Non-intrusive cluster diagnosis method and device based on AI and electronic equipment

Abstract

The invention provides a non-intrusive cluster diagnosis method and device based on AI and electronic equipment, and relates to the technical field of data processing.The method comprises the steps that real-time event monitoring is conducted on a POD cluster through a local network in a POD, cluster operation data are collected, and the cluster operation data comprise POD service operation data and peripheral equipment state data; obtaining a service resource dependency relationship among the POD services through the POD service operation data; obtaining a fault root cause service according to all the service resource dependency relationships; based on multi-model collaborative prediction, predicting the peripheral equipment state data to obtain peripheral equipment prediction data; and when the peripheral equipment prediction data is greater than an equipment threshold value, generating equipment fault early warning. According to the invention, the efficiency and precision of cluster diagnosis are improved.

Description

AI-based non-invasive cluster diagnostic methods, devices, and electronic equipment Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to an AI-based non-invasive cluster diagnostic method, apparatus, and electronic device. Background Technology

[0002] In containerized clusters (such as Kubernetes), diagnosing the service status, load, and resources of multiple containers within a Point of Container (POD) is crucial for ensuring system stability. Current diagnostic methods for containerized clusters typically involve directly executing commands within the target POD or capturing network packets on the nodes. These methods are highly invasive, potentially interfering with running services and hindering accurate fault reproduction. Furthermore, the diagnostic process heavily relies on manual intervention and the subjective experience of operations personnel, resulting in low efficiency and difficulty in standardization. Summary of the Invention

[0003] The present invention aims to solve at least one of the above-mentioned problems.

[0004] To address the aforementioned issues, this invention provides an AI-based non-invasive cluster diagnostic method, apparatus, and electronic device.

[0005] In a first aspect, the present invention provides an AI-based non-intrusive cluster diagnosis method, comprising: real-time event monitoring of the POD cluster via the local network within the POD, and collection of cluster operation data, wherein the cluster operation data includes POD service operation data and peripheral device status data; obtaining service resource dependencies between each POD service through the POD service operation data; obtaining the root cause service of the fault based on all the service resource dependencies; predicting the peripheral device status data based on multi-model collaborative prediction to obtain peripheral device prediction data; and generating a device fault warning when the peripheral device prediction data is greater than a device threshold.

[0006] Optionally, obtaining the service resource dependencies between each POD service through the POD service runtime data includes: obtaining the service resource dependencies between each POD service through the POD service runtime data; wherein, the service resource dependencies include: ,in, For service m to call the downstream service n, the service resource dependencies are defined. The average latency for service m to call downstream service n. The error rate of service m calling downstream service n. The number of requests per unit time that service m calls downstream service n. , , They are respectively , , The maximum value, , , These are the average latency weight, error rate weight, and request volume weight per unit time, respectively.

[0007] Optionally, obtaining the root cause service based on all the service resource dependencies includes: obtaining the corresponding resource consumption coefficients based on the service resource dependencies; obtaining the root cause score through the resource consumption coefficients; sorting all the root cause scores in descending order, and taking the POD service corresponding to the largest root cause score as the root cause service.

[0008] Optionally, obtaining the corresponding resource consumption coefficients based on the service resource dependencies includes: obtaining the corresponding resource consumption coefficients based on the service resource dependencies using attenuation coefficients; wherein the resource consumption coefficients include: ,in, The resource consumption coefficient for serving m. For service m, a collection of downstream services are called. For service m to call the downstream service n, the service resource dependencies are defined. The attenuation coefficient is... The resource consumption coefficient for serving downstream n. For local resource exceptions for service m.

[0009] Optionally, obtaining the root cause score through the resource consumption coefficient includes: obtaining the root cause score through the resource consumption coefficient; wherein, the root cause score includes: ,in, Score the root cause. The resource consumption coefficient for serving m. This represents downstream dependence.

