Server fault repairing method and device, host equipment, system and medium

Through the pre-trained language model, the fault prediction and automatic generation and repair operations of server operation data is solved, and the problem of single fault prediction function and manual intervention in the existing technology is solved, and the automation and flexibility of server fault repair is improved.

CN119065945APending Publication Date: 2024-12-03SHANDONG YUNHAI GUOCHUANG CLOUD COMPUTING EQUIP IND INNOVATION CENT CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411157036.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-22
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing server fault prediction function is single, and requires manual intervention to deal with the fault, which increases maintenance complexity and cost.

Method used

The pre-trained language model is used to predict the server's running data and automatically generate fault repair operations to achieve automatic fault repair.

Benefits of technology

Improves the flexibility and automation of server failure repair, reduces manual intervention, and reduces maintenance costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119065945A_ABST
    Figure CN119065945A_ABST
Patent Text Reader

Abstract

The invention provides a server fault repairing method and device, host equipment, a system and a medium, and relates to the field of servers, the method comprises the following steps: obtaining equipment model information and to-be-detected operation data of a server, and integrating the equipment model information and the to-be-detected operation data into an input text; inputting the input text into a pre-training language model to obtain a target fault type and a target fault repair operation which are output by the pre-training language model and correspond to the server; wherein the pre-training language model is trained by using a first preset corresponding relationship between the operation data of each model of server and the fault type and a second preset corresponding relationship between the fault type and the fault repair operation; performing automatic fault repair on the server according to the target fault repair operation; according to the method, the pre-training language model can be used for performing fault prediction on the dynamically changing server operation data, and the model can be used for automatically generating the fault repair operation, so that the flexibility of server fault repair can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of servers, and in particular to a server fault repair method, device, host equipment, system and medium. Background Art

[0002] The server is a core component of the information system infrastructure. The server motherboard is equipped with a baseboard management controller (BMC), which can monitor the server's voltage, temperature and other physical conditions in real time, and realize remote management, installation and restart of the server. However, the fault prediction function of the baseboard management controller is still relatively simple. In addition, once a server fails, it still needs to rely on manual intervention to handle it, which greatly increases the complexity and cost of maintenance. Summary of the invention

[0003] The purpose of the present invention is to provide a server fault repair method, apparatus, host device, system and medium, which can use a pre-trained language model to predict faults for dynamically changing server operation data, and can use the model to automatically generate fault repair operations, thereby improving the flexibility of server fault repair.

[0004] In order to solve the above technical problems, the present invention provides a server fault repair method, comprising:

[0005] Obtaining device model information and the operation data to be detected from the server, and integrating the device model information and the operation data to be detected into input text;

[0006] Inputting the input text into a pre-trained language model to obtain a target fault type and a target fault repair operation corresponding to the server output by the pre-trained language model; wherein the pre-trained language model is trained using a first preset correspondence between operation data of each model of server and a fault type, and a second preset correspondence between a fault type and a fault repair operation;

[0007] Automatically perform fault repair on the server according to the target fault repair operation.

[0008] Optionally, the pre-trained language model is deployed on a model deployment device;

[0009] Inputting the input text into a pre-trained language model to obtain a target fault type and a target fault repair operation corresponding to the server output by the pre-trained language model, including:

[0010] Sending the input text to the model deployment device so that the model deployment device inputs the input text into the pre-trained language model and the model deployment device returns the target fault type and the target fault repair operation;

[0011] Receive the target fault type and the target fault repair operation returned by the model deployment device.

[0012] Optionally, the process of determining the target fault type and the target fault repair operation by the pre-trained language model includes:

[0013] The pre-trained language model determines the predicted probability of the server encountering various types of failures based on the input text;

[0014] Determining whether the server will fail according to the predicted probability;

[0015] If it is determined that the server will fail, the failure type with the highest predicted probability is determined as the target failure type, and the target failure repair operation is determined according to the target failure type.

