), nothing is done. FIG. 31C shows an example of trace information 2203 at this time. From this information, the cause of the failure is SampleClass. It can be seen that the cause is that a negative number is assigned to id which is an internal variable of method_a ().

As described above, according to the processing of the present embodiment, detailed trace information of not only the method at the fault location but also other methods in the execution path in which the fault has occurred can be acquired. It is possible to more easily identify the failure occurrence process.

In the above example, the acquisition information type of each method is determined from the importance of the method alone, and then the trace information acquisition level importance of the caller method is updated according to the trace information acquisition level of the callee method. In another configuration example, as described in the first embodiment, the trace acquisition level (importance) in the method is determined from the importance of the execution path, and the acquisition level can be updated by the method of this embodiment. it can.

As mentioned above, although embodiment of this invention was described, this invention is not limited to said embodiment. A person skilled in the art can easily change, add, and convert each element of the above-described embodiment within the scope of the present invention.

Some of the above-described configurations, functions, processes, processing units, etc. may be realized by hardware. Information including programs and data for realizing each function is a storage device having a non-temporary storage medium such as a nonvolatile semiconductor memory, hard disk, SSD (Solid State Drive), or an IC card, SD card, DVD, etc. Can be stored in a computer-readable non-transitory storage medium.

The program is executed by the processor to perform the specified processing using the memory and port. In the above-described embodiment, the description with the processing unit (program) as the subject may be an explanation with the processor as the subject. The processing executed by the program is processing performed by a computer and a computer system on which the program operates. Some of the components in the above embodiment can be omitted within the scope of the present invention, and the components in different embodiments can be used in one system.

101 computer, 102 CPU, 103 memory, 114 Java virtual machine, 104 program expansion processing section, 105 importance calculation processing section, 107 trace data output processing section, 109 program, 110 expansion program, 111 profile information, 112 output trace information, 115 screen display processing unit, 116 program reading processing unit, 118 profile information update processing unit, 119 program monitoring processing unit, 120 program execution processing unit, 121 trace information acquisition level / importance level reference value correspondence table, 122 tracing target designation information, 123 Trace execution path designation information, 124 Acquired trace information, 125 Call tree information, 126 Monitoring information

Claims

A computer system including a computer that executes a program is a trace information acquisition method for acquiring trace information in the execution of the program,
In each of a plurality of procedures included in the execution path in the execution of the program, obtain the type of trace information set for the procedure,
Store the acquired trace information in a storage area,
Determining the importance of each of the plurality of procedures using trace information stored in the storage area;
Using the importance of the procedures included in the plurality of procedures, determine a new importance that determines the type of trace information to be acquired in other procedures included in the plurality of procedures,
According to the new importance determining the type of the determined trace information, update the type of trace information acquired in the other procedure,
In the execution of the other procedure, the updated type of trace information is acquired.
The trace information acquisition method according to claim 1,
The determination of the new importance is
Using the determined importance of each of the plurality of procedures, determine the importance of the execution path,
The trace information acquisition method, wherein the importance of the execution path is determined as the new importance that determines the type of trace information acquired in the other procedure.
The trace information acquisition method according to claim 2,
The determination of the new importance is
Determining the importance of each of the procedures included in the execution path using trace information in the procedure stored in the storage area;
Determining the importance of the execution path using the importance of each of all the procedures;
The trace information acquisition method, wherein the importance of the execution path is determined as the new importance of the end point procedure in the execution path which is the other procedure.
The trace information acquisition method according to claim 3,
The determination of the new importance is
A value representing the importance of each of the procedures included in the execution path is determined using the trace information stored in the storage area acquired in the procedure,
A trace information acquisition method, wherein the importance of the execution path is determined using a weighted sum of values representing importance of all the procedures.
The trace information acquisition method according to claim 1,
The plurality of procedures include a first procedure having a higher importance than the other procedures and called by the other procedures;
The trace information acquisition method according to claim 1, wherein the determination of the new importance level increases the importance level of the other procedure according to a predetermined condition.
The trace information acquisition method according to claim 5,
A trace information acquisition method characterized by sequentially increasing the importance of a caller procedure having a lower importance than that of a callee procedure in the execution path according to the predetermined condition.
The trace information acquisition method according to claim 6,
The predetermined condition includes that the number of calls of the callee procedure by the caller procedure reaches a specified number.
The trace information acquisition method according to claim 1,
Storing execution path importance designation information for designating an execution path in a plurality of execution paths and importance of the execution path in the program execution in a storage area;
A trace information acquisition method, wherein the type of trace information acquired in the specified execution path procedure is determined according to the specified importance level with reference to the execution path importance level specification information.
The trace information acquisition method according to claim 1,
Storing acquisition range specification information for specifying a range for acquiring trace information in the program execution in a storage area;
A trace information acquisition method for acquiring trace information limited to the specified range with reference to the acquisition range specification information.
The trace information acquisition method according to claim 1,
A trace information acquisition method, comprising: displaying an update result of the updated type of trace information on a display device.
A computer system for acquiring trace information in the execution of a program by a computer,
In each of a plurality of procedures included in the execution path in the execution of the program, a trace information acquisition unit that acquires the type of trace information set for the procedure;
A trace information storage area for storing the acquired trace information;
The importance level of the procedures included in the plurality of procedures is determined using the trace information stored in the storage area, and acquired in the other procedures included in the plurality of procedures using the determined importance level. An importance determination unit that determines the importance of determining the type of trace information to be performed,
A computer system comprising: an acquired trace information determination unit that updates a type of trace information acquired in the other procedure according to an importance level for determining the determined type of trace information.
12. A computer system according to claim 11, comprising:
The importance determining unit
Determining the importance of each procedure included in the plurality of procedures using the trace information stored in the storage area;
Determining the importance of the execution path;
The computer system, wherein the importance level of the execution path is determined as the new importance level that determines the type of trace information acquired in the other procedure.
12. A computer system according to claim 11, comprising:
The importance determining unit
Determining the importance of the first procedure included in the execution path using the trace information in the procedure stored in the storage area;
A computer system characterized in that, according to a predetermined condition, the first procedure is called in the execution path to increase the importance of the other procedure having a lower importance than the first procedure.
A program that causes a computer system including a computer that executes a program to execute processing for acquiring trace information in the execution of the program,
In each of a plurality of procedures included in the execution path in the execution of the program, obtain the type of trace information set for the procedure,
Store the acquired trace information in a storage area,
Determining the importance of each of the plurality of procedures using trace information stored in the storage area;
Using the importance of the procedures included in the plurality of procedures, determine a new importance that determines the type of trace information acquired in other procedures included in the plurality of procedures,
According to the new importance determining the type of the determined trace information, update the type of trace information acquired in the other procedure,
In the execution of the other procedure, the program includes obtaining the updated type of trace information.
The program according to claim 14, wherein
The determination of the new importance is
Using the determined importance of each of the plurality of procedures, determine the importance of the execution path,
The program which determines the said importance of the said execution path | route as the said new importance which determines the kind of trace information acquired in the said another procedure.