[0010] Optionally, the step of predicting the peripheral device status data to obtain peripheral device prediction data includes: inputting the peripheral device status data into multiple sub-models to obtain corresponding sub-model prediction data; and inputting all the sub-model prediction data into a weighted fusion model to obtain the peripheral device prediction data; wherein, the peripheral device prediction data includes: ,in, The peripheral device prediction data at time t. Let be the weight coefficient of the i-th sub-model at time t. The predicted data for the i-th sub-model at time t.

[0011] Optionally, the step of monitoring real-time events of the POD cluster and collecting cluster operation data through the local network within the POD includes: using DiagService to monitor real-time events of the POD cluster and collect cluster operation data through the local network within the POD.

[0012] Secondly, this invention provides an AI-based non-intrusive cluster diagnostic device, comprising: a data acquisition module for real-time event monitoring of the POD cluster via a local network within the POD, and for collecting cluster operation data, wherein the cluster operation data includes POD service operation data and peripheral device status data; a service resource dependency acquisition module for obtaining service resource dependencies between POD services through the POD service operation data; a fault root cause service acquisition module for obtaining fault root cause services based on all the service resource dependencies; a multi-model collaborative prediction module for predicting peripheral device status data based on multi-model collaborative prediction to obtain peripheral device prediction data; and a device fault warning generation module for generating a device fault warning when the peripheral device prediction data exceeds a device threshold.

[0013] Thirdly, the present invention provides an electronic device, including a memory and a processor; the memory is used to store a computer program; the processor is used to implement the AI-based non-invasive cluster diagnostic method as described in the first aspect when the computer program is executed.

[0014] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the AI-based non-intrusive cluster diagnostic method as described in the first aspect.

[0015] The beneficial effects of the AI-based non-intrusive cluster diagnosis method, device, and electronic equipment of this invention are as follows: Real-time event monitoring of the POD cluster is performed through the local network within the POD, collecting cluster operation data without modifying application code in the container, thus avoiding the intrusion of traditional diagnosis methods into business logic and not affecting operational services. Service resource dependencies between POD services are obtained through POD service operation data, dynamically reflecting the relationships between POD services and laying the foundation for subsequent fault diagnosis. The root cause service of the problem is accurately located based on the service resource dependencies, obtaining the root cause service of the fault, and more clearly reflecting the impact of the faulty service. Based on multi-model collaborative prediction, the peripheral device status data is predicted to obtain peripheral device prediction data. When the peripheral device prediction data exceeds the device threshold, a device fault warning is generated, enabling real-time response to abnormal states and significantly improving the efficiency and accuracy of cluster diagnosis. Attached Figure Description

[0016] Figure 1 is a flowchart illustrating a non-invasive cluster diagnostic method based on AI according to an embodiment of the present invention; Figure 2 is a structural diagram illustrating a non-invasive cluster diagnostic device based on AI according to an embodiment of the present invention; Figure 3 is a structural diagram illustrating an electronic device according to an embodiment of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0018] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0019] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0020] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0021] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0022] In related technologies, containerized clusters (such as Kubernetes) are at the core of modern cloud-native infrastructure, used for the automated deployment, scaling, and management of containerized applications. A container is a lightweight, portable software packaging technology that encapsulates an application and its dependencies (code, runtime, libraries, configuration, etc.) in an isolated runtime environment (such as Docker containers). A POD is the smallest deployable unit created and managed in Kubernetes. A POD is like a logical host, a collection of containers sharing a common context, sharing storage and network, and being scheduled together. With the popularization of microservice architectures, the scale of POD clusters has surged, significantly increasing operational complexity. Existing solutions mainly rely on: a single log analysis system or a POD resource monitoring system, resulting in low efficiency in post-event traceability and difficulty in associating multi-dimensional faults. Basic health checks are performed through command monitoring, only verifying process survival and ignoring performance bottlenecks, network quality, and business efficiency. Passive response relies on manual intervention and cannot predict resource exhaustion or business congestion in advance. Monitoring metrics, logs, and network topology data are separated, fault location is time-consuming, there is no autonomous decision-making, and the ability to generate optimization strategies based on historical data is lacking.

