Methods, systems, storage media, and program products for identifying the root causes of anomalies

By using service agents and system agents in the target application system, combining the operating status data and call relationships of multiple modalities, the abnormal analysis results are automatically determined, and the problems of high manual labeling cost and inaccurate detection are solved, achieving high efficiency , accurate abnormal root cause location.

CN118473904BActive Publication Date: 2024-11-22ALIBABA CLOUD COMPUTING CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410920405.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-10
Publication Date
2024-11-22
Estimated Expiration
2044-07-10

AI Technical Summary

Technical Problem

The prior art has problems such as high manual labeling cost and inaccurate abnormal detection in the positioning of abnormal root causes, which affects the accuracy of the positioning results of the root cause.

Method used

By setting up service agents and system agents in the target application system, using the operation status data of multiple modals and the call relationship between the service modules, the abnormal analysis results of the service module are automatically determined, and comprehensively analyzed to determine system abnormal analysis As a result, the need for manual labeling is reduced and accuracy is improved.

Benefits of technology

It realizes efficient and accurate abnormal root cause determination, reduces costs, improves the robustness and accuracy of system abnormality analysis, and avoids the need for manual labeling.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118473904B_ABST
    Figure CN118473904B_ABST
Patent Text Reader

Abstract

The present invention provides a method, system, storage medium and program product for determining the root cause of an abnormality, and relates to the field of artificial intelligence technology. The method includes: a first service agent corresponding to a first service module in a target application system obtains first available analysis information of the first service module, obtains neighbor available analysis information of a neighbor service module from a neighbor service agent according to the calling relationship of multiple service modules, determines the module abnormality analysis result of the first service module according to the first available analysis information of the first service module and the neighbor available analysis information of the neighbor service module, sends the module abnormality analysis result of the first service module to a system agent, so that the system agent determines the system abnormality analysis result of the target application system according to the module abnormality analysis results of the corresponding service modules sent by multiple service agents. The embodiment of the present invention can accurately determine the system abnormality analysis result of the target application system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of Internet technology, and in particular to a method, system, storage medium and program product for determining an abnormality root cause. Background Art

[0002] Abnormal root cause location is a key step in the operation and maintenance of many application systems (such as cloud computing systems). In actual applications, a common way to locate the root cause of an abnormal application system is to collect some abnormal operation status data generated during the operation of the application system, such as log data, and input the abnormal operation status data into a trained neural network model to complete the root cause location through the neural network model.

[0003] First, in the above-mentioned method of locating the root cause of anomalies, a large number of training samples need to be manually labeled, that is, the abnormal root cause information corresponding to the abnormal operating status data is labeled, and the labeling cost is high.

[0004] Secondly, an anomaly detection algorithm (such as the k-σ anomaly detection algorithm) is used to detect anomalies in a certain operating state data in advance. When the operating state data is determined to be abnormal, the neural network model is input to locate the root cause. If the anomaly detection result is inaccurate, it will inevitably affect the accuracy of the root cause location result.

[0005] In view of this, it is urgent to propose an efficient and accurate method to determine the root cause of abnormalities. Summary of the invention

[0006] The embodiments of the present invention provide a method, system, storage medium and program product for determining an abnormality root cause, which are used to efficiently and accurately determine a system abnormality analysis result of a target application system.

[0007] In a first aspect, an embodiment of the present invention provides a method for determining an abnormality root cause, which is applied to a first service agent corresponding to a first service module in a target application system, wherein the target application system includes multiple service modules, each service module corresponds to a service agent, and the first service module is one of the multiple service modules in the target application system in which abnormal operation status data is detected, and the method includes:

[0008] Acquire first available analysis information of the first service module, where the first available analysis information includes abnormal operation status data of the first service module;

[0009] Obtaining neighbor availability analysis information of a neighbor service module from a neighbor service agent according to the calling relationship of the multiple service modules, so as to determine a module abnormality analysis result of the first service module according to the first availability analysis information of the first service module and the neighbor availability analysis information of the neighbor service module, wherein the neighbor service module is a service module having a calling relationship with the first service module, and the neighbor availability analysis information includes abnormal operation status data of the neighbor service module;

[0010] The module exception analysis result of the first service module is sent to the system agent, so that the system agent determines the system exception analysis result of the target application system according to the module exception analysis results of the corresponding service modules sent by each of the multiple service agents, and the system exception analysis result indicates the target service module that causes the abnormal operation of the target application system and the root cause of the exception of the target service module.

[0011] In a second aspect, an embodiment of the present invention provides an abnormality root cause determination device, which is applied to a first service agent corresponding to a first service module in a target application system, wherein the target application system includes multiple service modules, each service module corresponds to a service agent, and the first service module is one of the multiple service modules in the target application system in which abnormal operation status data is detected, and the device includes:

[0012] An acquisition module, configured to acquire first available analysis information of a first service module, wherein the first available analysis information includes abnormal operation status data of the first service module;

[0013] an analysis module, configured to obtain neighbor available analysis information of a neighbor service module from a neighbor service agent according to a calling relationship among a plurality of service modules, so as to determine a module abnormality analysis result of the first service module according to the first available analysis information of the first service module and the neighbor available analysis information of the neighbor service module, wherein the neighbor service module is a service module having a calling relationship with the first service module, and the neighbor available analysis information includes abnormal operation status data of the neighbor service module;

[0014] A determination module is used to send the module exception analysis result of the first service module to the system agent, so that the system agent determines the system exception analysis result of the target application system according to the module exception analysis results of the corresponding service modules sent by each of the multiple service agents, and the system exception analysis result indicates the target service module that causes the abnormal operation of the target application system and the root cause of the exception of the target service module.

[0015] In a third aspect, an embodiment of the present invention provides a system for determining an abnormality root cause, the system comprising:

[0016] The abnormal data detection unit is used to detect the operation status data of each service module in the target application system to determine the abnormal operation status data of each of the multiple service modules;

[0017] A plurality of service agents corresponding to the plurality of service modules, wherein the service agent corresponding to any service module is used to: obtain available analysis information of any service module, obtain neighbor available analysis information of a neighbor service module from a neighbor service agent according to the calling relationship of the plurality of service modules, so as to determine a module abnormality analysis result of any service module according to the available analysis information of any service module and the neighbor available analysis information of the neighbor service module, wherein the neighbor service module is a service module having a calling relationship with any service module, the available analysis information of any service module includes abnormal operation status data of any service module, and the neighbor available analysis information includes abnormal operation status data of the neighbor service module;

[0018] A system agent is used to obtain module exception analysis results of the multiple service modules through the multiple service agents, and determine the system exception analysis results of the target application system according to the module exception analysis results of the multiple service modules, wherein the system exception analysis results indicate the target service module that causes the abnormal operation of the target application system and the root cause of the exception of the target service module.

[0019] In a fourth aspect, an embodiment of the present invention provides a non-temporary machine-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor executes the method for determining the root cause of an abnormality as described in the first aspect.

[0020] In a fifth aspect, an embodiment of the present invention provides a computer program product, comprising: a computer program, which, when executed by a processor of an electronic device, causes the processor to execute the method for determining a root cause of an abnormality as described in the first aspect.

[0021] The abnormal root cause determination method provided by the embodiment of the present invention can more comprehensively reflect whether the service module has abnormal operation status data by detecting whether each service module in the target application system uses one or more modal operation status data to detect whether it is abnormal. For multiple service modules with abnormal operation status data, a service agent (agent based on a large language module) is set accordingly. For the first service module with abnormal operation status data, the corresponding first service agent comprehensively considers the abnormal operation status data of the first service module and the abnormal operation status data of its abnormal neighbor service module when determining the module abnormal analysis result of the first service module, that is, the call relationship between the first service module and the neighbor service module is considered, thereby ensuring the accuracy of the obtained module abnormal analysis result. In addition, the use of multiple modal operation data makes it possible that even if the abnormal operation status data detection result of the operation status data of a certain mode of some service modules is incorrect, the service agent can accurately determine the abnormal analysis result of the corresponding service module based on the abnormal operation status data of other modes and the abnormal operation status data of various modes of the neighbor service module, which can better resist the influence of the incorrect detection result of some abnormal operation status data. After determining the module anomaly analysis result of the first service module, the module anomaly analysis results corresponding to multiple service modules are comprehensively analyzed by using a system agent to obtain a target service module indicating the abnormal operation of the target application system and the root cause of the abnormality of the target service module, so that a more accurate system anomaly analysis result can be obtained. The embodiment of the present invention adopts a two-level agent configuration of a service agent and a system agent. In the process of determining the root cause of the abnormality of the target application system, each agent can intelligently output its own judgment information based on a large language model, without the need for manual labeling operations, with low cost and high efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 A schematic diagram of the structure of an abnormality root cause location system provided by an embodiment of the present invention;

[0024] Figure 2 A schematic diagram of an operating environment of an abnormality root cause locating system provided by an embodiment of the present invention;