[0016] Optionally, judging whether the server will fail according to the predicted probability includes:

[0017] Determining an overall predicted probability that the server will fail using the predicted probabilities of various failure types;

[0018] Determining whether the total predicted probability is greater than a preset threshold;

[0019] If the total predicted probability is greater than the preset threshold, it is determined that the server will fail;

[0020] If the total predicted probability is not greater than the preset threshold, it is determined that the server will not fail.

[0021] Optionally, before automatically repairing the server according to the target fault repair operation, the method further includes:

[0022] Determining an operation type of the target fault repair operation;

[0023] If the target fault repair operation is a manual operation, outputting text information of the target fault repair operation to a display device;

[0024] If the target fault repair operation is a code execution operation, the step of automatically repairing the server according to the target fault repair operation is entered.

[0025] Optionally, after automatically repairing the server according to the target fault repair operation, the method further includes:

[0026] Integrate the execution result of the target fault repair operation and the target fault type into an output text;

[0027] The output text is output to a display device.

[0028] Optionally, after automatically repairing the server according to the target fault repair operation, the method further includes:

[0029] Acquire current operation data generated by the server after completing automatic fault repair;

[0030] Inputting the input text including the device model information and the current operation data into a pre-trained language model, and judging whether the fault of the server is successfully repaired according to the output of the pre-trained language model;

[0031] If the fault of the server is not successfully repaired, the device model information and the to-be-detected operation data are integrated into manually processed data, the manually processed data is saved in a preset database, and preset alarm information is output;

[0032] Determining whether the manually processed data in the preset database adds a fault type and a fault repair operation;

[0033] If the manually processed data has been added with fault types and fault repair operations, the pre-trained language model is trained using the manually processed data and the fault types and fault repair operations added to the manually processed data.

[0034] Optionally, after the server is repaired according to the fault repair operation, the method further includes:

[0035] If the fault of the server has been successfully repaired, the pre-trained language model is trained using the device model information, the operation data to be detected, and the target fault type.

[0036] The present invention also provides a server fault repair device, comprising:

[0037] An acquisition module, used for acquiring device model information and the operation data to be detected from the server, and integrating the device model information and the operation data to be detected into input text;

[0038] A model processing module, used for inputting the input text into a pre-trained language model to obtain the target fault type and target fault repair operation corresponding to the server output by the pre-trained language model; wherein the pre-trained language model is trained using a first preset correspondence between the operating data of each model of server and the fault type, and a second preset correspondence between the fault type and the fault repair operation;

[0039] A fault repair module is used to automatically repair the server according to the target fault repair operation.

[0040] The present invention also provides a host device, comprising:

[0041] Memory for storing computer programs;

[0042] A processor is used to implement the server fault repair method as described above when executing the computer program.

[0043] Optionally, the host device is a baseboard management controller.

[0044] The present invention also provides a server fault repair system, comprising:

[0045] A host device, used to execute the server failure repair method as described above;

[0046] A model deployment device, in which a pre-trained language model is deployed, is used to receive input text sent by the host device, input the input text into the pre-trained language model, and return the target fault type and target fault repair operation output by the pre-trained language model to the host device.

[0047] The present invention also provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are loaded and executed by a processor, the server fault repair method as described above is implemented.

[0048] The present invention provides a server fault repair method, comprising: obtaining device model information and operating data to be detected of a server, and integrating the device model information and the operating data to be detected into an input text; inputting the input text into a pre-trained language model to obtain a target fault type and a target fault repair operation corresponding to the server output by the pre-trained language model; wherein the pre-trained language model is trained using a first preset corresponding relationship between operating data and fault types of servers of various models, and a second preset corresponding relationship between fault types and fault repair operations; and automatically repairing the server according to the target fault repair operation.