[0023] To address the problems existing in the aforementioned related technologies, this embodiment provides an AI-based non-invasive cluster diagnostic method, device, and electronic device.

[0024] As shown in Figure 1, an AI-based non-intrusive cluster diagnosis method provided by an embodiment of the present invention includes: step 110, real-time event monitoring of the POD cluster through the local network within the POD, and collection of cluster operation data, wherein the cluster operation data includes POD service operation data and peripheral device status data.

[0025] Specifically, the runtime status and business metrics of business applications deployed within the same Pod are collected via the local network and through event listening. After initial aggregation and formatting, the collected data is provided externally by DiagService through the NBUS interface. Downstream data consumers such as external monitoring systems and AI diagnostic engines can subscribe to or periodically pull this data through the NBUS interface to obtain accurate Pod-level runtime metrics.

[0026] Step 120: Obtain the service resource dependencies between each POD service through the POD service operation data.

[0027] Specifically, based on service operation data collected non-intrusively from the local network within the POD, a resource dependency topology between services is constructed. For any two POD services m and n, if there is a call from m to n, it is determined that there is a dependency relationship between the two, and the strength of the dependency is further quantified.

[0028] Step 130: Obtain the root cause service based on all the service resource dependencies.

[0029] Specifically, after obtaining service resource dependencies, the root cause service most likely to cause the current cluster anomaly is identified through these dependencies. Taking into account three dimensions—call latency, error rate, and request volume—the resource consumption coefficient, call attenuation coefficient, and local resource anomaly items are calculated. Based on a comparison of actual values ​​and thresholds for metrics such as CPU and memory, a quadratic amplification calculation is used to ultimately determine the root cause service of the failure.

[0030] Step 140: Based on multi-model collaborative prediction, predict the peripheral device status data to obtain peripheral device prediction data.

[0031] In some more specific embodiments, the method for constructing a multi-model collaborative prediction model includes: acquiring historical device data; obtaining a training dataset and a validation dataset based on the historical device data; inputting the training dataset into an initial prediction model for training to obtain initial prediction data; performing loss calculation on the initial prediction data and the validation dataset to obtain a loss function output; adjusting the model parameters of the initial prediction model according to the loss function output until the loss function input meets preset conditions, and using the parameter-adjusted initial diagnostic model as the multi-model collaborative prediction model.

[0032] Step 150: When the predicted data of the peripheral device is greater than the device threshold, a device fault warning is generated.

[0033] Specifically, in real-time monitoring, by comparing the deviation of predicted data from peripheral devices from the equipment threshold, early warnings of latent equipment faults (such as efficiency decline and early mechanical wear) are achieved, generating equipment fault warnings. Secondly, for discrete data points that cannot be covered by the correlation model (such as cylinder temperature and switch status), clear physical thresholds are set. Once the data exceeds the limit (such as excessively high temperature), an over-limit warning is immediately triggered, enabling immediate response to explicit abnormal equipment conditions.

[0034] In some more specific embodiments, diagnostic reports can be generated based on root cause services and peripheral device prediction data, which can further form operational guidance (resource data examples: expanding the number of PODs, migrating to low-load nodes; business data examples: equipment valve pressure overload, requiring a reduction in valve opening, etc.), decision execution and alarms: automatic policy execution (automatic execution of policies when trigger conditions are met), and push of over-limit alarms to designated platforms, along with repair guidance.

[0035] In this embodiment, real-time event monitoring of the POD cluster is performed via the local network within the POD, collecting cluster operation data without modifying the application code in the container. This avoids the intrusion of traditional diagnostic methods into business logic and does not affect operational services. The service resource dependencies between POD services are obtained through the POD service operation data, dynamically reflecting the relationships between them and laying the foundation for subsequent fault diagnosis. The root cause service is accurately located based on the service resource dependencies, revealing the root cause service and more clearly reflecting the impact of the faulty service. Based on multi-model collaborative prediction, the peripheral device status data is predicted to obtain peripheral device prediction data. When the peripheral device prediction data exceeds a device threshold, a device fault warning is generated, enabling real-time response to abnormal states and significantly improving the efficiency and accuracy of cluster diagnosis.