[0025] Figure 3 A schematic diagram of the composition of a target application system for which abnormality root cause location is required according to an embodiment of the present invention;

[0026] Figure 4 A flowchart of a method for detecting abnormal operating status data provided by an embodiment of the present invention;

[0027] Figure 5 A flowchart of a method for determining an abnormal root cause provided by an embodiment of the present invention;

[0028] Figure 6a A flowchart of an abnormal analysis result of a determination module provided by an embodiment of the present invention;

[0029] Figure 6b A schematic diagram of an application of determining module abnormal analysis results provided by an embodiment of the present invention;

[0030] Figure 7a Another flow chart of determining module abnormality analysis results provided by an embodiment of the present invention;

[0031] Figure 7b A schematic diagram of an application of determining module abnormal analysis results provided by an embodiment of the present invention;

[0032] Figure 8a Another flow chart of determining module abnormality analysis results provided by an embodiment of the present invention;

[0033] Figure 8b A schematic diagram of an application of determining module abnormal analysis results provided by an embodiment of the present invention;

[0034] Fig. 9 A schematic diagram of the structure of an abnormality root cause determination device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of the present invention are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0037] Some embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the case where there is no conflict between the embodiments, the following embodiments and the features in the embodiments can be combined with each other. In addition, the step sequence in the following method embodiments is only an example and not a strict limitation.

[0038] First, the terms or concepts involved in the embodiments of the present invention are explained:

[0039] AI Agent: It is an intelligent agent system with a large language model as the core controller. Its essence is an agent system that controls the large language model to solve user problems. It relies on the large language model as its core decision-making and processing unit, and can use the logical reasoning ability of the large language model to realize the decomposition of the solution steps of user problems, tool calls and summary of the solution results of multiple steps.

[0040] The embodiment of the present invention provides the following ideas to solve the problem of locating the root cause of abnormalities in the application system: perform abnormal detection on the operating status data corresponding to each service module in the target application system to obtain multiple service modules with abnormal operating status data. By setting a service agent corresponding to each abnormal service module, a service agent is used to determine the abnormal analysis result of the service module corresponding to itself. When the service agent is working, it will comprehensively consider the abnormal operating status data of the service module corresponding to itself and the neighboring modules based on the calling relationship of multiple service modules, and determine the module abnormal analysis result of the service module corresponding to itself. After that, each service agent sends the module abnormal analysis result determined by itself to the global system agent, so that the system agent determines the system abnormal analysis result based on the module abnormal analysis results of the corresponding service modules sent by multiple service agents, which indicates the target service module that causes the abnormal operation of the target application system and the abnormal root cause of the target service module.

[0041] The following is an introduction to the abnormality root cause determination solution provided by an embodiment of the present invention.

[0042] Figure 1 A schematic diagram of the structure of an abnormality root cause location system provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the system includes: an abnormal data detection unit, a plurality of service agents corresponding to a plurality of service modules that detect abnormal operation status data in the target application system (i.e. Figure 1 Service agent 1-service agent N corresponding to service module 1-service module N in the system, as well as system agent.

[0043] The above-mentioned abnormal root cause location system is used to determine the abnormal root cause of the target application system, and the target application system may include several service modules. The abnormal data detection unit is used to detect the operating status data of each service module in the target application system to determine the abnormal operating status data of each of the multiple service modules. When performing the detection, optionally, the operating status data of multiple modes of the service module can be detected.

[0044] It should be noted that the abnormal operation status data of a service module in the embodiment of the present invention refers to some evaluation data that can be defined to evaluate whether the operation status of the service module is abnormal. These evaluation data are called operation status data, such as the CPU utilization, response time, task success rate, etc. mentioned below. If the actual detected operation status data of a service module reflects that the service module is working normally, then the operation status data is normal operation status data, otherwise, it is abnormal operation status data. It can be understood that these operation status data are evaluation data associated with the service module, rather than data directly output during the operation of the service module.

[0045] It should be noted that, in order to facilitate description, in the embodiments of the present invention, Figure 1 Only the service agents corresponding to the N abnormal service modules (i.e., the service modules that detect abnormal operation status data) are illustrated. In fact, a corresponding service agent can be set for each service module included in the target application system, but when some of the service modules are detected to have no abnormal operation status data, the corresponding service agent is not triggered for use.

[0046] In summary, each service agent is used to determine the module anomaly analysis result of the service module to which it corresponds, that is, the anomaly analysis result at the service module level. The system agent is used to summarize the module anomaly analysis results of each service agent to obtain the system anomaly analysis result, that is, the anomaly analysis result of the target application system as a whole. The system anomaly analysis result will provide the root cause determination information that causes the target application system to run abnormally: the target service module that causes the target application system to run abnormally and the abnormal root cause of the target service module.

[0047] In practical applications, the above three components can be used as cloud services and deployed in the same or different computing nodes (i.e., cloud servers) in the cloud. For example, Figure 2As shown, the abnormal data detection unit can be deployed as a service in computing node 201, N service agents corresponding to N service modules are deployed as N services in computing node 202, and the system agent is deployed as a service in computing node 203. The entire abnormal root cause locating system can provide a service interface for users to call, such as inputting access address information related to the target application system to trigger the abnormal root cause locating system to locate the abnormal root cause of the target application system.

[0048] Optionally, when each of the above services is deployed on a corresponding computing node, the operating environment of the service can be configured in the computing node, such as building a container group (pod) corresponding to each service in the computing node and allocating corresponding memory resources, processor resources, etc. Figure 2 As described in, the above different services need to interact with each other so that the abnormal root cause of the target application system can be located through the cooperation of these services.

[0049] Understandably, Figure 2 The correspondence between the services and computing nodes shown in the figure is only an example and is not limited to this.

[0050] In addition, the target application system in the embodiment of the present invention may be an application, such as a video application, an e-commerce application, etc. In this case, the service modules contained therein may be functional modules that provide different functions, such as a home page loading module, a product navigation module, a shopping cart module, a payment module, a logistics management module, etc.

[0051] The target application system can also be a distributed system, such as a cloud computing system, a cloud storage system, or a big data analysis system. Figure 3 The target application system shown in the figure includes the following service modules: load balancing module a, multiple distributed application server instances b1-bm, and distributed cloud storage modules c1-ck. In actual applications, the load balancing module a can be used to receive access requests triggered by users, determine an application server instance from multiple application server instances b1-bm through a set load balancing strategy, and forward the access request to the application server instance. When processing the access request, the application server instance needs to access some data stored in the cloud storage module.

[0052] As can be seen from the above description, in order to locate the root cause of the abnormality of the target application system, two levels of agents are set up in the embodiment of the present invention: an agent at the service module level, namely, a service agent, and an agent at the system level, namely, a system agent. Through the collaborative work of these two levels of agents, the root cause information that causes the abnormality of the target application system can be determined.

[0053] In summary, taking any abnormal service module i as an example, its corresponding service agent i is used to: obtain the available analysis information of service module i, obtain the neighbor available analysis information of the neighbor service module from the neighbor service agent according to the calling relationship between multiple service modules, so as to determine the module abnormality analysis result of service module i according to the available analysis information of service module i and the neighbor available analysis information of the neighbor service module. The neighbor service module is a service module that has a calling relationship with the service module i. The available analysis information of service module i includes the abnormal operation status data of any service module, and the neighbor available analysis information includes the abnormal operation status data of the neighbor service module.

[0054] The system agent is used to obtain module exception analysis results of multiple service modules through multiple service agents, and determine the system exception analysis results of the target application system based on the module exception analysis results of multiple service modules. The system exception analysis results indicate the target service module that causes the abnormal operation of the target application system and the root cause of the exception of the target service module.

[0055] It can be seen that in the process of determining the root cause of the abnormality of the target application system, the embodiment of the present invention uses operating status data of multiple modes, which will more comprehensively reflect whether the service module has abnormalities. The comprehensive consideration of the abnormal operating status data of multiple modes and the calling relationship between service modules makes the abnormal analysis results at the service module level more accurate, and can resist the interference of inaccurate detection results of some abnormal operating status data, and has better robustness. Further comprehensive analysis of the abnormal analysis results of multiple service modules from a global perspective will help to obtain more accurate system abnormality analysis results. In addition, both the service agent and the system agent are agents based on large language models, and there is no need for manual labeling of information.

[0056] The working processes of the different components in the above-mentioned abnormality root cause location system are described in detail below.

[0057] Among them, Figure 4 As shown, the abnormal operation status data detection process of the abnormal data detection unit may include the following steps:

[0058] 401. Obtain operation status data of multiple modes corresponding to the service module i within a set time period.

[0059] 402. Determine the to-be-detected numerical value sequences corresponding to the operating status data of the multiple modes, wherein different numerical values ​​in the to-be-detected numerical value sequences correspond to different time slices within a set time period.

