Fault diagnosis method, device, equipment and medium for data processing link

Abstract

The disclosed embodiment provides a fault diagnosis method, device, equipment and medium for a data processing link. After obtaining fault alarm information, the fault information matrix corresponding to each data processing node in the data processing link is obtained. The fault information matrix includes the combined state characteristics corresponding to the data processing node and the thought chain prompt information. The combined state characteristics represent at least two types of operating states of the data processing node. The thought chain prompt information represents the causal relationship between the data processing nodes with upstream and downstream relationships that trigger the fault. The fault information matrix corresponding to each data processing node is sequentially input into the fault diagnosis model in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents the fault handling suggestions for the target fault. The accuracy of fault diagnosis and the fault coverage are improved.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of artificial intelligence technology, and in particular to a fault diagnosis method, device, equipment and medium for a data processing link. Background Art

[0002] Cross-end data synchronization usually refers to synchronizing data from the source end to the target end through multiple data components. The data synchronization process often involves multiple data processing steps, such as data forwarding, compression, detection, and format conversion. Therefore, failures are very likely to occur during the data synchronization process, such as timeliness and accuracy issues. How to effectively troubleshoot is a technical issue of great concern in data synchronization scenarios.

[0003] At present, the commonly used fault location and troubleshooting solutions are: one is to rely on hundreds of monitoring indicators and perform fault diagnosis for each indicator status. The diagnosis timeliness is relatively low, and single-point indicator troubleshooting is also prone to generate alarm noise; the other is to simplify the above thousands of indicators. Compared with the previous solution, this solution has improved timeliness, but its alarm accuracy is often poor. Therefore, it is necessary to provide a troubleshooting technology solution with high timeliness and accuracy for data synchronization scenarios. Summary of the invention

[0004] The embodiments of the present disclosure provide a fault diagnosis method, apparatus, device and medium for a data processing link to overcome the problems of poor accuracy of fault diagnosis results and small fault coverage.

[0005] In a first aspect, an embodiment of the present disclosure provides a fault diagnosis method for a data processing link, comprising:

[0006] After obtaining the fault alarm information, obtain the fault information matrix corresponding to each data processing node in the data processing link, wherein the fault information matrix includes the combined state characteristics and thought chain prompt information corresponding to the data processing node, wherein the combined state characteristics represent the operating state of the data processing node in at least two dimensions, and the thought chain prompt information represents the causal relationship of triggering faults between data processing nodes with upstream and downstream relationships; input the fault information matrix corresponding to each of the data processing nodes into the fault diagnosis model in turn in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents the fault handling suggestions for the target fault.

[0007] In a second aspect, an embodiment of the present disclosure provides a fault diagnosis device for a data processing link, including:

[0008] An acquisition module is used to acquire a fault information matrix corresponding to each data processing node in the data processing link after obtaining the fault alarm information, wherein the fault information matrix includes a combination state feature and a thought chain prompt information corresponding to the data processing node, wherein the combination state feature represents the operation state of the data processing node in at least two dimensions, and the thought chain prompt information represents the causal relationship of triggering the fault between the data processing nodes having an upstream and downstream relationship;

[0009] A processing module is used to input the fault information matrix corresponding to each of the data processing nodes into the fault diagnosis model in turn in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents the fault handling suggestions for the target fault.

[0010] In a third aspect, an embodiment of the present disclosure provides an electronic device, including: a processor and a memory;

[0011] The memory stores computer-executable instructions;

[0012] The processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the fault diagnosis method for the data processing link as described in the first aspect and various possible designs of the first aspect.

[0013] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, in which computer execution instructions are stored. When a processor executes the computer execution instructions, the fault diagnosis method for the data processing link as described in the first aspect and various possible designs of the first aspect is implemented.

[0014] In a fifth aspect, an embodiment of the present disclosure provides a computer program product, including a computer program, which, when executed by a processor, implements the fault diagnosis method for a data processing link as described in the first aspect and various possible designs of the first aspect.

[0015] The fault diagnosis method, device, equipment and medium for the data processing link provided in this embodiment, after obtaining the fault alarm information, obtain the fault information matrix corresponding to each data processing node in the data processing link, the fault information matrix includes the combined state characteristics and the thinking chain prompt information corresponding to the data processing node, the combined state characteristics represent the operation state of the data processing node in at least two dimensions, and the thinking chain prompt information represents the causal relationship of triggering faults between data processing nodes with upstream and downstream relationships; the fault information matrix corresponding to each of the data processing nodes is sequentially input into the fault diagnosis model in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents the fault handling suggestions for the target fault. By inputting the fault information matrix composed of the combined state characteristics and the thinking chain prompt information into the fault diagnosis model in multiple rounds, the ability of the fault diagnosis model is fully utilized, the diagnosis of each data processing node in the data processing link is realized in a multi-round interactive manner, the corresponding fault handling suggestions are generated, and the accuracy of fault diagnosis and the fault coverage are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0017] Figure 1 An application scenario diagram of a fault diagnosis method for a data processing link provided in an embodiment of the present disclosure;

[0018] Figure 2 A schematic diagram of a method for diagnosing a fault in a data processing link according to an embodiment of the present disclosure Figure 1 ;

[0019] Figure 3 A schematic diagram of the structure of a thought chain prompt information provided by an embodiment of the present disclosure;

[0020] Figure 4 A schematic diagram of a method for diagnosing a fault in a data processing link according to an embodiment of the present disclosure Figure 2 ;

[0021] Figure 5 for Figure 4 A flowchart of a specific implementation method of step S203 in the illustrated embodiment;

[0022] Figure 6 for Figure 4 A flowchart of a specific implementation method of step S204 in the illustrated embodiment;

[0023] Figure 7 A schematic diagram of a process for generating a fault information matrix provided by an embodiment of the present disclosure;

[0024] Figure 8 for Figure 4 A flowchart of a specific implementation method of step S207 in the illustrated embodiment;

[0025] Fig. 9 A schematic diagram of a process for generating fault diagnosis information provided by an embodiment of the present disclosure;

[0026] Fig.10 A structural block diagram of a fault diagnosis device for a data processing link provided in an embodiment of the present disclosure;

[0027] Fig.11 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure;

[0028] Fig.12 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0030] 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 this disclosure 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.

[0031] The application scenarios of the embodiments of the present disclosure are explained below:

[0032] Figure 1An application scenario diagram of the fault diagnosis method for a data processing link provided in an embodiment of the present disclosure. The fault diagnosis method for a data processing link provided in an embodiment of the present disclosure can be applied to application scenarios of cross-national and cross-regional data synchronization. More specifically, it can be applied to application scenarios of fault monitoring and diagnosis during cross-regional data synchronization based on a data synchronization system. The execution subject of this embodiment can be a fault monitoring server deployed in the above-mentioned data synchronization system, a terminal device that communicates with the fault monitoring server, or other electronic devices that perform similar functions.

[0033] Among them, in some embodiments, the above-mentioned fault monitoring server or electronic device can implement the fault diagnosis method for the data processing link provided by the embodiment of the present application by running various computer executable instructions or computer programs. For example, the computer executable instructions can be program-level commands, machine instructions or software instructions. The computer program can be a native program or software module in the operating system; it can be a local application, that is, a program that needs to be installed in the operating system to run, or it can be a program that runs based on a browser environment. In summary, the above-mentioned computer executable instructions can be instructions in any form, and the above-mentioned computer program can be an application, module or plug-in in any form, and the specific implementation form can be configured as needed. Further, in some embodiments, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud services, cloud storage, cloud communication, cloud database, cloud computing, cloud functions, network services, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and basic cloud computing services such as big data and artificial intelligence platforms, wherein the cloud service can be an interactive processing service for terminal devices to call.