[0049] It can be seen that the present invention can first obtain the device model information and the operating data to be detected of the server, and integrate the device model information and the operating data to be detected into input text. Subsequently, the present invention can input the input text into a pre-trained language model, use the model to perform fault detection on the operating data to be detected of the server of this model, and obtain the target fault type and target fault repair operation corresponding to the server output by the pre-trained language model; wherein the pre-trained language model is trained using the first preset correspondence between the operating data and the fault type of each model server, and the second preset correspondence between the fault type and the fault repair operation. In this way, the present invention can not only use the pre-trained language model to predict faults for dynamically changing server operating data, thereby improving the server prediction performance, but also use the pre-trained language model to automatically generate fault repair operations, so that the server can be automatically repaired according to the target fault repair operation, so as to improve the flexibility of server fault repair. The present invention also provides a server fault repair device, host device, system and medium, which have the above-mentioned beneficial effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention 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 only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0051] Figure 1 A flow chart of a server fault repair method provided by an embodiment of the present invention;

[0052] Figure 2 A flowchart of intelligent fault prediction and analysis provided by an embodiment of the present invention;

[0053] Figure 3 A flowchart of a natural language interaction provided by an embodiment of the present invention;

[0054] Figure 4 A structural block diagram of a server intelligent management system provided by an embodiment of the present invention;

[0055] Figure 5 A structural block diagram of a server fault repair device provided by an embodiment of the present invention;

[0056] Figure 6 A structural block diagram of a host device provided by an embodiment of the present invention;

[0057] Figure 7 A structural block diagram of a server fault recovery system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] The server is a core component in the information system infrastructure. The server motherboard is equipped with a baseboard management controller (BMC), which can monitor the physical status of the server such as voltage and temperature in real time, and realize functions such as remote management, installation and restart of the server. However, the fault prediction function of the current baseboard management controller is still relatively simple. In addition, once a server fails, it still needs to rely on manual intervention to handle it, which greatly increases the complexity and cost of maintenance. In view of this, the present invention can provide a server fault repair method, which can use a pre-trained language model to predict faults for dynamically changing server operation data, and can use the model to automatically generate fault repair operations, thereby improving the flexibility of server fault repair.

[0060] It should be noted that the embodiment of the present invention does not limit the hardware device for executing the present method, and can be set according to actual application requirements. For example, the above-mentioned hardware device can be a personal computer, a dedicated server dedicated to controlling other servers, or other dedicated devices. In addition, there can be multiple hardware devices for executing the present method, and these multiple hardware devices can be of different types, and these multiple hardware devices can form a system. For example, the present method can be jointly executed by a host device and a model deployment device, and the host device is used to obtain the data to be detected from the server and input the data to be detected into the model deployment device; and is also used to perform corresponding operations according to the detection results returned by the model deployment device. The model deployment device is deployed with a pre-trained language model for detecting the data input by the host device and returning the detection results to the host device. The embodiment of the present invention also does not limit the specific types of the above-mentioned host device and model deployment device, and can be set according to actual application requirements. For example, the above-mentioned host device can be a baseboard management controller, and the model deployment device can be a personal computer, a server, etc.

[0061] For easier understanding, please refer to Figure 1 , Figure 1 A flow chart of a server fault repair method provided by an embodiment of the present invention, the method may include:

[0062] S101, obtaining device model information and operation data to be detected from a server, and integrating the device model information and the operation data to be detected into input text.

[0063] In the embodiment of the present invention, the device model information is used to record the device model of the server, and the operation data to be detected is used to record the operation status of the server. In the embodiment of the present invention, the device model information and the operation data to be detected can be integrated into input text to input the input text into the pre-trained language model for detection.

[0064] It should be noted that the embodiments of the present invention do not limit the specific device model information. The information may only record the device model of the server itself, or may further include the device models of various components in the server (such as the central processing unit, memory, hard disk, graphics card, etc.). It is understandable that the greater the amount of information contained in the device model information, the more helpful it is for the analysis and detection of the pre-trained language model.

[0065] Furthermore, the embodiments of the present invention do not limit the collection method of the operation data to be detected. For example, the operation data to be detected may be the operation data generated by the server at a certain moment, that is, only the operation data at that moment is collected; the operation data to be detected may be the operation data generated by the server within a certain time period, that is, all the operation data generated within the time period is collected. In order to facilitate the detection of the pre-trained language model, the operation data to be detected should contain a sufficient amount of information. Therefore, the embodiments of the present invention can use all the operation data generated by the server within a preset time period as the operation data to be detected. The embodiments of the present invention do not limit the length of the preset time period, and can be set according to actual application requirements.