[0060] 403. Determine the abnormality detection thresholds corresponding to the multiple modes according to the to-be-detected numerical sequences corresponding to the multiple modes.

[0061] 404. If it is determined that an abnormal value exists in the to-be-detected numerical value sequence corresponding to the target mode according to the abnormal detection threshold corresponding to the target mode, the operating status data corresponding to the abnormal value is determined to be the abnormal operating status data of the service module i.

[0062] The target mode is any one of the multiple modes. The service module i is any service module in the target application system.

[0063] In an embodiment of the present invention, optionally, the operating status data of multiple modes include at least the following parts: resource consumption data, log data, and call chain data. Correspondingly, the numerical sequence to be detected includes at least the following parts: resource consumption index value sequence, log quantity sequence, response time sequence. In practical applications, the operating status data of multiple modes may also include operating performance indicators of service modules, such as task processing success rate, etc. The corresponding numerical sequence to be detected is composed of operating performance indicator values ​​corresponding to different times.

[0064] When detecting abnormal operation status data of a service module, the data generated by each service module in the target application system within a set time period (such as a set time period of one day, one week, or one month) to reflect the operation status of the service module can be first collected, which is called operation status data. The mode (or data type) of the operation status data can be multiple as mentioned above. Among them, the data mode corresponding to the resource consumption data is a numerical type, the data mode corresponding to the log data is a text type, the data mode corresponding to the call chain data is a graph type, and the data mode corresponding to the operation performance indicator data is a numerical type.

[0065] As for resource consumption data, it may include some host-level resource consumption indicators, such as central processing unit (CPU) utilization, memory usage, etc. These indicators are all numerical mode data.

[0066] In actual applications, different service modules in the target application system may be located in different computing nodes. The above resource consumption data of each computing node can be collected every set time slice (for example, 1 minute) within a set time period as the resource consumption data of the service module deployed therein, that is, as the resource consumption index value sequence of the service module. It can be understood that if different service modules are deployed in the same computing node, and the different service modules are independently allocated corresponding resources, then the resource consumption data of each service module can also be collected separately. For example, service module a and service module b are deployed on the same computing node, and service module a is allocated to use CPU1, and service module b is allocated to use CPU2, then the utilization of CPU1 and CPU2 can be collected every set time slice (for example, 1 minute) within a set time period to obtain the corresponding resource consumption index value sequence of service module a and service module b within the set time period.

[0067] As for log data, the target application system will generate various styles of log files during the working process, which contain log records for recording various status information such as input data, processing process, output results, etc. of the target application system. Of course, when running to a certain service module, a log data record corresponding to the service module will be generated in the log file. Thus, the abnormal data detection unit can collect the log files generated by the target application system within the set time period, and parse out the log data corresponding to each service module. Afterwards, optionally, a time slice (such as 1 minute) can be set, and the set time period can be divided according to the time slice to obtain multiple time slices. For any service module i, determine the number of logs corresponding to the service module i in each time slice, that is, the number of log data corresponding to the service module i generated in each time slice. Thus, the number of logs corresponding to each of the multiple time slices forms a sequence of the number of logs corresponding to the service module i.

[0068] In fact, a log data (i.e., a log record) often contains a fixed content part and a variable content part. For example, a log data is "the CPU utilization rate of the current target application system is 50%", and another log data is "the CPU utilization rate of the current target application system is 60%". The same fixed content part contained in the two log data is "the CPU utilization rate of the current target application system is", and "50%" and "60%" are variable text content in the log data. Based on this, in an optional embodiment, several log templates can be pre-defined, and each log template includes a fixed content part that will be included in a certain log data actually generated. In this way, after collecting several original log data corresponding to the service module i within the set time period, the log template corresponding to each log data is determined, and the number of occurrences of each log template in each defined time slice is counted as the log quantity sequence corresponding to the service module i. For service module i, assuming that only one log template L1 is involved, taking the time slice of 1 minute as an example, log template L1 appears 2 times in the 1st minute, 5 times in the second minute, 20 times in the third minute..., then, based on the number of times the log template L1 appears in the above different time slices, the log quantity sequence corresponding to the above service module i can be obtained.

[0069] As for the call chain data, the call chain data belongs to graph data, in which the nodes represent each service module, and the directed edges represent the call relationship between the service modules. In actual applications, when the target application system receives an input request, it will trigger different service modules one by one as needed to make corresponding responses, and finally feedback the output results, thus forming a call chain data. This call chain data records the service modules that are called in sequence, the call relationship between the service modules, and the response time of each service module, where the response time is the time from the service module receiving the input to giving the output. Taking service module i as an example, by collecting the call chain data related to service module i within the set time period (that is, the corresponding call chain contains service module i), the response time corresponding to service module i is parsed out, thereby forming a response time sequence corresponding to service module i. It should be noted that, optionally, at this time, the time when the service module i is called each time within the set time period can be used as different time slices within the set time period, or the set time period can be evenly divided into multiple time slices according to the size of the set time slice (for example, 1 minute), and the response time statistics of the service module i in each time slice are counted, so that the multiple response time statistics corresponding to the multiple time slices form the response time sequence corresponding to the service module i. The response time statistics can be a mean value, a maximum value, etc.

[0070] It should be noted that the above examples of the numerical value sequences to be detected are only provided for ease of understanding and are not limited thereto.

[0071] After determining the numerical sequences to be detected corresponding to the various modes, the abnormal operation status data of the multiple service modules can be detected according to the abnormal detection algorithm (such as the k-σ abnormal detection algorithm). Specifically:

[0072] When detecting the target service module according to the sequence of values ​​to be detected, determine the mean value (μ) and standard deviation (σ) of the values ​​in the sequence of values ​​to be detected. Then, set an error coefficient k (k can be 2 or 3), define the range of normal values ​​(i.e., the abnormal detection threshold) as the interval [μ- kσ, μ+ kσ]. If a value in the sequence of values ​​to be detected exceeds this range, then the value is considered to be an abnormal value, and its corresponding operating status data is the abnormal operating status data. Therefore, if an abnormal value in a sequence of values ​​to be detected corresponding to service module i is not within the above threshold range, then the operating status data corresponding to the abnormal value in each sequence to be detected constitutes a type of abnormal operating status data corresponding to service module i. For example, if the log quantity sequence and response time sequence corresponding to service module i both contain abnormal values, then service module i is ultimately detected with two modes of abnormal operating status data: log data and call chain data.

[0073] Furthermore, for the abnormal detection of log data, a neural network model can also be used for detection. For example, for service module i, service module i inputs the corresponding log data into a pre-trained abnormal data detection model within a set time period, so that it outputs the detection result of whether each log data is abnormal. Among them, the training process of the abnormal data detection model can refer to the training process of the conventional neural network model, which will not be introduced in detail here.

[0074] After the abnormal data detection unit determines multiple service modules with abnormal operation status data from all service modules included in the target application system, the abnormal operation status data corresponding to each of the multiple service modules can be sent to the service agents corresponding to each of the multiple service modules. Specifically, taking service module i as an example, each of the abnormal operation status data corresponding to the service module i can be written into a text and sent to the corresponding service agent i.

[0075] The above mainly introduces the composition of the abnormality root cause locating system. For the specific execution process of multiple service agents and system agents in the abnormality root cause locating system, please refer to the subsequent embodiments.

[0076] Based on the above-mentioned abnormal root cause positioning system composition, the abnormal root cause determination process is described in detail below.

[0077] Figure 5 A flowchart of a method for determining an abnormal root cause provided by an embodiment of the present invention is applied to a first service agent corresponding to a first service module in a target application system. The first service module is one of multiple service modules in the target application system that have abnormal operation status data. The detection process of abnormal operation status data is described above. Figure 5 As shown, the method comprises the following steps:

[0078] 501. A first service agent obtains first available analysis information of a first service module, where the first available analysis information includes abnormal operation status data of the first service module.

[0079] 502. The first service agent obtains the neighbor availability analysis information of the neighbor service module from the neighbor service agent according to the calling relationship among multiple service modules, so as to determine the module abnormality analysis result of the first service module according to the first availability analysis information of the first service module and the neighbor availability analysis information of the neighbor service module. The neighbor service module is a service module that has a calling relationship with the first service module, and the neighbor availability analysis information includes the abnormal operation status data of the neighbor service module.

[0080] 503. The first service agent sends the module anomaly analysis result of the first service module to the system agent, so that the system agent determines the system anomaly analysis result of the target application system according to the module anomaly analysis results of the corresponding service modules sent by multiple service agents.

[0081] The system abnormality analysis result indicates the target service module causing the abnormal operation of the target application system and the abnormal root cause of the target service module.