[0034] refer to Figure 1As shown in , in the application scenario of cross-regional data synchronization, data is synchronized through a distributed data synchronization system. The data synchronization system includes multiple data processing nodes, such as the P1 node, P2 node, P3 node, etc. in the figure. The above-mentioned data processing nodes are, for example, data storage components, cache components, and forwarding components. According to specific needs, each data processing node in the data synchronization system can realize corresponding functions and tasks. After the data is transmitted from the A end (source end) through the above-mentioned P1 node, P2 node, P3 node and other data processing nodes, it finally reaches the B end (target end) to achieve data synchronization. At the same time, in the above-mentioned data synchronization system, a fault monitoring server is also provided. The fault monitoring server collects the operating data of the A end, the B end, and each data processing node to perform fault monitoring and alarm on the above-mentioned data synchronization system, and sends the generated alarm information to the terminal device communicating with it, thereby achieving the purpose of fault monitoring for the data synchronization system and the data synchronization process.

[0035] In the prior art, in order to achieve the normal execution of the above-mentioned data synchronization process, the fault detection of the data synchronization system and the data synchronization process is usually carried out by obtaining the operating status indicators of each data processing node and comparing it with the threshold of the preset value to determine whether it has a fault and realize fault judgment. However, the method scheme in the prior art can only judge the data processing node that has a fault, but cannot diagnose the cause of the fault, that is, it is impossible to confirm the upstream data processing node that actually causes the fault, so that the operation and maintenance personnel cannot accurately repair the data synchronization system according to the fault data. At the same time, due to the large number of data processing nodes, each data processing node corresponds to more than one operating status indicator. Monitoring and processing the above data in full will lead to huge resource overhead and require the operation and maintenance personnel to have very high requirements for the understanding of the entire system. Reducing the amount of collected operating status indicators will lead to the problem of not being able to detect faults in a timely and accurate manner. Therefore, it leads to problems such as poor accuracy of fault diagnosis results and small fault coverage.

[0036] The embodiment of the present disclosure provides a fault diagnosis method for a data processing link to solve the above-mentioned problem.

[0037] refer to Figure 2 , Figure 2 A schematic diagram of a method for diagnosing a fault in a data processing link according to an embodiment of the present disclosure Figure 1 The method of this embodiment can be applied in a fault monitoring server. The fault diagnosis method for a data processing link includes:

[0038] Step S101: After obtaining the fault alarm information, obtain the fault information matrix corresponding to each data processing node in the data processing link. The fault information matrix includes the combined state characteristics corresponding to the data processing nodes and the thinking chain prompt information. The combined state characteristics represent the operating state of the data processing nodes in at least two dimensions, and the thinking chain prompt information represents the causal relationship between the data processing nodes with upstream and downstream relationships that trigger the fault.

[0039] refer to Figure 1 The schematic diagram of the application scenario shown in the figure, in this embodiment, the fault monitoring server (hereinafter referred to as the server) is used as the execution subject to introduce the fault diagnosis method for the data processing link provided. For example, a functional unit for monitoring the data processing node will be provided in the data synchronization system, and the functional unit can be implemented by the server, that is, the above-mentioned server collects data from each data processing node, and determines whether a data processing node fails based on the collected monitoring data. When it is determined that a data processing node fails, a fault alarm message is generated; of course, in another possible implementation method, the functional unit can be an independent component set outside the above-mentioned server, which collects data from each data processing node and determines whether there is a data processing node that fails. When it detects a data processing node with a fault, a fault alarm message is generated and sent to the server. The specific generation method of the above-mentioned fault alarm message can be set as needed without limitation.

[0040] Afterwards, the server responds to the fault alarm information and further obtains the fault information matrix corresponding to each data processing node in the data processing link, wherein the fault information matrix includes the combination state characteristics corresponding to the data processing node and the thought chain prompt information. The combination state characteristics corresponding to the data processing node and the thought chain prompt information corresponding to the data processing node are further introduced in detail below.

[0041] Specifically, on the one hand, the combined state feature characterizes the operating state of the data processing node in at least two dimensions, that is, the combination of the operating states of multiple discrete time points. Specifically, for example, the combined state feature may include the "CPU utilization" (first dimension) and "read and write speed" (second dimension) of the data processing node, that is, the operating state in two dimensions. The combined state feature can be obtained by the server by accessing the data processing node. Based on the previous introduction, based on the specific implementation functions, the data processing node can be a data storage component, a cache component, and a forwarding component in the data synchronization system. As an independently running component, the data processing node can generate a variety of data indicators that characterize the operating state during operation. The server obtains a variety of data indicators corresponding to each data processing node by accessing each data processing node, thereby obtaining the combined state feature.

[0042] On the other hand, the thought chain prompt information characterizes the causal relationship of triggering faults between data processing nodes with upstream and downstream relationships. Specifically, the content in the thought chain prompt information is used to characterize how the data processing node located upstream affects the downstream data processing node to cause it to fail. For example, the content in the thought chain prompt information Info_1 of data processing node C includes: "The increase in the number of transactions per second indicator of the upstream data processing node A will cause the outflow throughput indicator of data processing node C to increase." Therefore, through the content in the thought chain prompt information, it is possible to infer the fault characteristics of the downstream data processing node C based on the fault characteristics of the upstream data processing node A, thereby realizing fault reasoning between data processing nodes with upstream and downstream relationships.

[0043] Furthermore, in a possible implementation, the thinking chain prompt information includes causal prompt information, functional prompt information and risk prompt information. The causal prompt information is used to characterize the rule that the upstream data processing node of the current data processing node triggers the second fault feature corresponding to the current data processing node based on the first fault feature, and / or the current data processing node triggers the rule that the upstream data processing node of the current data processing node triggers the second fault feature corresponding to the second fault feature based on the second fault feature; the functional prompt information is used to indicate the target data obtained by the current data processing node, and the processing rules for the target data, the target data being the output data generated after processing by the upstream data processing node; the risk prompt information is used to indicate the target fault feature that has appeared in the upstream data processing node. Figure 3 A schematic diagram of a structure of a thought chain prompt information provided by an embodiment of the present disclosure, such as Figure 3As shown, taking the thinking chain prompt information Info_P corresponding to the data processing node P as an example, Info_P includes causal prompt information Info_P_1, functional prompt information Info_P_2 and risk prompt information Info_P_3. Among them, the content of the causal prompt information Info_P_1 is, for example: "1. The increase of the cpu.busy index of the instance machine of the upstream data processing node A will cause the increase of the data pulling delay index of the data processing node P; 2. The increase of the write qps index of the upstream data processing node A may cause the outflow throughput index of the data processing node P to increase; 3. The current limiting of the data processing node P will cause the delay index of the upstream data processing node A to increase". Based on the above example, three causal relationship rules are recorded in the causal prompt information Info_P_1, namely: first, the upstream data processing node A triggers the characteristic of the increase in the data pulling delay indicator of the data processing node P based on the characteristic of the increase in the cpu.busy indicator (the first fault characteristic); second, the upstream data processing node A triggers the rule that the outflow throughput indicator of the data processing node P increases based on the characteristic of the increase in the write qps indicator (the first fault characteristic); third, the data processing node P triggers the characteristic of the increase in the delay indicator of the data processing node A (the second fault characteristic) based on the current limiting characteristic (the first fault characteristic).