[0066] Furthermore, the embodiments of the present invention do not limit how to integrate the device model information and the running data to be detected into the input text, and can be set according to the actual application requirements. For example, in order to facilitate the pre-trained language model to understand the contextual meaning of the device model information and the running data to be detected, and to understand the user's intention, an input text template can be preset, and the input text template contains a filling position for the device model information, a filling position for the running data to be detected, and a preset instruction text, and the preset instruction text is used to indicate how the pre-trained language model performs the detection. Furthermore, it is only necessary to add the device model information and the running data to be detected to the corresponding filling position of the input text template. The embodiments of the present invention also do not restrict the text content of the preset instruction text, and can be set according to the actual application requirements. For example, the text content of the preset instruction text can be used to indicate the information type of the device model information and the running data to be detected, so that the pre-trained language model can analyze the device model information and the running data to be detected; the text content can also be used to indicate the content to be analyzed by the pre-trained language model (such as the fault type and fault repair operation corresponding to the analysis server); the text content can also be used to indicate the pre-trained language model in which preset format to output the detection result.

[0067] S102. Input the input text into a pre-trained language model to obtain a target fault type and a target fault repair operation corresponding to the server output by the pre-trained language model; wherein the pre-trained language model is trained using a first preset correspondence between operating data of each model of server and the fault type, and a second preset correspondence between the fault type and the fault repair operation.

[0068] In this step, the input text obtained can be input into the pre-trained language model so that the pre-trained language model can detect and analyze the input text to obtain the target fault type and target fault repair operation corresponding to the server. In order to ensure that the pre-trained language model has detection capabilities, the embodiment of the present invention can use the first preset correspondence between the operating data of each model of server and the fault type, and the second preset correspondence between the fault type and the fault repair operation as training data, and use the training data to train the model to embed the above two preset correspondences into the model parameters of the model. In this way, the pre-trained language model can detect the input text, determine the target fault type corresponding to the server and the target fault repair operation required to repair the fault of the target fault type.

[0069] It should be noted that the embodiments of the present invention do not limit specific fault types or specific fault repair operations, and can be set according to actual application requirements. It is worth noting that the fault repair operations here can include only operations automatically executed by the host device, such as instructions, script files, etc., and can also include operations that need to be manually executed by operation and maintenance personnel, and can be set according to actual application requirements.

[0070] It should also be noted that the embodiments of the present invention do not limit the model structure of the pre-trained language model, nor the training process of the model, and reference may be made to relevant techniques of machine learning. In addition, the pre-trained language model may also be fine-tuned based on an existing open source language model (such as various open source large language models LLM, Large Language Model).

[0071] Furthermore, it is understandable that the above-mentioned operating data to be detected may not indicate that there is a fault in the server, or may only indicate that there is a minor fault in the server. In the case where there is no fault or there is a minor fault, it does not make sense to perform a fault repair operation on the server, or it does not make much sense. Therefore, the pre-trained language model in the embodiment of the present invention can also determine the above-mentioned target fault type and target fault operation in two steps. First, the pre-trained language model can determine the predicted probability of the server encountering various fault types based on the input text; then, the model can determine whether the server will fail based on the predicted probability. If it is determined that the server will fail, the fault type with the highest predicted probability can be determined as the target fault type, and the target fault repair operation can be determined based on the target fault type. Of course, if it is determined that the server will not fail, the step of determining the target fault repair operation may not be performed.

[0072] Based on this, the process of determining the target fault type and the target fault repair operation by the pre-trained language model may include:

[0073] Step 11: The pre-trained language model determines the predicted probability of the server encountering various types of failures based on the input text.

[0074] Step 12: Determine whether the server will fail based on the predicted probability; if so, proceed to step 13; if not, proceed to step 14.

[0075] Step 13: Determine the fault type with the highest predicted probability as the target fault type, and determine the target fault repair operation according to the target fault type.

[0076] Step 14: Return the detection result indicating that the server is not faulty.