[0082] In order to determine the system anomaly analysis results of the target application system, first of all, all service modules in the target application system must be detected for anomalies. The specific anomaly detection process can be found in the relevant description above and will not be repeated here. After determining that there are multiple service modules with abnormal operating status data in the target application system, multiple service agents can be used to perform anomaly cause analysis on multiple service modules respectively, and the module anomaly analysis results obtained by the analysis are sent to the system agent, and finally the system anomaly analysis results of the target application system are obtained.

[0083] Taking the first service agent corresponding to the first service module as an example, in order to realize the abnormal analysis of the first service module, firstly, the available analysis information of the first service module is obtained, which is called the first available analysis information, wherein the first available analysis information includes the abnormal operation status data of the first service module. Of course, the first available analysis information may also include the functional description information of the first service module. Secondly, according to the calling relationship of multiple service modules, the neighbor available analysis information of the neighbor service module is obtained from the neighbor service agent of the first service agent, which includes the abnormal operation status data of the neighbor service module. Among them, the available analysis information of the neighbor service module is called the neighbor available analysis information.

[0084] It should be noted that the first service agent can obtain the calling relationship of the above-mentioned multiple service modules in advance, and determine the upstream and / or downstream service modules adjacent to the first service module (i.e., the service modules called by the first service module, or the service modules that call the first service module) as neighbor service modules based on the calling relationship. The calling relationship of these multiple service modules can optionally be determined by the abnormal operation status data detection unit based on the collected call chain data and sent to the service agents corresponding to the above-mentioned multiple service modules. Therefore, the first available analysis information can also include the calling relationship specifically presented by the first service module and the neighbor service module.

[0085] Furthermore, the first service agent determines the abnormal analysis result of the first service module based on the first available analysis information of the first service module and the neighbor available analysis information of its neighbor service module, which is called the module abnormal analysis result. The module abnormal analysis result includes the determination result of whether the first service module is abnormal and the corresponding abnormal cause. It should be understood that since there is a calling relationship between the first service module and its neighbor service module, if the first service module has abnormal operation status data, it is very likely caused by the abnormality of its neighbor service module. When determining the module abnormal analysis result of the first service module, the embodiment of the present invention comprehensively considers the abnormal operation status data of the first service module and the abnormal operation status data of its neighbor service module, thereby ensuring the accuracy of the obtained module abnormal analysis result. The specific process of determining the module abnormal analysis result of the first service module can be found in the subsequent embodiments, which will not be described in detail here.

[0086] In actual applications, the first available analysis information of the first service module and the neighbor available analysis information of its neighbor service module can be added to the prompt word template corresponding to the first service agent to generate corresponding prompt words, which are input into the large language model contained in the first agent. Under the action of the prompt words, the large language model outputs the module abnormality analysis result of the first service module. The above prompt word template will give task description information: based on the given available analysis information of the first service module and the neighbor service module, determine whether the first service module is abnormal and the cause of the abnormality. The prompt word template can also give some examples of abnormalities and abnormal causes for the large language model to learn. The first available analysis information and the neighbor available analysis information can also include the constraint information of the normal input parameters and output parameters of the corresponding service modules, such as the type and value range of the input and output parameters.

[0087] Further, after determining the module abnormality analysis result of the first service module, the first service agent can send the module abnormality analysis result of the first service module to the system agent. Optionally, the first service agent can also send the function description information of the first service module, the call relationship corresponding to the first service module, and the abnormal operation status data of the first service module to the system agent.

[0088] Since multiple service agents corresponding to multiple service modules all execute the above process, the system agent can finally obtain the module exception analysis results corresponding to the multiple service modules, and can also obtain the functional description information, corresponding call relationships and abnormal operation status data of the multiple service modules. By summarizing this information, the system agent finally outputs the system exception analysis results corresponding to the target application system, which indicates the target service module that causes the abnormal operation of the target application system and the root cause of the exception of the target service module.

[0089] It should be understood that in this process, the system agent can conduct a comprehensive analysis of the module anomaly analysis results of the corresponding service modules determined by multiple service agents to further improve the accuracy of the system anomaly analysis.

[0090] The system agent has a corresponding prompt word template, and the information received by the system agent can be added to the prompt word template to input the generated prompt words into the large language model contained in the system agent, especially output the system abnormality analysis results. In actual applications, the structure of the large language model used in the system agent and each service agent can be the same or different, and the scale of the model parameters can be the same or different.

[0091] It should be noted that if the module abnormality analysis result of the first service module determined by the first service agent is that the first service module is abnormal and the cause of the abnormality is xxx, while the system abnormality analysis result determined by the system agent is that the first service module is not abnormal, or the first service module is abnormal but the cause of the abnormality is yyy, then the system abnormality analysis result output by the system agent shall prevail.

[0092] Based on the abnormal root cause determination method provided by the above-mentioned embodiment of the present invention, by detecting the abnormal operation state data of the operation state data of various modes for each service module, the accuracy of the abnormal identification of the service module and the robustness of the abnormal root cause positioning system are improved. Because in the process of determining the module abnormal analysis result of the service module by the service agent corresponding to each service module, the calling relationship between different service modules is considered, and the multi-modal abnormal operation state data is used, and the abnormal operation state data of various modes of the service module and the neighboring service module are analyzed comprehensively, so as to ensure the accuracy of the obtained module abnormal analysis result, and resist the interference of inaccurate detection results of some abnormal operation state data. The system agent then performs a comprehensive and comprehensive analysis on the module abnormal analysis results of multiple service modules, makes full use of the calling relationship between different service modules, and can obtain more accurate system abnormal analysis results. In addition, by adopting the two-level agent configuration of the service agent and the system agent, the embodiment of the present invention does not need to perform manual labeling in the process of determining the abnormal root cause of the target application system, so the cost is low and the efficiency is high.

[0093] With respect to the process of determining the module abnormality analysis result of the service module by the service agent, in the embodiment of the present invention, two different information exchange strategies between service agents are provided to guide the information exchange process between the service agent corresponding to a service module and the service agent corresponding to its neighboring service module. The two information exchange strategies are respectively called: sequential propagation strategy and information aggregation strategy. The following describes the two strategies respectively.

[0094] Figure 6a A flowchart of a determination module abnormality analysis result provided by an embodiment of the present invention, such as Figure 6a As shown, the following steps may be included:

[0095] 601. The first service agent obtains the calling relationship between multiple service modules and the set first information propagation direction, where the first information propagation direction indicates a first direction in which multiple service agents exchange available analysis information.

[0096] 602. The first service agent obtains second available analysis information of the second service module from the second service agent based on the calling relationship between multiple service modules and the first information propagation direction. The second service module is a neighbor service module that is arranged in front of the first service module in the first information propagation direction. The first available analysis information includes functional description information of the first service module and the calling relationship with the second service module. The second available analysis information includes abnormal operation status data of the second service module and the module abnormality analysis result corresponding to the second service module in the first information propagation direction.

[0097] 603. The first service agent determines the module anomaly analysis result corresponding to the first service module in the first information propagation direction according to the first available analysis information and the second available analysis information.

[0098] This embodiment introduces an optional implementation of the sequential propagation strategy. Optionally, in the sequential propagation strategy, a first information propagation direction is defined, indicating the first direction in which multiple service agents exchange available analysis information, assuming that it is a forward information propagation direction, and is defined as: in the forward information propagation process, information is propagated from the called service module (callees) to the service module (callers) that calls them. Based on this, it can be understood that if in the call chain describing the call relationship of multiple service modules, the node represents the service module, the directed edge represents the call relationship between the service modules, and it is assumed that the directed edge is from the caller to the callee, then in the forward information propagation process, the working order of multiple service agents corresponding to multiple service modules is: start working in reverse order along the call chain, that is, the service agent corresponding to the leaf node starts working first, and the service agent corresponding to the root node starts working last. This is the meaning of "order" under the sequential propagation strategy, that is, under the sequential propagation strategy, multiple service agents are to start working in order, and the latter service agent will use the output result of the previous service agent.

[0099] In practical applications, the first information propagation direction may also be a reverse information propagation direction opposite to the forward information propagation direction, which is defined as: in the reverse information propagation process, information propagates from the calling service module (callers) to the service module (callees) it calls. Based on this, it can be understood that under the definition of the above call chain, in the reverse information propagation process, the working order of the multiple service agents corresponding to the multiple service modules is: start working in sequence along the positive order of the call chain, that is, the service agent corresponding to the root node starts working first, and the service agent corresponding to the leaf node starts working last.

[0100] This embodiment illustrates the situation where only the forward information propagation direction or only the reverse information propagation direction is used. In fact, in another optional embodiment, both the forward information propagation direction and the reverse information propagation direction may be used at the same time, which will be explained in other embodiments below.