[0044] Furthermore, the content of the function prompt information Info_P_2 is, for example: "Pull the native binlog data of the upstream data processing node A and perform format conversion." Based on the above example, the causal prompt information records that the current data processing node obtains the native binlog data (target data) from the data processing node A, and performs format conversion (processing rules) on the native binlog data. After the data processing node P processes the native binlog data (target data) based on the function prompt information Info_P_2, the generated data is the output data generated by the data processing node P. The output data can be used as the target data of the downstream data processing node, and the above process is repeated based on the function prompt information of the downstream data processing node.

[0045] Furthermore, the content of the risk warning information Info_P_3 is, for example: "The write qps index of the upstream data processing node A has increased." Based on the above example, the causal warning information records the target fault characteristics that have appeared in the upstream data processing node A, that is, the characteristics of the increase in the qps index. In the subsequent processing steps, the above content in the risk warning information Info_P_3 will be used as reference information to guide the fault diagnosis model to perform fault diagnosis. Based on the above introduction, it can be seen that the causal prompt information and functional prompt information in the risk warning information are static information corresponding to the data processing node, which can be predetermined by loading a configuration file, etc. The risk warning information is dynamic information. In the subsequent process of executing multiple rounds of input into the fault diagnosis model in sequence, the risk warning information will be updated according to the multiple rounds of diagnosis of the fault diagnosis model, so as to achieve more accurate fault diagnosis and provide the accuracy of the generated fault diagnosis information.

[0046] Among them, further, the thinking chain prompt information can be pre-generated information, which is set corresponding to the data processing node and changes with the running state of the data processing node. The generation method of the thinking chain prompt information can be manual configuration by the user or dynamic configuration based on preset rules, which is not limited in this embodiment.

[0047] Step S102: The fault information matrix corresponding to each data processing node is input into the fault diagnosis model in turn in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents the fault handling suggestions for the target fault.

[0048] Exemplarily, after obtaining the fault information matrix, the fault information matrix is ​​used as input data, and is sequentially input into the fault diagnosis model in multiple rounds, and processed using the fault diagnosis model. Among them, the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round. The multiple rounds of sequential input referred to here means that each time the fault information matrix corresponding to a data processing node is input into the fault diagnosis model, and after obtaining the corresponding model output, the model output and the fault information matrix corresponding to the next data processing node are input into the fault diagnosis model, and so on, until the fault information matrix corresponding to the last data processing node is input into the fault diagnosis model, and the corresponding model output is obtained.

[0049] In one possible implementation, the model output of each round will be stored separately, and in the next round, the model output and the fault information matrix are sent to the fault diagnosis model as input data for processing. In another possible implementation, the model output of each round is used to update the fault information matrix of the next data processing node. More specifically, for example, the thinking chain prompt information in the fault information matrix of the next data processing node is updated to obtain an updated fault information matrix. After that, in the next round, the updated fault information matrix corresponding to the next data processing node is sent to the fault diagnosis model as input data for processing, that is, combined with the collected information of the next data processing node to obtain the corresponding model output.

[0050] Finally, when the fault information matrix corresponding to the last data processing node is processed by the fault diagnosis model, the fault diagnosis model will further perform an additional text generation step based on the model output of the last round to generate fault diagnosis information that represents the fault handling suggestions for the target fault. The target fault may refer to the fault characteristics of each data processing node determined in the previous multiple rounds, or the key fault characteristics of the upstream data processing node that causes the downstream data processing node to have fault characteristics. For example, the fault characteristic T1 of the upstream data processing node A causes different fault characteristics to appear in multiple data processing nodes such as the downstream data processing node A and the data processing node B, then the fault characteristic T1 of the data processing node A is determined as the key fault characteristic, that is, the target fault, and this process can be implemented by the fault diagnosis model. Afterwards, the fault diagnosis model generates a description text of the corresponding fault handling suggestions for the target fault, that is, generates fault diagnosis information, for example, "It is recommended to restart the XX data processing node". Afterwards, optionally, the server sends the fault diagnosis information to the terminal device by actively pushing or passively accessing the fault diagnosis information, so that the operation and maintenance user on the terminal device side can maintain the corresponding data processing node based on the above fault diagnosis information.

[0051] In this embodiment, after obtaining the fault alarm information, the fault information matrix corresponding to each data processing node in the data processing link is obtained, and the fault information matrix includes the combined state characteristics and the thinking chain prompt information corresponding to the data processing node, the combined state characteristics represent the operating state of the data processing node in at least two dimensions, and the thinking chain prompt information represents the causal relationship between the data processing nodes with upstream and downstream relationships that trigger the fault; the fault information matrix corresponding to each data processing node is input into the fault diagnosis model in turn in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents the fault handling suggestions for the target fault. By inputting the fault information matrix composed of the combined state characteristics and the thinking chain prompt information into the fault diagnosis model in multiple rounds, the ability of the fault diagnosis model is fully utilized, and the diagnosis of each data processing node in the data processing link is realized in a multi-round interactive manner, and the corresponding fault handling suggestions are generated, thereby improving the accuracy of fault diagnosis and the coverage of faults.

[0052] refer to Figure 4 , Figure 4 A schematic diagram of a method for diagnosing a fault in a data processing link according to an embodiment of the present disclosure Figure 2 In this embodiment Figure 2 On the basis of the embodiment shown, step S102 is further refined, and a step of generating fault alarm information is added. The fault diagnosis method for the data processing link includes:

[0053] Step S201: Based on the data acquisition system for the data processing link, monitoring data corresponding to the data processing link is acquired.

[0054] Step S202: Split the monitoring data based on the data processing nodes to obtain the indicator features corresponding to each data processing node.

[0055] Step S203: Call the fault alarm model to process the indicator characteristics corresponding to each data processing node in turn through the health monitoring dialogue method, and generate fault alarm information when an abnormality occurs. The fault alarm model has a first number of neuron parameters, and the fault diagnosis model has a second number of neuron parameters, and the first number is less than the second number.

[0056] For example, based on Figure 2In the illustrated embodiment, a functional unit for monitoring data processing nodes, i.e., a fault alarm unit, is provided in the data synchronization system. The fault alarm unit collects data from each data processing node in the data processing link based on the data acquisition system to obtain corresponding monitoring data. In one possible implementation, the monitoring data is a set of index characteristics of each data processing node; in another possible implementation, the monitoring data is a set of operating parameters that characterize the operating status of each data processing node, such as CPU occupancy, memory occupancy, abnormal information, logs, and other information. Among them, the fault alarm unit can be deployed inside the data monitoring server or in other electronic devices. Among them, the monitoring data can be obtained after the data acquisition system collects data according to the preset sampling rules, and the specific content can be set based on the role of each data processing node and the specific implementation method. Further, the data acquisition system is provided with an interface for obtaining the above-mentioned monitoring data. The fault alarm unit periodically obtains the above-mentioned monitoring data from the data acquisition system by calling the interface. After that, the fault alarm unit splits all the monitoring data in the above-mentioned monitoring cycle according to the node identifier, thereby obtaining the index characteristics corresponding to each data processing node. Among them, the indicator characteristics refer to various indicators in the operation process of the data processing unit, and different data processing units may correspond to different indicators. Afterwards, the fault alarm unit compares the indicator characteristics in the operation process of each data processing unit with the corresponding preset indicator threshold to determine whether the data processing unit has an abnormality. If so, a fault alarm information representing the abnormality of the data processing unit is generated.