[0077] It should be noted that the embodiments of the present invention do not limit how to determine whether a server will fail based on the predicted probabilities of the server encountering various types of failures. For example, it can be determined whether each predicted probability exceeds a preset threshold. If so, it is determined that the server will fail, otherwise it is determined that the server will not fail. The predicted probabilities of various types of failures can also be used to determine the total predicted probability that the server will fail, and determine whether the total predicted probability exceeds a preset threshold. If so, it is determined that the server will fail, otherwise it is determined that the server will not fail. In order to improve comprehensiveness, the embodiments of the present invention will determine whether a server will fail based on the total predicted probability.

[0078] Based on this, judging whether the server will fail according to the predicted probability may include:

[0079] Step 21: Determine the total predicted probability that the server will fail using the predicted probabilities of various failure types;

[0080] Step 22: Determine whether the total predicted probability is greater than a preset threshold; if so, proceed to step 23; if not, proceed to step 24;

[0081] Step 23: Determine that the server will fail;

[0082] Step 24: Determine whether the server will fail.

[0083] It should be noted that the embodiment of the present invention does not limit the specific value of the preset threshold, which can be set according to actual application requirements, for example, it can be 50%.

[0084] S103: Automatically perform fault repair on the server according to the target fault repair operation.

[0085] After obtaining the target fault repair operation output by the pre-trained language model, it is only necessary to automatically execute the target fault repair operation to automatically repair the existing server faults, thereby improving the flexibility of server fault detection and repair.

[0086] Of course, as mentioned above, the fault repair operations learned by the pre-trained language model can include only operations that need to be performed by the host device, or manual operations that need to be performed manually by the operation and maintenance personnel. Therefore, before executing the target fault repair operation, the type of the target fault repair operation can also be determined. If the target fault repair operation is a manual operation, the text information of the target fault repair operation is output to the display device for relevant personnel to view; if the target fault repair operation is a code execution operation, the server is automatically repaired according to the target fault repair operation.

[0087] Based on this, before automatically repairing the server according to the target fault repair operation, the following may be included:

[0088] Step 31: Determine the operation type of the target fault repair operation;

[0089] Step 32: if the target fault repair operation is a manual operation, outputting text information of the target fault repair operation to a display device;

[0090] Step 33: If the target fault repair operation is a code execution operation, then enter the step of automatically repairing the server according to the target fault repair operation.

[0091] Furthermore, after the target fault repair operation is executed, the execution result of the target fault repair operation and the target fault type may be integrated into an output text, and the output text may be output to a display device for viewing by operation and maintenance personnel.

[0092] Based on this, after automatically repairing the server according to the fault repair operation, the method may further include:

[0093] Step 41: Integrate the execution result of the target fault repair operation and the target fault type into an output text;

[0094] Step 42: Output the output text to a display device.

[0095] Based on the above embodiments, the present invention can first obtain the device model information and the operating data to be detected of the server, and integrate the device model information and the operating data to be detected into input text. Subsequently, the present invention can input the input text into a pre-trained language model, and use the model to perform fault detection on the operating data to be detected of the server of this model, and obtain the target fault type and target fault repair operation corresponding to the server output by the pre-trained language model; wherein the pre-trained language model is trained using the first preset correspondence between the operating data and the fault type of each model server, and the second preset correspondence between the fault type and the fault repair operation. In this way, the present invention can not only use the pre-trained language model to predict faults for dynamically changing server operating data, thereby improving server prediction performance, but also use the pre-trained language model to automatically generate fault repair operations, so that the server can be automatically repaired according to the target fault repair operation, so as to improve the flexibility of server fault repair.

[0096] Based on the above embodiments, it can be understood that the operation data of the server has dynamic and changeable characteristics, so the conventional training data collection method is difficult to cover various situations of server operation. Therefore, there must be unprocessed operation data that the pre-trained language model has not been exposed to. At this time, the model may have problems such as fault type analysis errors and fault repair operation determination errors, which may cause the server fault to be unable to be repaired. In order to further enhance the processing capability of the pre-trained language model, the embodiment of the present invention also provides a feedback mechanism, which can automatically collect and manually process and manually annotate the operation data to be detected that is incorrectly detected by the pre-trained language model, and can use the newly annotated operation data to be detected to train the pre-trained language model. The specific steps of the above feedback mechanism are introduced below.