[0101] For ease of understanding, the following Figure 6b For the example in Figure 6a The steps in the following sections are explained with specific examples. Figure 6b In the example, it is assumed that the first information propagation direction is the forward information propagation direction, and multiple service modules include: service module A, service module B and service module C, and the corresponding service agents are represented as service agent a, service agent b and service agent c respectively. The calling relationship of the three service modules is assumed to be: service module A calls service module B, and service module B calls service module C, thus forming the following Figure 6b The call chain shown in the figure: the root node is service module C, and the leaf node is service module A. Based on this, in the forward information propagation direction, Figure 6b As shown in the timeline, the working order of the service agents corresponding to these service modules is: service agent c, service agent b, service agent a. The working process of these service agents is as follows:

[0102] 1. Service agent c works first. Due to the calling relationship based on multiple service modules, service agent c determines that in the forward information propagation direction, there is no service module downstream of service module C that it can call. Therefore, in the forward information propagation direction, service agent c determines that the available analysis information I_C1 of service module C may include abnormal operation status data of service module C and functional description information of service module C. Based on the available analysis information I_C1, the service agent c outputs the module abnormal analysis result R_C1 corresponding to service module C.

[0103] 2. After that, service agent b starts working. Based on the known information of the calling relationship between multiple service modules, service agent b determines that the service module that is ranked before service module B in the forward information propagation direction is service module C. Therefore, service agent b interacts with service agent c to obtain the abnormal operation status data of service module C and the module abnormal analysis result R_C1 of service module C that has been output by service agent c from service agent c. It is thus determined that the available analysis information I_B1 of service module B in the forward information propagation direction may include the abnormal operation status data and module abnormal analysis result R_C1 of service module C, as well as the abnormal operation status data and function description information of service module B itself, and the calling relationship between service module B and service module C. Based on the above available analysis information I_B1, service agent b outputs the module abnormal analysis result R_B1 corresponding to service module B.

[0104] 3. After that, service agent a starts working. Based on the known information of the calling relationship between multiple service modules, service agent a determines that the service module that is ranked before service module A in the forward information propagation direction is service module B. Therefore, service agent a interacts with service agent b to obtain the abnormal operation status data of service module B and the module abnormal analysis result R_B1 of service module B that has been output by service agent b from service agent b. Thus, service agent a determines that the available analysis information I_A1 of service module A in the forward information propagation direction may include the abnormal operation status data and module abnormal analysis result R_B1 of service module B, the abnormal operation status data and function description information of service module A itself, and the calling relationship between service module A and service module B. Based on the above available analysis information I_A1, service agent a outputs the module abnormal analysis result R_A1 corresponding to service module A.

[0105] In this way, the information exchange process of the above three service agents in the forward information propagation direction is completed. Finally, service agent a, service agent b and service agent c send the module abnormality analysis results R_A1, R_B1 and R_C1 of their corresponding service modules in the forward information propagation direction to the system agent. In addition, the three service agents can also send the abnormal operation status data of their corresponding service modules and the calling relationship of multiple modules to the system agent. The system agent integrates the various information received and finally outputs the system abnormality analysis results of the target application system.

[0106] The above example only illustrates the forward information propagation direction, and the reverse information propagation direction is similar.

[0107] Based on the above information propagation strategy, in the process of determining the module abnormality analysis result of a service module, based on a given information propagation direction, relevant information of other adjacent service modules (including abnormal operation status data, module abnormality analysis results that have been determined under the information propagation direction, etc.) can be obtained to assist in the accurate determination of the module abnormality analysis result of the service module.

[0108] Figure 7a Another flow chart of determining module abnormality analysis results provided by an embodiment of the present invention is as follows: Figure 7a As shown, the following steps may be included:

[0109] 701. The first service agent obtains the calling relationship between multiple service modules and sets the first information propagation direction and the second information propagation direction. The first information propagation direction indicates the first direction in which multiple service agents exchange available analysis information. The second information propagation direction indicates the second direction in which multiple service agents exchange available analysis information. The first information propagation direction is executed before the second information propagation direction.

[0110] 702. The first service agent obtains second available analysis information of the second service module from the second service agent based on the calling relationship between multiple service modules and the first information propagation direction, and determines that the first available analysis information corresponding to the first service module in the first information propagation direction includes the functional description information of the first service module and the calling relationship with the second service module, the second service module is a neighbor service module that is arranged in front of the first service module in the first information propagation direction, and the second available analysis information includes the abnormal operation status data of the second service module and the module abnormality analysis result corresponding to the second service module in the first information propagation direction.

[0111] 703. The first service agent determines the module anomaly analysis result corresponding to the first service module in the first information propagation direction according to the first available analysis information and the second available analysis information.

[0112] 704. The first service agent obtains the third available analysis information of the third service module from the third service agent based on the calling relationship between the multiple service modules and the second information propagation direction, and determines that the first available analysis information corresponding to the first service module in the second information propagation direction includes the functional description information of the first service module and the calling relationship with the third service module, the third service module is a neighbor service module that is arranged in front of the first service module in the second information propagation direction, and the third available analysis information includes the abnormal operation status data of the third service module and the module abnormality analysis result corresponding to the third service module in the second information propagation direction.

[0113] Among them, if the first service module is the service module ranked first in the second information propagation direction, the first available analysis information in the second information propagation direction also includes the module abnormality analysis result corresponding to the first service module in the first information propagation direction.

[0114] 705. The first service agent determines the module anomaly analysis result corresponding to the first service module in the second information propagation direction based on the first available analysis information and the third available analysis information.

[0115] 706. The first service agent sends the module anomaly analysis result corresponding to the first service module in the second information propagation direction to the system agent.

[0116] In this embodiment, it is assumed that the above sequential propagation strategy stipulates that the first information propagation direction and the second information propagation direction need to be executed in sequence.

[0117] The following description is made by taking the first information propagation direction as the forward information propagation direction and the second information propagation direction as the reverse information propagation direction as an example.

[0118] For ease of understanding, the following Figure 7b The examples in the above steps are explained in detail. Figure 6b For example, in Figure 7b In the example, it is assumed that the multiple service modules include: service module A, service module B and service module C, and the corresponding service agents are represented as service agent a, service agent b and service agent c respectively. The calling relationship of the three service modules is assumed to be: service module A calls service module B, and service module B calls service module C, thus forming the following Figure 7b The call chain shown: the root node is service module C, and the leaf node is service module A.

[0119] Based on this, in the forward information propagation direction executed first, such as Figure 7b According to the timeline shown in , the working order of the service agents corresponding to these service modules is: service agent c, service agent b, service agent a. In the reverse information propagation direction executed later, Figure 7b In the timeline shown in , the working order of the service agents corresponding to these service modules is: service agent a, service agent b, service agent c. That is, considering the two directions comprehensively, the overall working order of multiple service agents is: service agent c, service agent b, service agent a, service agent a, service agent b, service agent c.

[0120] In the forward information transmission direction, the specific working process of service agent c, service agent b, and service agent a working in sequence refers to the above description of Figure 6b The description of the situation shown in (see steps 1, 2, and 3 above for details) is not repeated here. Here we only describe the working process of these service agents in the reverse information propagation direction executed later:

[0121] 4. After the last service agent a working in the forward information propagation direction is executed, the information transmission process in the reverse information propagation direction begins. The first service agent a determines that there is no other service module in front of service module A in the reverse information propagation direction based on the calling relationship of multiple service modules. Therefore, it determines that the available analysis information I_A2 of service module A in the reverse information propagation direction can include the module abnormality analysis result R_A1 corresponding to service module A outputted in the forward information propagation direction, as well as the abnormal operation status data and function description information of service module A itself. Based on the above available analysis information I_A2, service agent a outputs the module abnormality analysis result R_A2 corresponding to service module A.

[0122] 5. After that, service agent b starts working. Based on the known information of the calling relationship between multiple service modules, service agent b determines that the service module that is ranked before service module B in the reverse information propagation direction is service module A. Therefore, service agent b interacts with service agent a to obtain the abnormal operation status data of service module A and the module abnormal analysis result R_A1 of service module A that is most recently output by service agent a from service agent a. Thus, service agent b determines that the available analysis information I_B2 of service module B in the reverse information propagation direction may include the abnormal operation status data and module abnormal analysis result R_A1 of service module A, the abnormal operation status data and function description information of service module B itself, and the calling relationship between service module B and service module A. Based on the above available analysis information I_B2, service agent b outputs the module abnormal analysis result R_B2 corresponding to service module B.

[0123] 6. After that, the service agent c starts to work. Based on the known information of the calling relationship between multiple service modules, the service agent c determines that the service module that is ranked before the service module C in the reverse information propagation direction is the service module B. Therefore, the service agent c interacts with the service agent b to obtain the abnormal operation status data of the service module B and the module abnormal analysis result R_B2 of the service module B that is most recently output by the service agent b from the service agent b. Therefore, the service agent c determines that the available analysis information I_C2 of the service module C in the reverse information propagation direction may include the abnormal operation status data and the module abnormal analysis result R_B2 of the service module B, the abnormal operation status data and function description information of the service module C itself, and the calling relationship between the service module C and the service module B. Based on the above available analysis information I_C2, the service agent c outputs the module abnormal analysis result R_C2 corresponding to the service module C.