[0057] Exemplarily, the indicator feature includes a category identifier representing the indicator category, and a corresponding indicator value, wherein the indicator category includes a time domain indicator and a corresponding frequency domain indicator, and the frequency domain indicator is generated based on the frequency domain conversion of the time domain operation data of the data processing node in the current detection cycle. Specifically, the indicator feature obtained by the fault alarm unit includes a time domain indicator and a corresponding frequency domain indicator. For example, for a certain data processing unit, it includes a "CPU occupancy rate" indicator. For the "CPU occupancy rate" indicator, the indicator feature includes a "CPU occupancy rate" time domain indicator and a "CPU occupancy rate" frequency domain indicator, wherein the "CPU occupancy rate" time domain indicator represents the characteristics of the CPU occupancy rate of the data processing node in the time domain, such as the curve of the CPU occupancy rate changing over time, thereby showing the overall trend of the CPU occupancy rate changing over time; and the "CPU occupancy rate" frequency domain indicator represents the characteristics of the CPU occupancy rate of the data processing node in the frequency domain, such as the spectrum of the CPU occupancy rate over time, thereby showing the jump frequency of the CPU occupancy rate.

[0058] In the steps of this embodiment, by setting the time domain indicator and the corresponding frequency domain indicator in the indicator characteristics, the indicator characteristics of the data processing node in the two dimensions of time domain and frequency domain can be expressed, thereby improving the content richness of the indicator characteristics, and further providing the accuracy of the fault alarm performed by the fault alarm unit based on the indicator characteristics.

[0059] For example, Figure 5 As shown, the specific implementation of step S203 includes:

[0060] Step S2031: Obtain a pre-trained fault alarm model.

[0061] Step S2032: Call the fault alarm model to process the indicator characteristics corresponding to each data processing node in turn through multiple rounds of health monitoring dialogues. When the inspection results of this round of health dialogues are normal, the inspection results of this round are used as the thinking chain prompts for the next round of health checks and the next round of time series data combination to trigger the next round of health monitoring until an abnormality is determined to generate a fault alarm information or all rounds are traversed to confirm that the data is synchronized to a healthy state.

[0062] Exemplarily, a fault alarm model for fault alarm is provided inside or outside the fault alarm unit, and the fault alarm is realized and fault alarm information is generated by calling the fault alarm model. Among them, the fault alarm information can be a vector data used to characterize different data processing nodes and different types of faults. The alarm model is a pre-trained neural network model, and the fault alarm model determines whether the data processing node is abnormal by processing the indicator characteristics corresponding to the data processing node. Since the specific content of the indicator characteristics is pre-defined, it is less difficult to map the corresponding fault state based on the indicator characteristics. Therefore, the fault alarm model is smaller in size and has fewer neurons than the fault diagnosis model, that is, the fault alarm model has a first number of neuron parameters, and the fault diagnosis model has a second number of neuron parameters, and the first number is less than the second number. In a possible implementation method, the first number is much smaller than the second number, and the difference between the two is more than 10 times.

[0063] Furthermore, in the specific process, the fault alarm model detects each data processing node in a multi-round detection manner. Specifically, if the inspection result of this round of monitoring dialogue is normal, the inspection result of this round of dialogue will be used as the thinking chain prompt for the next round of fault alarm detection and the next round of combined state characteristics to trigger the next round of fault alarm detection until all upstream and downstream inspections are completed, and it is determined that all data processing nodes at this moment are in a healthy state; if any detection result is abnormal in the process, a fault alarm message is generated.

[0064] In this embodiment, a smaller fault alarm model is set to perform fault monitoring and alarm. After the fault is discovered (fault alarm information is generated), a larger and more complex fault diagnosis model is used to perform fault diagnosis. This can improve the real-time nature of the fault alarm, reduce the overall resource overhead of the system, and improve resource utilization.

[0065] Step S204: after obtaining the fault alarm information, generate a fault information matrix corresponding to each data processing node in the data processing link.

[0066] Exemplarily, after the fault alarm information is generated by the fault alarm unit, the fault monitoring server responds to the fault alarm information and further generates a fault information matrix corresponding to each data processing node in the data processing link, wherein the fault information matrix includes the combined state characteristics corresponding to the data processing node and the thinking chain prompt information, and the thinking chain prompt information includes causal prompt information, functional prompt information and risk prompt information. Among them, the causal prompt information and functional prompt information in the thinking chain prompt information are preset information, and the risk prompt information is the information generated in the process of inputting multiple rounds of fault diagnosis model. The specific implementation method is as follows: Figure 2 The embodiments shown have been described in detail and will not be repeated here. In one possible implementation, the combined state feature is obtained by the fault monitoring server by accessing each data processing node, or accessing a data acquisition system for collecting monitoring data of each data processing node. Exemplarily, the combined state feature includes at least two of an indicator feature, a log feature, and an operation feature, wherein the indicator feature is generated based on the data indicator of the data processing node; the log feature is generated based on the error log recorded by the data processing node; and the operation feature is generated based on the process operation information of the data processing node. In a more specific implementation, the combined state feature includes an indicator feature, a log feature, and an operation feature, and the combined state feature is obtained when executing step S202, that is, the monitoring data contains the indicator feature, the log feature, and the operation feature corresponding to the data processing node. Based on the node identification of the data processing node, the monitoring data is split to obtain the combined state feature corresponding to each data processing node.

[0067] Furthermore, if Figure 6 As shown, in a possible implementation, the fault information matrix also includes fault reference information. The specific implementation of step S204 includes:

[0068] Step S2041: Acquire a fault knowledge base based on the retrieval enhancement generation model.

[0069] Step S2042: query the fault knowledge base based on the combined state feature to obtain fault reference information.

[0070] Step S2043: Generate a fault information matrix according to the fault reference information.

[0071] Exemplarily, after obtaining the combined state feature based on the previous steps, the fault monitoring server further uses the combined state feature to query the fault knowledge base based on the retrieval-augmented generation (RAG) model. Among them, the retrieval-augmented generation model is a text generation model that combines language models and information retrieval technology. Specifically, when the model needs to generate text or answer questions, it will first retrieve relevant information from a large document collection, and then use the retrieved information to guide the generation of text, thereby improving the quality and accuracy of the prediction. The fault knowledge base is a model that stores a large amount of prior knowledge of fault diagnosis, and the prior knowledge of fault diagnosis is the mapping relationship between the fault state and the fault type and the fault handling suggestion. By querying the fault knowledge base based on the combined state feature, the fault reference information for the combined state feature can be obtained, wherein the fault reference information may include description information of the fault phenomenon and fault type, and may also include description information of the rules for handling specific faults. The specific implementation method is determined by the specific implementation method of the fault knowledge base, which will not be repeated this time.

[0072] Afterwards, based on the fault reference information, the combined state characteristics obtained in the previous steps, and the thought chain prompt information, a fault information matrix consisting of the fault reference information, the combined state characteristics, and the thought chain prompt information is constructed.