[0036] Optionally, obtaining the service resource dependencies between each POD service through the POD service runtime data includes: obtaining the service resource dependencies between each POD service through the POD service runtime data; wherein, the service resource dependencies include: ,in, For service m to call the downstream service n, the service resource dependencies are defined. The average latency for service m to call downstream service n. The error rate of service m calling downstream service n. The number of requests per unit time that service m calls downstream service n. , , They are respectively , , The maximum value, , , These are the average latency weight, error rate weight, and request volume weight per unit time, respectively.

[0037] Specifically, construct the POD and the inter-service call relationship matrix X[m][m] = { The service resource dependencies of service m calling downstream service n, where, .

[0038] Optionally, obtaining the root cause service based on all the service resource dependencies includes: obtaining the corresponding resource consumption coefficients based on the service resource dependencies; obtaining the root cause score through the resource consumption coefficients; sorting all the root cause scores in descending order, and taking the POD service corresponding to the largest root cause score as the root cause service.

[0039] Optionally, obtaining the corresponding resource consumption coefficients based on the service resource dependencies includes: obtaining the corresponding resource consumption coefficients based on the service resource dependencies using attenuation coefficients; wherein the resource consumption coefficients include: ,in, The resource consumption coefficient for serving m. For service m, a collection of downstream services are called. For service m to call the downstream service n, the service resource dependencies are defined. The attenuation coefficient is... The resource consumption coefficient for serving downstream n. For local resource exceptions for service m.

[0040] Specifically, , .

[0041] In this optional embodiment, the overall resource consumption of the service not only comes from local anomalies, but is also affected by the resource dependencies of downstream services, and decreases exponentially with the depth of the call chain. The corresponding resource consumption coefficient is obtained by using the decay coefficient according to the resource dependencies of the services.

[0042] Optionally, obtaining the root cause score through the resource consumption coefficient includes: obtaining the root cause score through the resource consumption coefficient; wherein, the root cause score includes: ,in, Score the root cause. The resource consumption coefficient for serving m. This represents downstream dependence.

[0043] Specifically, the downstream dependency includes: ,in, To calculate the set of downstream services that are directly or indirectly called by service m in the downstream dependency, The length of the call path from which service m calls downstream service n.

[0044] In this optional embodiment, traditional methods may only detect high CPU usage in a Pod, while this method can quickly pinpoint that a downstream database proxy Pod is experiencing slow processing due to insufficient memory, which in turn triggers an avalanche of the entire call chain. Root cause localization goes from the "symptom occurrence point" to the "root cause of the problem." Furthermore, it can clearly depict the list of all upstream services directly and indirectly affected by a faulty Pod and the degree of impact, transforming the "scope of impact" from a vague description to a precise quantification, greatly improving the accuracy of fault reporting.

[0045] Optionally, the step of predicting the peripheral device status data to obtain peripheral device prediction data includes: inputting the peripheral device status data into multiple sub-models to obtain corresponding sub-model prediction data; and inputting all the sub-model prediction data into a weighted fusion model to obtain the peripheral device prediction data; wherein, the peripheral device prediction data includes: ,in, The peripheral device prediction data at time t. Let be the weight coefficient of the i-th sub-model at time t. The predicted data for the i-th sub-model at time t.