[0124] In this way, the information exchange process of the three service agents in the forward and reverse information propagation directions is completed. Finally, service agent a, service agent b and service agent c send the module abnormality analysis results R_A2, R_B2 and R_C2 corresponding to their respective service modules in the reverse information propagation direction executed later to the system agent. In addition, the three service agents can also send the abnormal operation status data of their respective service modules and the calling relationship of multiple modules to the system agent. The system agent integrates the various information received and finally outputs the system abnormality analysis results of the target application system.

[0125] Based on the above information propagation strategy, in the process of determining the module abnormality analysis result of a service module, based on the given two information propagation directions executed successively, the relevant information of other upstream or downstream adjacent service modules (including abnormal operation status data, module abnormality analysis results that have been determined under the current information propagation direction, etc.) can be obtained successively to assist in the accurate determination of the module abnormality analysis result of the service module.

[0126] Figure 8a Another flow chart of determining module abnormality analysis results provided by an embodiment of the present invention is as follows: Figure 8a As shown, the following steps may be included:

[0127] 801. The first service agent obtains the calling relationship between multiple service modules.

[0128] 802. The first service agent determines, based on the calling relationship among the plurality of service modules, that the neighboring service modules of the first service module include a fourth service module upstream and a fifth service module downstream.

[0129] 803. The first service agent obtains fourth available analysis information of the fourth service module from the fourth service agent, and obtains fifth available analysis information of the fifth service module from the fifth service agent, so as to determine the module abnormality analysis result of the first service module based on the first available analysis information, the fourth available analysis information and the fifth available analysis information.

[0130] This embodiment introduces an optional implementation method of the information aggregation strategy.

[0131] Under the aforementioned sequential information propagation strategy, in one propagation direction, only the relevant information of the neighboring service modules in one direction can be used, and there is a dependency relationship between the service agents: the service agent executed later needs to rely on the results output by the service agent executed before it. Therefore, the overall time consumption may be longer. When the user is sensitive to the diagnosis time of the root cause of the abnormality, this strategy is more suitable for the situation where there are fewer service modules with abnormal operating status data in the target application system.

[0132] The information aggregation strategy provided in this embodiment allows different service agents to work in parallel and can simultaneously use the relevant information of upstream and downstream adjacent service modules of a service module, which is more time-efficient overall.

[0133] Taking the first service module as an example, the first service agent can determine the upstream fourth service module and the downstream fifth service module adjacent to the first service module based on the calling relationship of multiple service modules. Thus, the fourth available analysis information of the fourth service module is obtained from the fourth service agent, and the fifth available analysis information of the fifth service module is obtained from the fifth service agent, so as to determine the module abnormality analysis result of the first service module according to the first available analysis information, the fourth available analysis information and the fifth available analysis information.

[0134] Optionally, the fourth available analysis information may include abnormal operation status data of the fourth service module, the fifth available analysis information may include abnormal operation status data of the fifth service module, and the first available analysis information may include abnormal operation status data of the first service module and the specific call relationship between the fourth and fifth service modules. At this time, it is equivalent to that the first service agent only performs an iterative process based on the abnormal operation status data of the fourth and fifth service modules adjacent to the upstream and downstream to obtain the module abnormality analysis result of the first service module.

[0135] In order to further improve the accuracy of the module anomaly analysis result of the first service module, in an embodiment of the present invention, a method for determining the module anomaly analysis result of the first service module through multiple iterations is also provided. At this time, the module anomaly analysis result of the first service module determined in the last iteration may be output to the system agent. The number of iterations may be set in advance according to the requirements. At this time, the process of determining the module anomaly analysis result of the first service module may be:

[0136] In the process of determining the module abnormality analysis result of the first service module in the last iteration process through the set multiple iteration processes, the fourth available analysis information of the fourth service module provided by the fourth agent, the fifth available analysis information of the fifth service module provided by the fifth agent, and the first available analysis information of the first service module in the current iteration process are obtained, so as to determine the module abnormality analysis result of the first service module in the current iteration process according to the first available analysis information, the fourth available analysis information, and the fifth available analysis information obtained in the current iteration process. Wherein, in the first iteration process, the first available analysis information includes the abnormal operation status data of the first service module, the functional description information of the first service module, and the calling relationship with the fourth service module and the fifth service module, the fourth available analysis information includes the abnormal operation status data of the fourth service module, and the fifth available analysis information includes the abnormal operation status data of the fifth service module. Wherein, in the other iteration processes except the first iteration process, the first available analysis information also includes the module abnormality analysis result of the first service module determined in the previous iteration process, the fourth available analysis information also includes the module abnormality analysis result of the fourth service module determined in the previous iteration process, and the fifth available analysis information also includes the module abnormality analysis result of the fifth service module determined in the previous iteration process.

[0137] For ease of understanding, the following Figure 8b The example in the above steps is used to illustrate the above steps. Figure 8b In the example, it is assumed that the multiple service modules include: service module A, service module B and service module C, and the corresponding service agents are represented as service agent a, service agent b and service agent c respectively. The calling relationship of the three service modules is assumed to be: service module A calls service module B, and service module B calls service module C, thus forming the following Figure 8b The call chain shown: the root node is service module C, and the leaf node is service module A.

[0138] For service module C, it only has a downstream adjacent service module B; for service module B, it includes an upstream adjacent service module C and a downstream adjacent service module A; for service module A, it includes an upstream adjacent service module B.

[0139] Based on this, in the first iteration, service agent c obtains the available analysis information X_B1 of service module B from service agent b, and determines the available analysis information X_C1 of service module C. The available analysis information X_C1 includes the abnormal operation status data of service module C, the functional description information of service module C, and the calling relationship between service module C and service module B. The available analysis information X_B1 includes the abnormal operation status data of service module B. Based on the above available analysis information X_B1 and available analysis information X_C1, service agent c outputs the module abnormal analysis result Y_C1 corresponding to service module C.

[0140] During the first iteration, service agent b obtains the available analysis information X_A2 of service module A and the available analysis information X_C2 of service module C from service agent a and service agent c, respectively, and determines the available analysis information X_B2 of service module B. The available analysis information X_B2 includes the abnormal operation status data of service module B, the functional description information of service module B, and the calling relationship between service module B, service module A, and service module C. The available analysis information X_A2 includes the abnormal operation status data of service module A, and the available analysis information X_C2 includes the abnormal operation status data of service module C. Based on the above available analysis information X_A2, available analysis information X_B2, and available analysis information X_C2, service agent b outputs the module abnormal analysis result Y_B1 corresponding to service module B.

[0141] During the first iteration, service agent a obtains the available analysis information X_B3 of service module B from service agent b, and determines the available analysis information X_A3 of service module A. The available analysis information X_A3 includes the abnormal operation status data of service module A, the functional description information of service module A, and the calling relationship between service module A and service module B. The available analysis information X_B3 includes the abnormal operation status data of service module B. Based on the above available analysis information X_A3 and available analysis information X_B3, service agent a outputs the module abnormal analysis result Y_A1 corresponding to service module A.

[0142] The first iteration process of the above three service agents is carried out in parallel, and the first iteration process of the three service agents is now completed.

[0143] During the second iteration, service agent c obtains the available analysis information X_B4 of service module B from service agent b, and determines the available analysis information X_C4 of service module C. The available analysis information X_C4 includes the abnormal operation status data of service module C, the functional description information of service module C, and the calling relationship between service module C and service module B, and also includes the module abnormal analysis result Y_C1 of service module C determined in the previous iteration. The available analysis information X_B4 includes the abnormal operation status data of service module B and the module abnormal analysis result Y_B1 of service module B determined in the previous iteration. Based on the above available analysis information X_B4 and available analysis information X_C4, service agent c outputs the module abnormal analysis result Y_C2 corresponding to service module C.

[0144] During the second iteration, service agent b obtains the available analysis information X_A5 of service module A and the available analysis information X_C5 of service module C from service agent a and service agent c, respectively, and determines the available analysis information X_B5 of service module B. The available analysis information X_B5 includes the abnormal operation status data of service module B, the functional description information of service module B, and the calling relationship between service module B, service module A, and service module C, and also includes the module abnormal analysis result Y_B1 of service module B determined in the previous iteration. The available analysis information X_A5 includes the abnormal operation status data of service module A and the module abnormal analysis result Y_A1 of service module A determined in the previous iteration, and the available analysis information X_C5 includes the abnormal operation status data of service module C and the module abnormal analysis result Y_C1 of service module C determined in the previous iteration. Service agent b outputs the module abnormal analysis result Y_B2 corresponding to service module B based on the above available analysis information X_A5, available analysis information X_B5, and available analysis information X_C5.