[0073] Figure 7 A schematic diagram of a process for generating a fault information matrix provided by an embodiment of the present disclosure is shown in FIG. Figure 7 As shown, exemplarily, first, by accessing each data processing node, the combined state characteristics and thinking chain prompt information corresponding to each data processing node are obtained, wherein the combined state characteristics include indicator characteristics, log characteristics and operation characteristics; the thinking chain prompt information includes cause-effect prompt information, function prompt information and risk prompt information; then, based on the combined state characteristics, the RAG knowledge base is queried, and the relevant prior knowledge in the RAG knowledge base is used to obtain the corresponding fault reference information; finally, the combined state characteristics, thinking chain prompt information and fault reference information constitute a fault information matrix, and are input into the fault diagnosis model in the subsequent steps.

[0074] Step S205: Obtain the fault information matrix corresponding to the current data processing node, and input the fault information matrix into the fault diagnosis model to obtain model prompt information corresponding to the current data processing node. The model prompt information is used to indicate the fault characteristics of the current data processing node.

[0075] Step S206: If the target stop condition is not triggered, execute step S207; if the target stop condition is triggered, execute step S209.

[0076] Step S207: configuring the fault information matrix of the next data processing node based on the model prompt information.

[0077] Exemplarily, when obtaining the fault information matrix, starting from the first current data processing node, the fault information matrix is ​​sequentially input into the fault diagnosis model, and after being processed by the fault diagnosis model, the model prompt information corresponding to the current data processing node output by the fault diagnosis model is obtained, that is, the model output. Afterwards, it is determined whether the target stop condition is triggered, wherein the target stop condition may be that the current data processing node is the last current data processing node, or that the target fault represented by the model prompt information is a preset specific fault (e.g., a more serious fault). According to the judgment result, if the above-mentioned target stop condition is not triggered, the fault information matrix of the next data processing node is configured based on the model prompt information, so that the fault feature matrix of the next data processing node contains information representing the fault feature of the current data processing node, thereby realizing the utilization of context information.

[0078] For example, Figure 8 As shown, the specific implementation steps of step S207 include:

[0079] Step S2071: Generate risk warning information based on the model prompt information.

[0080] Step S2072: Record the risk warning information to the thinking chain prompt information corresponding to the next data processing node to obtain the fault information matrix after the next data processing node is configured.

[0081] Exemplarily, after the fault diagnosis model generates model prompt information, risk prompt information is generated according to the fault characteristics represented by the model prompt information. The risk prompt information is information used to indicate the fault characteristics, which is the same as the content of the model prompt information, but can be expressed in different data forms and structures. Afterwards, the risk prompt information is written into the thinking chain prompt information corresponding to the next data processing node. Specifically, exemplary, the risk prompt information in the thinking chain prompt information in the default state can be empty. For example, the risk prompt information of the thinking chain prompt information corresponding to the first data processing node is empty. Afterwards, after the fault information matrix corresponding to the first data processing node is input into the fault diagnosis model and the corresponding model prompt information is obtained, risk prompt information is generated based on the model prompt information, and the risk prompt information is written into the thinking chain prompt information of the second data processing node, thereby realizing the update of the thinking chain prompt information. Afterwards, based on the updated thinking chain prompt information, as well as the causal prompt information and function prompt information corresponding to the next data processing node, the fault information matrix after the configuration of the next data processing node is obtained.

[0082] Step S208: Set the next data processing node as the current data processing node, and return to execute step S205.

[0083] Step S209: Obtain fault diagnosis information based on the model prompt information.

[0084] Furthermore, after that, the thread polls the next data processing node, and the next data processing node is set as the current data processing node, and the above processing steps are repeated, and the fault diagnosis model processes the corresponding fault information matrix to obtain the model prompt information. And so on, after multiple rounds of interaction, until the target stop condition is finally triggered, according to the model prompt information obtained in the last round, the corresponding fault diagnosis information is generated by using the ability of the fault diagnosis model. Due to the multiple rounds of reasoning process, the difficulty of fault diagnosis reasoning based on the fault diagnosis model can be reduced, and the accuracy of the fault diagnosis information finally generated can be improved.

[0085] Fig. 9 A schematic diagram of a process for generating fault diagnosis information provided by an embodiment of the present disclosure is shown below in conjunction with Fig. 9 The above process is further introduced as follows: Fig. 9As shown, exemplarily, first, the monitoring data corresponding to the data processing link is obtained through the data acquisition system, and then, after the monitoring data is split, the indicator features corresponding to each data processing node are generated, such as the data processing node P_1, data processing node P_2, data processing node P_n, etc. shown in the figure. The corresponding indicator features M_1, indicator features M_2, indicator features M_n, etc. are respectively input into the fault alarm model in a multi-round interactive manner, wherein the model output of the fault alarm model in each round is the model input of the fault diagnosis model in the next round. For example, after the indicator feature M_1 is input into the fault alarm model, the fault alarm model outputs Out_1. Afterwards, Out_1 output by the fault alarm model and the indicator feature M_2 are input into the fault alarm model together, and Out_2 output by the fault alarm model in the second round is obtained, and so on, until the last indicator feature M_n is input into the fault alarm model. In this process, the fault alarm model makes an alarm judgment on the operation status of the corresponding data processing node according to each indicator feature. When it is judged that a certain data processing node has a fault (shown as the Y path in the figure), the above process of receiving and processing the indicator feature is interrupted, and a fault alarm information is generated; if all data processing nodes have a fault (shown as the N path in the figure), the above steps are repeated to continue monitoring. After that, the fault diagnosis server obtains the indicator features corresponding to each of the above data processing nodes, queries the RAG knowledge base for each characteristic indicator, and generates fault reference information according to the query results. At the same time, it combines the thinking chain prompt information and the risk prompt information to generate the fault information matrix corresponding to the data processing node, such as Mat_1, Mat_2, Mat_n, etc. shown in the figure. After that, the above fault information matrices Mat_1, Mat_2, and Mat_n are sequentially input into the fault diagnosis model in a multi-round interactive manner, and the model prompt information output by the fault diagnosis model in each round is used to update the thinking chain prompt information in the fault information matrix of the next round (specifically, the risk prompt information in the thinking chain prompt information is updated). This is repeated until the last fault information matrix Mat_n is input into the fault diagnosis model, which generates fault alarm information by calling the fault diagnosis component. At the same time, optionally, the above fault alarm information can also be stored in the RAG knowledge base to update the fault diagnosis prior knowledge in the RAG knowledge base.

[0086] Corresponding to the fault diagnosis method for the data processing link in the above embodiment, Fig.10A structural block diagram of a fault diagnosis device for a data processing link provided in an embodiment of the present disclosure. The method described in the above embodiment can be executed by the fault diagnosis device for a data processing link, which can be implemented by software and / or hardware, and can be integrated in an electronic device with certain data processing functions. The electronic device can include but is not limited to a mobile terminal with big data processing capabilities, and a fixed terminal with big data processing capabilities such as a desktop computer and a supercomputer.

[0087] For the sake of convenience, only the parts related to the embodiments of the present disclosure are shown. Fig.10 , the fault diagnosis device 3 for the data processing link includes:

[0088] The acquisition module 31 is used to acquire the fault information matrix corresponding to each data processing node in the data processing link after obtaining the fault alarm information, wherein the fault information matrix includes the combined state characteristics and the thought chain prompt information corresponding to the data processing node, wherein the combined state characteristics represent the operation state of the data processing node in at least two dimensions, and the thought chain prompt information represents the causal relationship between the data processing nodes having the upstream and downstream relationship that trigger the fault;

[0089] The processing module 32 is used to input the fault information matrix corresponding to each data processing node into the fault diagnosis model in turn in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents the fault handling suggestions for the target fault.