[0046] Specifically, the sub-models can be LSTM (Long Short-Term Memory) models, XGBoost (Ensemble Learning) models, ARIMA (Autoregressive Integral Moving Average) models, or Transformer models. A time-series model allows the program to understand normal behavioral patterns over time, a weighted fusion model performs decision diagnosis, and an uncertainty quantification model calculates credibility, ultimately identifying hidden equipment fault points. Let i be the weight coefficients of the i sub-models at time t: Where, softmax is the softmax function. Let be the root mean square error of the i-th sub-model over the last 300 seconds, and t be time. The scaling factor is defined as follows: The uncertainty quantification model includes: , ,in, Let be the confidence interval of the prediction result at time t, and K be the comprehensive error estimate. Let be the prediction variance of the i-th sub-model. The confidence interval, used to calculate the prediction result, is one of the core technologies for uncertainty quantification. In multi-model collaborative prediction, it is used to assess the reliability of predicted values ​​and provide statistical basis for "over-limit warnings." The confidence interval represents the range of credible prediction values, indicating that the actual value has a high probability of falling within this range.

[0047] In some more specific embodiments, intelligent analysis and anomaly identification are performed on the operating data of equipment or process systems. First, the operating parameters of the equipment (such as valve opening V, inlet pressure P_in, outlet pressure P_out, flow rate F, motor current I, etc.) are structured and categorized. Then, based on a large amount of historical data of the equipment in a healthy state, normal operating conditions are imported (e.g., [I, P_out] = AI_Model(V, P_in, F)). This model defines the normal output range that the system should have under a given input. In real-time monitoring, by comparing the deviation between the actual measured values ​​and the model's predicted values, early warnings of latent equipment faults (such as efficiency decline, early mechanical wear) are achieved. Secondly, for discrete data points that cannot be covered by the correlation model (such as cylinder temperature, switch status), explicit physical thresholds are set. Once the data exceeds the limit (such as excessively high temperature), an over-limit warning is immediately triggered, achieving immediate response to explicit abnormal states of the equipment.

[0048] In this optional embodiment, the method has significant advantages in equipment health monitoring and latent fault early warning. By using a time-series model to enable the program to understand normal behavior patterns in the "time dimension," a weighted fusion model for decision diagnosis, and an uncertainty quantification model for reliability calculation, the method comprehensively obtains the equipment latent fault points, significantly improving the accuracy and reliability of early warning compared to traditional single models.

[0049] Optionally, the step of monitoring real-time events of the POD cluster and collecting cluster operation data through the local network within the POD includes: using DiagService to monitor real-time events of the POD cluster and collect cluster operation data through the local network within the POD.

[0050] Specifically, a lightweight DiagService service is deployed inside each container for real-time data collection. The Diagnostic Agent Service (DiagService) is a lightweight background process embedded within or closely cooperating with the application container. It is used to non-intrusively and automatically collect multi-dimensional runtime data, providing input for AI-driven cluster health diagnostics, fault prediction, and self-healing.

[0051] As shown in Figure 2, an embodiment of the present invention provides an AI-based non-intrusive cluster diagnostic device, comprising: a data acquisition module 10, used to monitor real-time events of the POD cluster through the local network within the POD and collect cluster operation data, wherein the cluster operation data includes POD service operation data and peripheral device status data; a service resource dependency acquisition module 20, used to obtain the service resource dependency relationships between each POD service through the POD service operation data; a fault root cause service acquisition module 30, used to obtain the fault root cause service based on all the service resource dependency relationships; a multi-model collaborative prediction module 40, used to predict the peripheral device status data based on multi-model collaborative prediction to obtain peripheral device prediction data; and a device fault warning generation module 50, used to generate a device fault warning when the peripheral device prediction data is greater than a device threshold.

[0052] The AI-based non-invasive cluster diagnostic device in this embodiment is used to implement the AI-based non-invasive cluster diagnostic method described above. Its advantages over the prior art are the same as the advantages of the AI-based non-invasive cluster diagnostic method over the prior art, and will not be repeated here.

[0053] Optionally, the service resource dependency acquisition module 20 is specifically used to: obtain the service resource dependencies between the POD services through the POD service operation data; wherein, the service resource dependencies include: ,in, For service m to call the downstream service n, the service resource dependencies are defined. The average latency for service m to call downstream service n. The error rate of service m calling downstream service n. The number of requests per unit time that service m calls downstream service n. , , They are respectively , , The maximum value, , , These are the average latency weight, error rate weight, and request volume weight per unit time, respectively.