[0145] During the second iteration, service agent a obtains the available analysis information X_B6 of service module B from service agent b, and determines the available analysis information X_A6 of service module A. The available analysis information X_A6 includes the abnormal operation status data of service module A, the functional description information of service module A, and the calling relationship between service module A and service module B, and also includes the module abnormal analysis result Y_A1 of service module A determined during the previous iteration. The available analysis information X_B6 includes the abnormal operation status data of service module B and the module abnormal analysis result Y_B1 of service module B determined during the previous iteration. Based on the above available analysis information X_A6 and available analysis information X_B6, service agent a outputs the module abnormal analysis result Y_A2 corresponding to service module A.

[0146] The second iteration process of the above three service agents is carried out in parallel, and the second iteration process of the three service agents is completed. The iteration process of other times is similar, and will not be repeated here.

[0147] Assuming that the total number of iterations = 2, based on the above example, service agent a, service agent b and service agent c send the module anomaly analysis results Y_A2, Y_B2, Y_C2 determined by their respective service modules during the last iteration to the system agent.

[0148] The above respectively introduces the execution process of the sequential propagation strategy and the information aggregation strategy. In practical applications, these two strategies can be used selectively or simultaneously. When these two strategies are used simultaneously, each service agent can send the module anomaly analysis result of the corresponding service module determined under each strategy to the system agent, so that the system agent determines the system anomaly analysis result according to the module anomaly analysis results corresponding to the two strategies uploaded by multiple service agents.

[0149] The following will describe in detail the abnormal root cause determination device of one or more embodiments of the present invention. Those skilled in the art will appreciate that these devices can be configured using commercially available hardware components through the steps taught in this solution.

[0150] Fig. 9 A schematic diagram of the structure of an abnormal root cause determination device provided by an embodiment of the present invention, the abnormal root cause determination device is applied to a first service agent corresponding to a first service module in a target application system, wherein the target application system includes multiple service modules, each service module corresponds to a service agent, and the first service module is one of the multiple service modules in the target application system in which abnormal operation status data is detected. Fig. 9 As shown, the device includes: an acquisition module 11, an analysis module 12 and a determination module 13.

[0151] The acquisition module 11 is used to acquire first available analysis information of the first service module, where the first available analysis information includes abnormal operation status data of the first service module.

[0152] The analysis module 12 is used to obtain the neighbor available analysis information of the neighbor service module from the neighbor service agent according to the calling relationship between the multiple service modules, so as to determine the module abnormality analysis result of the first service module according to the first available analysis information of the first service module and the neighbor available analysis information of the neighbor service module. The neighbor service module is a service module that has a calling relationship with the first service module, and the neighbor available analysis information includes abnormal operation status data of the neighbor service module.

[0153] Determination module 13 is used to send the module exception analysis result of the first service module to the system agent, so that the system agent determines the system exception analysis result of the target application system according to the module exception analysis results of the corresponding service modules sent by each of the multiple service agents, and the system exception analysis result indicates the target service module that causes the abnormal operation of the target application system and the root cause of the exception of the target service module.

[0154] Among them, optionally, the analysis module 12 is specifically used to: obtain the calling relationship between the multiple service modules and the set first information propagation direction, the first information propagation direction indicating the first direction in which the multiple service agents exchange available analysis information; obtain the second available analysis information of the second service module from the second service agent according to the calling relationship and the first information propagation direction, determine that the first available analysis information corresponding to the first service module in the first information propagation direction includes the functional description information of the first service module and the calling relationship with the second service module, the second service module is a neighbor service module that is arranged in front of the first service module in the first information propagation direction, the second available analysis information includes the abnormal operation status data of the second service module and the module abnormality analysis result corresponding to the second service module in the first information propagation direction; determine the module abnormality analysis result corresponding to the first service module in the first information propagation direction according to the first available analysis information and the second available analysis information, the first available analysis information includes the functional description information of the first service module and the calling relationship with the second service module.

[0155] Optionally, the analysis module 12 can also be used to: obtain third available analysis information of a third service module from a third service agent according to the calling relationship and the set second information propagation direction, the third service module being a neighbor service module that is ranked ahead of the first service module in the second information propagation direction, the third available analysis information including abnormal operation status data of the third service module and module abnormality analysis results corresponding to the third service module in the second information propagation direction, the second information propagation direction indicating the second direction in which the multiple service agents exchange available analysis information, the first information propagation direction being executed before the second information propagation direction; determine that the first available analysis information corresponding to the first service module in the second information propagation direction includes function description information of the first service module and a calling relationship with the third service module; wherein, if the first service module is the service module ranked first in the second information propagation direction, the first available analysis information in the second information propagation direction also includes module abnormality analysis results corresponding to the first service module in the first information propagation direction; determine the module abnormality analysis results corresponding to the first service module in the second information propagation direction according to the first available analysis information and the third available analysis information.

[0156] Wherein, optionally, the determination module 13 is specifically used to: send the module abnormality analysis result corresponding to the first service module in the second information propagation direction to the system agent.

[0157] Among them, optionally, the analysis module 12 is also used to: obtain the calling relationship among the multiple service modules; determine that the neighbor service module includes an upstream fourth service module and a downstream fifth service module based on the calling relationship; obtain the fourth available analysis information of the fourth service module from the fourth service agent, and obtain the fifth available analysis information of the fifth service module from the fifth service agent, so as to determine the module abnormality analysis result of the first service module based on the first available analysis information, the fourth available analysis information and the fifth available analysis information.

[0158] Wherein, optionally, the analysis module 12 is specifically used to: in the process of determining the module abnormality analysis result of the first service module in the last iteration process through the set multiple iteration processes, obtain the fourth available analysis information of the fourth service module provided by the fourth agent and the fifth available analysis information of the fifth service module provided by the fifth agent and the first available analysis information of the first service module in the current iteration process, so as to determine the module abnormality analysis result of the first service module in the current iteration process according to the first available analysis information, the fourth available analysis information and the fifth available analysis information obtained in the current iteration process. Wherein, in the first iteration process, the first available analysis information includes the abnormal operation status data of the first service module, the functional description information of the first service module and the calling relationship with the fourth service module and the fifth service module, the fourth available analysis information includes the abnormal operation status data of the fourth service module, and the fifth available analysis information includes the abnormal operation status data of the fifth service module. Among them, except for the initial iteration process, in other iteration processes, the first available analysis information also includes the module abnormality analysis result of the first service module determined in the previous iteration process, the fourth available analysis information also includes the module abnormality analysis result of the fourth service module determined in the previous iteration process, and the fifth available analysis information also includes the module abnormality analysis result of the fifth service module determined in the previous iteration process.

[0159] Optionally, the determination module 13 is specifically used to send the module abnormality analysis result determined by the first service module in the last iteration process to the system agent.

[0160] In addition, the determination module 13 is also used to send the module abnormality analysis result of the first service module, the functional description information of the first service module, the call relationship corresponding to the first service module and the abnormal operation status data of the first service module to the system agent.

[0161] Fig. 9 The device shown can execute the steps provided in the aforementioned embodiments. For detailed execution process and technical effects, please refer to the description in the aforementioned embodiments, which will not be repeated here.

[0162] In one possible design, an embodiment of the present invention provides a non-temporary machine-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor can at least implement the method for determining the root cause of an abnormality as provided in the aforementioned embodiment.

[0163] In addition, an embodiment of the present invention provides a computer program product, which includes: a computer program, when the computer program is executed by a processor of an electronic device, the processor can at least implement the abnormality root cause determination method provided in the above embodiment.

[0164] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. Those of ordinary skill in the art may understand and implement the present invention without creative effort.

[0165] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by adding a necessary general hardware platform, and of course can also be implemented by combining hardware and software. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a computer product, and the present invention can be implemented in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0166] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining an abnormal root cause, characterized in that: A method for applying to a first service agent corresponding to a first service module in a target application system, wherein the target application system includes a plurality of service modules, each service module corresponds to a service agent, and the first service module is one of a plurality of service modules in which abnormal operation status data is detected among the plurality of service modules, and the method comprises: Acquire first available analysis information of the first service module, where the first available analysis information includes abnormal operation status data of the first service module; Obtaining neighbor availability analysis information of a neighbor service module from a neighbor service agent according to the calling relationship of the multiple service modules, so as to determine a module abnormality analysis result of the first service module according to the first availability analysis information of the first service module and the neighbor availability analysis information of the neighbor service module, wherein the neighbor service module is a service module having a calling relationship with the first service module, and the neighbor availability analysis information includes abnormal operation status data of the neighbor service module; Sending the module abnormality analysis result of the first service module to the system agent, so that the system agent determines the system abnormality analysis result of the target application system according to the module abnormality analysis results of the corresponding service modules sent by the plurality of service agents, wherein the system abnormality analysis result indicates the target service module causing the abnormal operation of the target application system and the abnormal root cause of the target service module; The detection process of abnormal operation status data of the first service module includes the following steps: obtaining operation status data of multiple modes corresponding to the first service module within a set time period; determining a sequence of values ​​to be detected corresponding to each of the operation status data of the multiple modes, and different values ​​in the sequence of values ​​to be detected correspond to different time slices within the set time period; determining an abnormal detection threshold corresponding to each of the multiple modes according to the sequence of values ​​to be detected corresponding to each of the multiple modes; if it is determined that there is an abnormal value in the sequence of values ​​to be detected corresponding to the target mode according to the abnormal detection threshold corresponding to the target mode, then determining that the operation status data corresponding to the abnormal value is the abnormal operation status data of the first service module.