[0090] According to one or more embodiments of the present disclosure, the processing module 32 is specifically used to: loop the following steps: obtain the fault information matrix corresponding to the current data processing node, and input the fault information matrix into the fault diagnosis model to obtain the model prompt information corresponding to the current data processing node, the model prompt information is used to indicate the fault characteristics of the current data processing node; if the target stop condition is not triggered, update the thinking chain prompt information in the fault information matrix based on the model prompt information to obtain the updated fault information matrix, and set the updated fault information matrix as the fault information matrix corresponding to the next data processing node; set the next data processing node to the current data processing node.

[0091] According to one or more embodiments of the present disclosure, when the processing module 32 updates the thinking chain prompt information in the fault information matrix based on the model prompt information to obtain the updated fault information matrix, it is specifically used to: record the model prompt information to the thinking chain prompt information to obtain the thinking chain prompt information corresponding to the next data processing node; and obtain the updated fault information matrix according to the thinking chain prompt information corresponding to the next data processing node.

[0092] According to one or more embodiments of the present disclosure, the thinking chain prompt information includes causal prompt information, functional prompt information and risk prompt information. The causal prompt information is used to characterize the rule that the upstream data processing node of the current data processing node triggers the second fault feature corresponding to the current data processing node based on the first fault feature, and / or the current data processing node triggers the rule of the second fault feature corresponding to the upstream data processing node of the current data processing node based on the second fault feature; the functional prompt information is used to indicate the target data obtained by the current data processing node, and the processing rules for the target data, the target data being the output data generated after processing by the upstream data processing node; the risk prompt information is used to indicate the target fault feature that has appeared in the upstream data processing node.

[0093] According to one or more embodiments of the present disclosure, the fault information matrix also includes fault reference information, and the processing module 32 is further used to: obtain a fault knowledge base based on a retrieval enhanced generation model; query the fault knowledge base based on combined state features to obtain fault reference information; and generate a fault information matrix based on the fault reference information.

[0094] According to one or more embodiments of the present disclosure, the combined status characteristics include at least two of the indicator characteristics, log characteristics and operation characteristics: the indicator characteristics are generated based on the data indicators of the data processing node; the log characteristics are generated based on the error logs recorded by the data processing node; and the operation characteristics are generated based on the process operation information of the data processing node.

[0095] According to one or more embodiments of the present disclosure, the acquisition module 31 is also used to: acquire monitoring data corresponding to the data processing link, split the monitoring data based on the data processing nodes to obtain indicator characteristics corresponding to each data processing node; call the fault alarm model to process the indicator characteristics corresponding to each data processing node in turn through a health monitoring dialogue, and generate fault alarm information when an abnormality occurs, the fault alarm model has a first number of neuron parameters, and the fault diagnosis model has a second number of neuron parameters, and the first number is less than the second number.

[0096] According to one or more embodiments of the present disclosure, when the acquisition module 31 calls the fault alarm model to process the indicator characteristics corresponding to each data processing node in sequence through a health monitoring dialogue, and determines that a fault alarm message is generated when an abnormality occurs, it is specifically used to: call the fault alarm model to process the indicator characteristics corresponding to each data processing node in sequence through multiple rounds of health monitoring dialogues, and when the inspection result of the health dialogue in this round is normal, use the inspection result of this round as a thinking chain prompt for the next round of health inspection and a combination of the next round of time series data to trigger the next round of health monitoring until it is determined that an abnormality is generated to generate a fault alarm message or all rounds are traversed to confirm that the data is synchronized to a healthy state.

[0097] According to one or more embodiments of the present disclosure, the indicator characteristics include a category identifier that characterizes the indicator category, and a corresponding indicator value, wherein the indicator category includes a time domain indicator and a corresponding frequency domain indicator, and the frequency domain indicator is generated based on the frequency domain conversion of the time domain operation data of the data processing node in the current detection cycle.

[0098] The acquisition module 31 and the processing module 32 are connected in sequence. The fault diagnosis device 3 for data processing link provided in this embodiment can implement the technical solution of the above method embodiment, and its implementation principle and technical effect are similar, which will not be described in detail in this embodiment.

[0099] Fig.11 A schematic diagram of the structure of an electronic device provided by an embodiment of the present disclosure is shown in FIG. Fig.11 As shown, the electronic device 4 includes:

[0100] A processor 41, and a memory 42 communicatively connected to the processor 41;

[0101] The memory 42 stores computer executable instructions;

[0102] The processor 41 executes the computer execution instructions stored in the memory 42 to implement the following Figure 2-Figure 9 The fault diagnosis method for a data processing link in the illustrated embodiment.

[0103] Optionally, the processor 41 and the memory 42 are connected via a bus 43 .

[0104] For related instructions, please refer to Figure 2-Figure 9 The relevant descriptions and effects corresponding to the steps in the corresponding embodiments can be understood, and no further elaboration is made here.

[0105] The present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, which are used to implement the present invention when executed by a processor. Figure 2-Figure 9 A fault diagnosis method for a data processing link provided by any one of the corresponding embodiments.

[0106] The present invention provides a computer program product, including a computer program. When the computer program is executed by a processor, the present invention is realized. Figure 2-Figure 9 A fault diagnosis method for a data processing link provided by any one of the corresponding embodiments.

[0107] In order to implement the above embodiment, the embodiment of the present disclosure also provides an electronic device.

[0108] refer to Fig.12, which shows a schematic diagram of the structure of an electronic device 900 suitable for implementing the embodiment of the present disclosure, and the electronic device 900 may be a terminal device or a server. The terminal device may include but is not limited to mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, personal digital assistants (PDAs), tablet computers (Portable Android Devices, PADs), portable multimedia players (PMPs), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig.12 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0109] like Fig.12 As shown, the electronic device 900 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage device 908 to a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the electronic device 900 are also stored. The processing device 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0110] Typically, the following devices may be connected to the I / O interface 905: input devices 906 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 907 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 908 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 909. The communication device 909 may allow the electronic device 900 to communicate with other devices wirelessly or by wire to exchange data. Although Fig.12 The electronic device 900 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0111] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device 909, or installed from a storage device 908, or installed from a ROM 902. When the computer program is executed by the processing device 901, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0112] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0113] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0114] The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0115] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0116] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0117] The units or modules involved in the embodiments described in the present disclosure may be implemented by software or hardware, wherein the name of a unit or module does not, in some cases, constitute a limitation on the unit itself.

[0118] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0119] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0120] In a first aspect, according to one or more embodiments of the present disclosure, a fault diagnosis method for a data processing link is provided, comprising:

[0121] After obtaining the fault alarm information, the fault information matrix corresponding to each data processing node in the data processing link is obtained, wherein the fault information matrix includes the combined state characteristics corresponding to the data processing nodes and the thinking chain prompt information, wherein the combined state characteristics represent the operating state of the data processing nodes in at least two dimensions, and the thinking chain prompt information represents the causal relationship between the data processing nodes with upstream and downstream relationships that trigger the fault; the fault information matrix corresponding to each of the data processing nodes is input into the fault diagnosis model in turn in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents the fault handling suggestions for the target fault.