[0097] Based on this, after automatically repairing the server according to the target fault repair operation, the method may further include:

[0098] S201. Acquire current operation data generated by the server after completing automatic fault repair.

[0099] S202: Inputting the input text including the device model information and the current operation data into a pre-trained language model, and judging whether the fault of the server is successfully repaired according to the output of the pre-trained language model.

[0100] Steps S201 and S202 are used to determine whether the server fault has been successfully repaired. The host device can collect the current operating data generated by the server after executing the target fault repair operation, and input the input text containing the current operating data into the pre-trained language model to determine whether the target fault type output by the model is the previous type. If it is not the previous fault type, it can be determined that the server fault has been repaired; if it is the previous type, it can be determined that the server fault has not been successfully repaired. For the operating data to be detected corresponding to the unsuccessful repair of the server fault, it can be determined that it is operating data that cannot be correctly analyzed by the pre-trained language model and needs to be manually processed again.

[0101] S203: If the fault of the server is not successfully repaired, the device model information and the to-be-detected operating data are integrated into manually processed data, the manually processed data is saved in a preset database, and preset alarm information is output.

[0102] In this step, a preset database can be pre-set to save the device model information and the running data to be tested that failed to be processed by the pre-trained language model. In this step, such device model information and the running data to be tested can be integrated into manually processed data, and the manually processed data can be saved in the preset database, and preset alarm information can be output to prompt the operation and maintenance personnel to handle it.

[0103] S204: Determine whether the manually processed data in the preset database has added a fault type and a fault repair operation.

[0104] In this step, after adding the above-mentioned manually processed data to the preset database, the host device can determine whether the manually processed data in the database has manually added fault types and fault repair operations. If so, the pre-trained language model can be trained using the perfected manually processed data to enhance the performance of the model.

[0105] S205: If the manually processed data has been added with fault types and fault repair operations, the pre-trained language model is trained using the manually processed data and the fault types and fault repair operations added to the manually processed data.

[0106] After steps S201 to S205, the present invention can automatically collect device model information and operating data to be detected that cannot be processed by the pre-trained language model, and can input them into a preset database for manual labeling. The pre-trained language model can also be further trained using the data after manual labeling, thereby facilitating the improvement of the detection performance of the model.

[0107] Of course, for the device model information and the operating data to be detected that are correctly processed by the pre-trained language model, in order to enhance the model's learning effect on such data, the embodiment of the present invention can also use the device model information and the operating data to be detected that have been correctly processed by the pre-trained language model and the target fault types corresponding to the two to train the pre-trained language model. In this way, the effect of improving the model learning effect and increasing the amount of model training data can also be achieved.

[0108] Based on this, after the server is repaired according to the fault repair operation, the method may further include:

[0109] S207: If the fault of the server has been successfully repaired, the pre-trained language model is trained using the device model information, the operation data to be detected, and the target fault type.

[0110] Based on the above embodiments, considering that the host device may have less computing power and is insufficient to deploy the pre-trained language model, the embodiment of the present invention can also add a model deployment device dedicated to deploying the pre-trained language model. In this way, the host device only needs to send the input text to the model deployment device, and perform corresponding processing according to the target fault type and target fault repair operation returned by the model deployment device, which can give full play to the characteristics of different hardware devices.

[0111] Based on this, the pre-trained language model is deployed on a model deployment device;

[0112] Inputting the input text into a pre-trained language model to obtain a target fault type and a target fault repair operation corresponding to the server output by the pre-trained language model may include:

[0113] S301, sending the input text to the model deployment device, so that the model deployment device inputs the input text into the pre-trained language model, and the model deployment device returns the target fault type and the target fault repair operation;

[0114] S302: Receive the target fault type and the target fault repair operation returned by the model deployment device.