[0054] Optionally, the service resource dependency acquisition module 20 is specifically used to: obtain the corresponding resource consumption coefficients according to the service resource dependencies; obtain the root cause score through the resource consumption coefficients; sort all the root cause scores in descending order, and take the POD service corresponding to the largest root cause score as the fault root cause service.

[0055] Optionally, the service resource dependency acquisition module 20 is specifically used to: obtain the corresponding resource consumption coefficients based on the service resource dependencies using attenuation coefficients; wherein, the resource consumption coefficients include: ,in, The resource consumption coefficient for serving m. For service m, a collection of downstream services are called. For service m to call the downstream service n, the service resource dependencies are defined. The attenuation coefficient is... The resource consumption coefficient for serving downstream n. For local resource exceptions for service m.

[0056] Optionally, the service resource dependency acquisition module 20 is specifically used to: obtain the root cause score through the resource consumption coefficient; wherein, the root cause score includes: ,in, Score the root cause. The resource consumption coefficient for serving m. This represents downstream dependence.

[0057] Optionally, the multi-model collaborative prediction module 40 is specifically used for: inputting the peripheral device status data into multiple sub-models to obtain corresponding sub-model prediction data; inputting all the sub-model prediction data into a weighted fusion model to obtain the peripheral device prediction data; wherein, the peripheral device prediction data includes: ,in, The peripheral device prediction data at time t. Let be the weight coefficient of the i-th sub-model at time t. The predicted data for the i-th sub-model at time t.

[0058] Optionally, the data acquisition module 10 is specifically used to: use DiagService to perform real-time event monitoring on the POD cluster through the local network within the POD, and collect the cluster's operating data.

[0059] As shown in Figure 3, an electronic device 300 provided in this embodiment of the invention includes a memory 310 and a processor 320; the memory 310 is used to store a computer program; the processor 320 is used to implement the AI-based non-intrusive cluster diagnostic method as described above when the computer program is executed.

[0060] Alternatively, an electronic device 300 includes a memory 310 and a processor 320 coupled to the memory 310; the memory 310 is configured to store a computer program; the processor 320 is configured to perform the following operations when executing the computer program: real-time event monitoring of the POD cluster via the local network within the POD, collecting cluster operation data, wherein the cluster operation data includes POD service operation data and peripheral device status data; obtaining service resource dependencies between each POD service through the POD service operation data; obtaining a fault root cause service based on all the service resource dependencies; predicting the peripheral device status data based on multi-model collaborative prediction to obtain peripheral device prediction data; and generating a device fault warning when the peripheral device prediction data exceeds a device threshold.

[0061] This invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the AI-based non-intrusive cluster diagnostic method described above.

[0062] Alternatively, a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following operations: real-time event monitoring of the POD cluster via the local network within the POD, collecting cluster operation data, wherein the cluster operation data includes POD service operation data and peripheral device status data; obtaining service resource dependencies between each POD service through the POD service operation data; obtaining the root cause service based on all the service resource dependencies; predicting the peripheral device status data based on multi-model collaborative prediction to obtain peripheral device prediction data; and generating a device fault warning when the peripheral device prediction data exceeds a device threshold.

[0063] The present invention will now be described an electronic device 300 that can serve as a server or client of the present invention, which is an example of a hardware device that can be applied to various aspects of the present invention. Electronic device 300 is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 300 can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0064] Electronic device 300 includes a computing unit that can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) or a computer program loaded from a storage unit into random access memory (RAM). The RAM may also store various programs and data required for device operation. The computing unit, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0065] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc. In this application, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention according to actual needs. Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units can be implemented in hardware or as software functional units.