[0122] According to one or more embodiments of the present disclosure, the fault information matrix corresponding to each of the data processing nodes is sequentially input into a fault diagnosis model in multiple rounds to obtain fault diagnosis information, including:

[0123] The following steps are executed in a loop until the target stop condition is triggered: obtaining a fault information matrix corresponding to the current data processing node, and inputting the fault information matrix into the fault diagnosis model to obtain model prompt information corresponding to the current data processing node, wherein the model prompt information is used to indicate the fault characteristics of the current data processing node; if the target stop condition is not triggered, updating the thinking chain prompt information in the fault information matrix based on the model prompt information to obtain an updated fault information matrix, and setting the updated fault information matrix as the fault information matrix corresponding to the next data processing node; and setting the next data processing node as the current data processing node.

[0124] According to one or more embodiments of the present disclosure, the thinking chain prompt information in the fault information matrix is ​​updated based on the model prompt information to obtain an updated fault information matrix, including: recording the model prompt information to the thinking chain prompt information to obtain the thinking chain prompt information corresponding to the next data processing node; and obtaining an updated fault information matrix according to the thinking chain prompt information corresponding to the next data processing node.

[0125] According to one or more embodiments of the present disclosure, the thought chain prompt information includes causal prompt information, functional prompt information and risk prompt information. The causal prompt information is used to characterize the rule that the upstream data processing node of the current data processing node triggers the second fault feature corresponding to the current data processing node based on the first fault feature, and / or the rule that the current data processing node triggers the second fault feature corresponding to the upstream data processing node of the current data processing node based on the second fault feature; the functional prompt information is used to indicate the target data acquired by the current data processing node, and the processing rules for the target data, and the target data is the output data generated after processing by the upstream data processing node; the risk prompt information is used to indicate the target fault feature that has appeared at the upstream data processing node.

[0126] According to one or more embodiments of the present disclosure, the fault information matrix also includes fault reference information, and the method further includes: obtaining a fault knowledge base based on a retrieval enhanced generation model; querying the fault knowledge base based on the combined state feature to obtain fault reference information; and generating the fault information matrix based on the fault reference information.

[0127] According to one or more embodiments of the present disclosure, the combined status characteristics include at least two of indicator characteristics, log characteristics and operation characteristics: the indicator characteristics are generated based on data indicators of the data processing node; the log characteristics are generated based on error logs recorded by the data processing node; and the operation characteristics are generated based on process operation information of the data processing node.

[0128] According to one or more embodiments of the present disclosure, the method also includes: obtaining monitoring data corresponding to the data processing link, splitting the monitoring data based on the data processing nodes to obtain indicator characteristics corresponding to each data processing node; calling the fault alarm model to process the indicator characteristics corresponding to each data processing node in turn through a health monitoring dialogue, and generating the fault alarm information when an abnormality is determined, the fault alarm model has a first number of neuron parameters, and the fault diagnosis model has a second number of neuron parameters, and the first number is less than the second number.

[0129] According to one or more embodiments of the present disclosure, the fault alarm model is called to process the indicator characteristics corresponding to each of the data processing nodes in sequence through a health monitoring dialogue, and the fault alarm information is generated when an abnormality is determined, including: calling the fault alarm model to process the indicator characteristics corresponding to each of the data processing nodes in sequence through multiple rounds of health monitoring dialogues, and when the inspection result of the health dialogue in this round is normal, the inspection result of this round is used as a thinking chain prompt for the next round of health inspection and a combination of the next round of time series data to trigger the next round of health monitoring until the abnormality is determined to generate the fault alarm information or all rounds are traversed to confirm that the data is synchronized to a healthy state.

[0130] According to one or more embodiments of the present disclosure, the indicator characteristics include a category identifier that characterizes an indicator category, and a corresponding indicator value, wherein the indicator category includes a time domain indicator and a corresponding frequency domain indicator, and the frequency domain indicator is generated based on frequency domain conversion of the time domain operation data of the data processing node in the current detection cycle.

[0131] In a second aspect, according to one or more embodiments of the present disclosure, a fault diagnosis device for a data processing link is provided, comprising:

[0132] An acquisition module is used to acquire a fault information matrix corresponding to each data processing node in the data processing link after obtaining the fault alarm information, wherein the fault information matrix includes a combination state feature and a thought chain prompt information corresponding to the data processing node, wherein the combination state feature represents the operation state of the data processing node in at least two dimensions, and the thought chain prompt information represents the causal relationship of triggering the fault between the data processing nodes having an upstream and downstream relationship;

[0133] A processing module is used to input the fault information matrix corresponding to each of the data processing nodes into the fault diagnosis model in turn in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents the fault handling suggestions for the target fault.

[0134] According to one or more embodiments of the present disclosure, the processing module is specifically used to: loop the following steps: obtain a fault information matrix corresponding to a current data processing node, and input the fault information matrix into the fault diagnosis model to obtain model prompt information corresponding to the current data processing node, wherein the model prompt information is used to indicate the fault characteristics of the current data processing node; if the target stop condition is not triggered, update the thinking chain prompt information in the fault information matrix based on the model prompt information to obtain an updated fault information matrix, and set the updated fault information matrix as the fault information matrix corresponding to the next data processing node; set the next data processing node to the current data processing node.

[0135] According to one or more embodiments of the present disclosure, when the processing module updates the thinking chain prompt information in the fault information matrix based on the model prompt information to obtain the updated fault information matrix, it is specifically used to: record the model prompt information to the thinking chain prompt information to obtain the thinking chain prompt information corresponding to the next data processing node; and obtain the updated fault information matrix according to the thinking chain prompt information corresponding to the next data processing node.

[0136] According to one or more embodiments of the present disclosure, the thought chain prompt information includes causal prompt information, functional prompt information and risk prompt information. The causal prompt information is used to characterize the rule that the upstream data processing node of the current data processing node triggers the second fault feature corresponding to the current data processing node based on the first fault feature, and / or the rule that the current data processing node triggers the second fault feature corresponding to the upstream data processing node of the current data processing node based on the second fault feature; the functional prompt information is used to indicate the target data acquired by the current data processing node, and the processing rules for the target data, and the target data is the output data generated after processing by the upstream data processing node; the risk prompt information is used to indicate the target fault feature that has appeared at the upstream data processing node.

[0137] According to one or more embodiments of the present disclosure, the fault information matrix also includes fault reference information, and the processing module is further used to: obtain a fault knowledge base based on a retrieval enhanced generation model; query the fault knowledge base based on the combined state feature to obtain fault reference information; and generate the fault information matrix based on the fault reference information.

[0138] According to one or more embodiments of the present disclosure, the combined status characteristics include at least two of indicator characteristics, log characteristics and operation characteristics: the indicator characteristics are generated based on data indicators of the data processing node; the log characteristics are generated based on error logs recorded by the data processing node; and the operation characteristics are generated based on process operation information of the data processing node.

[0139] According to one or more embodiments of the present disclosure, the acquisition module is also used to: acquire the monitoring data corresponding to the data processing link, split the monitoring data based on the data processing nodes to obtain the indicator characteristics corresponding to each data processing node; call the fault alarm model to process the indicator characteristics corresponding to each data processing node in turn through a health monitoring dialogue, and generate the fault alarm information when an abnormality is determined, the fault alarm model has a first number of neuron parameters, and the fault diagnosis model has a second number of neuron parameters, and the first number is less than the second number.

[0140] According to one or more embodiments of the present disclosure, when the acquisition module calls the fault alarm model to process the indicator characteristics corresponding to each of the data processing nodes in sequence through a health monitoring dialogue, and generates the fault alarm information when an abnormality is determined, it is specifically used to: call the fault alarm model to process the indicator characteristics corresponding to each of the data processing nodes in sequence through multiple rounds of health monitoring dialogues, and when the inspection result of the health dialogue in this round is normal, use the inspection result of this round as the thinking chain prompt for the next round of health inspection and the next round of time series data combination to trigger the next round of health monitoring until it is determined that an abnormality is generated to generate the fault alarm information or all rounds are traversed to confirm that the data is synchronized to a healthy state.