2. The method according to claim 1, characterized in that The acquiring, from a neighbor service agent, neighbor availability analysis information of a neighbor service module according to the calling relationship of the plurality of service modules, so as to determine a module abnormality analysis result of the first service module according to the first availability analysis information of the first service module and the neighbor availability analysis information of the neighbor service module, comprises: Acquire the calling relationship of the plurality of service modules and the set first information propagation direction, wherein the first information propagation direction indicates a first direction in which the plurality of service agents exchange available analysis information; According to the calling relationship and the first information propagation direction, second available analysis information of a second service module is obtained from a second service agent, where the second service module is a neighbor service module that is arranged in front of the first service module in the first information propagation direction, and the second available analysis information includes abnormal operation status data of the second service module and a module abnormality analysis result corresponding to the second service module in the first information propagation direction; Determine that the first available analysis information corresponding to the first service module in the first information propagation direction includes function description information of the first service module and a calling relationship with the second service module; A module abnormality analysis result corresponding to the first service module in the first information propagation direction is determined according to the first available analysis information and the second available analysis information.

3. The method according to claim 2, characterized in that The acquiring, from a neighbor service agent, neighbor availability analysis information of a neighbor service module according to the calling relationship of the plurality of service modules, so as to determine a module abnormality analysis result of the first service module according to the first availability analysis information of the first service module and the neighbor availability analysis information of the neighbor service module, comprises: According to the calling relationship and the set second information propagation direction, third available analysis information of a third service module is obtained from a third service agent, the third service module is a neighbor service module arranged before the first service module in the second information propagation direction, the third available analysis information includes abnormal operation status data of the third service module and a module abnormality analysis result corresponding to the third service module in the second information propagation direction, the second information propagation direction indicates a second direction in which the multiple service agents exchange available analysis information, and the first information propagation direction is executed before the second information propagation direction; Determine that the first available analysis information corresponding to the first service module in the second information propagation direction includes the function description information of the first service module and the calling relationship with the third service module; wherein, if the first service module is the first service module in the second information propagation direction, then the first available analysis information in the second information propagation direction also includes the module abnormality analysis result corresponding to the first service module in the first information propagation direction; A module abnormality analysis result corresponding to the first service module in the second information propagation direction is determined according to the first available analysis information and the third available analysis information.

4. The method according to claim 3, characterized in that The step of sending the module abnormality analysis result of the first service module to the system agent comprises: The module abnormality analysis result corresponding to the first service module in the second information propagation direction is sent to the system agent.

5. The method according to any one of claims 1 to 4, characterized in that The acquiring, from a neighbor service agent, neighbor availability analysis information of a neighbor service module according to the calling relationship of the plurality of service modules, so as to determine a module abnormality analysis result of the first service module according to the first availability analysis information of the first service module and the neighbor availability analysis information of the neighbor service module, comprises: Acquire the calling relationship of the multiple service modules; Determining, according to the calling relationship, that the neighbor service module includes an upstream fourth service module and a downstream fifth service module; The fourth available analysis information of the fourth service module is obtained from the fourth service agent, and the fifth available analysis information of the fifth service module is obtained from the fifth service agent, so as to determine the module abnormality analysis result of the first service module according to the first available analysis information, the fourth available analysis information and the fifth available analysis information.

6. The method according to claim 5, characterized in that The step of acquiring fourth available analysis information of the fourth service module from the fourth service agent and acquiring fifth available analysis information of the fifth service module from the fifth service agent, so as to determine a module abnormality analysis result of the first service module according to the first available analysis information, the fourth available analysis information and the fifth available analysis information, comprises: In the process of determining the module abnormality analysis result of the first service module in the last iteration process through the set multiple iteration processes, the fourth available analysis information of the fourth service module provided by the fourth service agent in the current iteration process, the fifth available analysis information of the fifth service module provided by the fifth service agent, and the first available analysis information of the first service module are obtained, so as to determine the module abnormality analysis result of the first service module in the current iteration process according to the first available analysis information, the fourth available analysis information and the fifth available analysis information obtained in the current iteration process; Wherein, during the first iteration, the first available analysis information includes the abnormal operation status data of the first service module, the functional description information of the first service module, and the calling relationship with the fourth service module and the fifth service module, the fourth available analysis information includes the abnormal operation status data of the fourth service module, and the fifth available analysis information includes the abnormal operation status data of the fifth service module; Among them, except for the initial iteration process, in other iteration processes, the first available analysis information also includes the module abnormality analysis result of the first service module determined in the previous iteration process, the fourth available analysis information also includes the module abnormality analysis result of the fourth service module determined in the previous iteration process, and the fifth available analysis information also includes the module abnormality analysis result of the fifth service module determined in the previous iteration process.

7. The method according to claim 6, characterized in that The step of sending the module abnormality analysis result of the first service module to the system agent comprises: The module anomaly analysis result determined by the first service module in the last iteration process is sent to the system agent.

8. The method according to claim 1, characterized in that: The step of sending the module abnormality analysis result of the first service module to the system agent comprises: The module abnormality analysis result of the first service module, the functional description information of the first service module, the call relationship corresponding to the first service module and the abnormal operation status data of the first service module are sent to the system agent.

9. An abnormality root cause location system, characterized in that: include: The abnormal data detection unit is used to detect the running status data of each service module in the target application system to determine the abnormal running status data of each of the multiple service modules, wherein the detection process of the abnormal running status data of each of the multiple service modules includes the following steps: obtaining the running status data of multiple modes corresponding to the target service module among the multiple service modules within a set time period; determining the to-be-detected numerical sequence corresponding to each of the running status data of the multiple modes, wherein different numerical values ​​in the to-be-detected numerical sequence correspond to different time slices within the set time period; determining the abnormal detection thresholds corresponding to each of the multiple modes according to the to-be-detected numerical sequence corresponding to each of the multiple modes; if it is determined according to the abnormal detection threshold corresponding to the target mode that there is an abnormal numerical value in the to-be-detected numerical sequence corresponding to the target mode, then determining that the running status data corresponding to the abnormal numerical value is the abnormal running status data of the target service module; A plurality of service agents corresponding to the plurality of service modules, wherein the service agent corresponding to any service module is used to: obtain available analysis information of any service module, obtain neighbor available analysis information of a neighbor service module from a neighbor service agent according to the calling relationship of the plurality of service modules, so as to determine a module abnormality analysis result of any service module according to the available analysis information of any service module and the neighbor available analysis information of the neighbor service module, wherein the neighbor service module is a service module having a calling relationship with any service module, the available analysis information of any service module includes abnormal operation status data of any service module, and the neighbor available analysis information includes abnormal operation status data of the neighbor service module; A system agent is used to obtain module exception analysis results of the multiple service modules through the multiple service agents, and determine the system exception analysis results of the target application system according to the module exception analysis results of the multiple service modules, wherein the system exception analysis results indicate the target service module that causes the abnormal operation of the target application system and the root cause of the exception of the target service module.

10. The system according to claim 9, characterized in that The abnormal data detection unit is specifically used for: Acquire the operation status data of multiple modes corresponding to any one of the service modules within a set time period, wherein the operation status data of multiple modes includes at least the following parts: resource consumption data, log data, and call chain data; Determine the to-be-detected numerical value sequences corresponding to the operation status data of the multiple modes, wherein different numerical values ​​in the to-be-detected numerical value sequences correspond to different time slices within the set time period, and the to-be-detected numerical value sequences include at least the following parts: a resource consumption index value sequence, a log quantity sequence, and a response time sequence; Determining the abnormality detection thresholds corresponding to the multiple modes respectively according to the to-be-detected numerical value sequences corresponding to the multiple modes respectively; If, according to the abnormal detection threshold corresponding to the target modality, it is determined that there is an abnormal value in the sequence of values ​​to be detected corresponding to the target modality, then the operating status data corresponding to the abnormal value is determined to be the abnormal operating status data of any of the service modules, and the target modality is any one of the multiple modalities.

11. A non-transitory machine-readable storage medium, characterized in that: The non-transitory machine-readable storage medium stores executable code, and when the executable code is executed by a processor of an electronic device, the processor is caused to perform the method as claimed in any one of claims 1 to 8.

12. A computer program product, characterized in that include: A computer program, when executed by a processor of an electronic device, causes the processor to execute the method according to any one of claims 1 to 8.