[0066] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A non-invasive cluster diagnostic method based on AI, characterized in that, include: Real-time event monitoring of the POD cluster is performed through the local network within the POD to collect cluster operation data, including POD service operation data and peripheral device status data; the service resource dependencies between each POD service are obtained through the POD service operation data; and the root cause service of the fault is obtained based on all the service resource dependencies. Based on multi-model collaborative prediction, the peripheral device status data is predicted to obtain peripheral device prediction data; when the peripheral device prediction data is greater than the device threshold, a device fault warning is generated.

2. The AI-based non-invasive cluster diagnostic method according to claim 1, characterized in that, The step of obtaining the service resource dependencies between each POD service through the POD service operation data includes: obtaining the service resource dependencies between each POD service through the POD service operation data; wherein, the service resource dependencies include: ,in, For service m to call the downstream service n, the service resource dependencies are defined. The average latency for service m to call downstream service n. The error rate of service m calling downstream service n. The number of requests per unit time that service m calls downstream service n. 、 、 They are respectively 、 、 The maximum value, 、 、 These are the average latency weight, error rate weight, and request volume weight per unit time, respectively.

3. The AI-based non-invasive cluster diagnostic method according to claim 1, characterized in that, The step of obtaining the root cause service based on all the service resource dependencies includes: obtaining the corresponding resource consumption coefficients based on the service resource dependencies; obtaining the root cause score through the resource consumption coefficients; sorting all the root cause scores in descending order, and taking the POD service corresponding to the root cause score with the highest score as the root cause service.

4. The AI-based non-invasive cluster diagnostic method according to claim 3, characterized in that, The step of obtaining the corresponding resource consumption coefficients based on the service resource dependencies includes: obtaining the corresponding resource consumption coefficients based on the service resource dependencies using attenuation coefficients; wherein, the resource consumption coefficients include: ,in, The resource consumption coefficient for serving m. For service m, a collection of downstream services are called. For service m to call the downstream service n, the service resource dependencies are defined. The attenuation coefficient is... The resource consumption coefficient for serving downstream n. For local resource exceptions for service m.

5. The AI-based non-invasive cluster diagnostic method according to claim 3, characterized in that, The step of obtaining the root cause score through the resource consumption coefficient includes: obtaining the root cause score through the resource consumption coefficient; wherein, the root cause score includes: ,in, Score the root cause. The resource consumption coefficient for serving m. This represents downstream dependence.

6. The AI-based non-invasive cluster diagnostic method according to claim 1, characterized in that, The step of predicting peripheral device status data to obtain peripheral device prediction data includes: inputting the peripheral device status data into multiple sub-models to obtain corresponding sub-model prediction data; and inputting all the sub-model prediction data into a weighted fusion model to obtain the peripheral device prediction data; wherein, the peripheral device prediction data includes: ,in, The peripheral device prediction data at time t. Let be the weight coefficient of the i-th sub-model at time t. The predicted data for the i-th sub-model at time t.

7. The AI-based non-invasive cluster diagnostic method according to claim 1, characterized in that, The step of monitoring real-time events of the POD cluster and collecting cluster operation data through the local network within the POD includes: using DiagService to monitor real-time events of the POD cluster and collect cluster operation data through the local network within the POD.

8. A non-invasive cluster diagnostic device based on AI, characterized in that, include: The data acquisition module is used to monitor real-time events of the POD cluster through the local network within the POD and collect cluster operation data, including POD service operation data and peripheral device status data. The service resource dependency acquisition module is used to obtain the service resource dependencies between each POD service through the POD service operation data. The fault root cause service acquisition module is used to obtain the fault root cause service based on all the service resource dependencies. The multi-model collaborative prediction module is used to predict the peripheral device status data based on multi-model collaborative prediction to obtain peripheral device prediction data. The device fault warning generation module is used to generate a device fault warning when the peripheral device prediction data exceeds a device threshold.

9. An electronic device, characterized in that, It includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the AI-based non-intrusive cluster diagnostic method as described in any one of claims 1 to 7 when the computer program is executed.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the AI-based non-intrusive cluster diagnostic method as described in any one of claims 1 to 7.

Images