[0141] According to one or more embodiments of the present disclosure, the indicator characteristics include a category identifier that characterizes an indicator category, and a corresponding indicator value, wherein the indicator category includes a time domain indicator and a corresponding frequency domain indicator, and the frequency domain indicator is generated based on frequency domain conversion of the time domain operation data of the data processing node in the current detection cycle.

[0142] In a third aspect, according to one or more embodiments of the present disclosure, there is provided an electronic device, comprising: at least one processor and a memory;

[0143] The memory stores computer-executable instructions;

[0144] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the fault diagnosis method for the data processing link as described in the first aspect and various possible designs of the first aspect.

[0145] In a fourth aspect, according to one or more embodiments of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer execution instructions. When a processor executes the computer execution instructions, the fault diagnosis method for the data processing link as described in the first aspect and various possible designs of the first aspect is implemented.

[0146] In a fifth aspect, according to one or more embodiments of the present disclosure, a computer program product is provided, including a computer program, which, when executed by a processor, implements the fault diagnosis method for a data processing link as described in the first aspect and various possible designs of the first aspect.

[0147] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

[0148] In addition, although each operation is described in a specific order, this should not be understood as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0149] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

Claims

1. A fault diagnosis method for a data processing link, characterized in that: include: After obtaining the fault alarm information, a fault information matrix corresponding to each data processing node in the data processing link is obtained, wherein the fault information matrix includes a combination state feature corresponding to the data processing node and a thought chain prompt information, wherein the combination state feature represents the operation state of the data processing node in at least two dimensions, and the thought chain prompt information represents the causal relationship of triggering the fault between the data processing nodes having an upstream and downstream relationship; Inputting the fault information matrix corresponding to each of the data processing nodes into the fault diagnosis model in turn in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents the fault handling suggestion for the target fault; The thought chain prompt information includes causal prompt information, functional prompt information and risk prompt information, and the causal prompt information is used to characterize the rule that the upstream data processing node of the current data processing node triggers the second fault feature corresponding to the current data processing node based on the first fault feature, and / or the rule that the current data processing node triggers the second fault feature corresponding to the upstream data processing node of the current data processing node based on the second fault feature; The function prompt information is used to indicate the target data acquired by the current data processing node and the processing rules for the target data, wherein the target data is output data generated after being processed by the upstream data processing node; The risk warning information is used to indicate target fault characteristics that have occurred in the upstream data processing node.

2. The method according to claim 1, characterized in that: The fault information matrix corresponding to each of the data processing nodes is sequentially input into the fault diagnosis model in multiple rounds to obtain fault diagnosis information, including: The following steps are executed repeatedly until the target stop condition is triggered: Obtaining a fault information matrix corresponding to a current data processing node, and inputting the fault information matrix into the fault diagnosis model to obtain model prompt information corresponding to the current data processing node, wherein the model prompt information is used to indicate a fault feature occurring at the current data processing node; If the target stop condition is not triggered, the fault information matrix of the next data processing node is configured based on the model prompt information; if the target stop condition is triggered, the fault diagnosis information is obtained based on the model prompt information; The next data processing node is set as the current data processing node.

3. The method according to claim 2, characterized in that The configuring the fault information matrix of the next data processing node based on the model prompt information includes: Generate risk warning information according to the model prompt information; The risk warning information is recorded in the thinking chain prompt information corresponding to the next data processing node to obtain the fault information matrix after the next data processing node is configured.

4. The method according to claim 1, characterized in that: The fault information matrix also includes fault reference information, and the method further includes: Obtaining a fault knowledge base based on a retrieval-enhanced generation model; Based on the combined state feature, the fault knowledge base is queried to obtain fault reference information; The fault information matrix is ​​generated according to the fault reference information.

5. The method according to claim 1, characterized in that The combined state feature includes at least two of an indicator feature, a log feature, and an operation feature: The indicator feature is generated based on the data indicator of the data processing node; The log feature is generated based on the error log recorded by the data processing node; The operation feature is generated based on process operation information of the data processing node.

6. The method according to claim 1, characterized in that The method further comprises: Acquire monitoring data corresponding to the data processing link, and split the monitoring data based on the data processing nodes to obtain indicator features corresponding to each data processing node; The fault alarm model is called to process the indicator characteristics corresponding to each of the data processing nodes in turn through a health monitoring dialogue, and the fault alarm information is generated when an abnormality is determined. The fault alarm model has a first number of neuron parameters, and the fault diagnosis model has a second number of neuron parameters, and the first number is less than the second number.

7. The method according to claim 6, characterized in that The calling of the fault alarm model sequentially processes the indicator features corresponding to each of the data processing nodes through a health monitoring dialogue mode, and generates the fault alarm information when an abnormality is determined, including: The fault alarm model is called to process the indicator characteristics corresponding to each data processing node in turn through multiple rounds of health monitoring dialogues. When the inspection result of this round of health dialogue is normal, the inspection result of this round is used as the thinking chain prompt of the next round of health inspection and the combination of the next round of time series data to trigger the next round of health monitoring until the abnormality is determined to generate the fault alarm information or all rounds are traversed to confirm that the data is synchronized to a healthy state.

8. The method according to claim 6, characterized in that The indicator characteristics include a category identifier that characterizes the indicator category and a corresponding indicator value, wherein the indicator category includes a time domain indicator and a corresponding frequency domain indicator, and the frequency domain indicator is generated based on the frequency domain conversion of the time domain operation data of the data processing node in the current detection cycle.

9. A fault diagnosis device for a data processing link, characterized in that: include: An acquisition module is used to acquire a fault information matrix corresponding to each data processing node in the data processing link after obtaining the fault alarm information, wherein the fault information matrix includes a combination state feature and a thought chain prompt information corresponding to the data processing node, wherein the combination state feature represents the operation state of the data processing node in at least two dimensions, and the thought chain prompt information represents the causal relationship of triggering the fault between the data processing nodes having an upstream and downstream relationship; A processing module, used to input the fault information matrix corresponding to each of the data processing nodes into the fault diagnosis model in turn in multiple rounds to obtain fault diagnosis information, wherein the model output of the fault diagnosis model in this round is the model input of the fault diagnosis model in the next round, and the fault diagnosis information represents a fault handling suggestion for the target fault; The thought chain prompt information includes causal prompt information, functional prompt information and risk prompt information, and the causal prompt information is used to characterize the rule that the upstream data processing node of the current data processing node triggers the second fault feature corresponding to the current data processing node based on the first fault feature, and / or the rule that the current data processing node triggers the second fault feature corresponding to the upstream data processing node of the current data processing node based on the second fault feature; The function prompt information is used to indicate the target data acquired by the current data processing node and the processing rules for the target data, wherein the target data is output data generated after being processed by the upstream data processing node; The risk warning information is used to indicate target fault characteristics that have occurred in the upstream data processing node.

10. An electronic device, characterized in that: include: Processor and memory; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor executes the fault diagnosis method for a data processing link according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions. When the processor executes the computer-executable instructions, the fault diagnosis method for a data processing link according to any one of claims 1 to 8 is implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the fault diagnosis method for a data processing link according to any one of claims 1 to 8 is implemented.

Images