[0115] Based on the above embodiments, the server fault repair method provided by the present invention is fully introduced below based on the schematic diagram. The present invention provides a server intelligent management method based on a pre-trained language model (such as a large language model), such as Figure 2 (Intelligent Fault Prediction and Analysis) and Figure 3 (Natural Language Interaction) shown.

[0116] The server intelligent management method based on the pre-trained language model of this embodiment includes two aspects:

[0117] 1. Intelligent Fault Prediction and Analysis

[0118] The process of intelligent fault prediction and analysis is as follows: Figure 2 As shown, the specific process is as follows:

[0119] (1) Obtain basic information of the server through the BMC chip, including the server's device information and real-time status information. The device information includes the server's entire machine information and information about each internal component (such as the CPU model, memory capacity, etc.), and the real-time status information includes temperature, voltage, CPU utilization, memory occupancy, etc.;

[0120] (2) The BMC chip passes the above information to the model deployment hardware unit (the hardware on which the pre-trained language model is deployed) as model input;

[0121] (3) The pre-trained language model outputs the fault prediction probability P of the server in the current state based on the above information; where the fault prediction probability P = 1-(1-P1)*(1-P2)*…*(1-P n ), where P i is the probability of the i-th type of failure occurring in the server (i=1,2,…,n).

[0122] (4) Determine whether the fault prediction probability P is greater than 50%. If P is less than 50%, it is considered that the server will not fail, and go to (1) to continue the next cycle. If P is greater than 50%, it is considered that the server will fail, and go to (5) to perform intelligent fault analysis and processing.

[0123] (5) Pre-train the language model for intelligent fault analysis and send the most likely fault type (i.e., P i the fault type with the largest value), and explain the possible cause of the fault;

[0124] (6) The BMC displays the fault type and possible fault cause on the BMC user interface, executes the corresponding fault handling procedure according to the fault type, and then goes to (1) to continue the next cycle.

[0125] 2. Natural Language Interaction

[0126] Users can manage servers through natural language, such as Figure 3 The specific process is as follows:

[0127] (1) The user uses natural language to describe the server information or operations to be performed in the dialog box of the BMC software interface. The interactive scheduling unit transmits the above natural language information to the model deployment device;

[0128] (2) The model deployment device generates corresponding information based on the input of the above natural language information and returns it to the interactive scheduling management unit.

[0129] (3) The interactive scheduling unit determines whether the information returned by the pre-trained language model contains program code (such as query code, shutdown code, or restart code, etc.). If it does not contain code, jump to (6). If it contains code, jump to (4).

[0130] (4) The interactive scheduling unit passes the program code to the execution unit in the BMC for execution, and the execution unit then returns the program running results to the interactive scheduling unit.

[0131] (5) The interactive scheduling unit combines the information returned by the large model in (2) and the operation results returned in (4) into one message.

[0132] (6) The interactive scheduling unit sends the information to the BMC, which displays the above information on the BMC software interface.

[0133] Please refer to Figure 4 , Figure 4 This is a structural block diagram of a server intelligent management system provided by an embodiment of the present invention. Figure 4As shown in the figure, the BMC chip is connected to other component vectors in the server through interfaces such as PCIe (Peripheral Component Interconnect express, a high-speed serial computer expansion bus standard), I3C (Improved Inter Integrated Circuit, an improved I2C bus), PECI (Platform Environment Control Interface, a platform environment control interface), and ADC (Analog to Digital Converte, an analog-to-digital converter), and obtains basic information and real-time operation status information of the server through these interfaces. The BMC is connected to the model deployment device through the bus. The interactive scheduling unit in the BMC is responsible for scheduling the intelligent interactive management function, including sending the natural language input information from the user to the large language model hardware; judging whether the generated content of the large language model contains program code. If it contains program code, the program code needs to be handed over to the execution unit for execution, and the program running results from the execution unit are received; the generated content of the large model and the program running results are combined and sent to the display unit, so as to be displayed in the software user interface.

[0134] The server fault repair device, host device, server fault repair system, computer program product, and computer-readable storage medium provided in an embodiment of the present invention are introduced below. The server fault repair device, host device, server fault repair system, computer program product, and computer-readable storage medium described below can be referenced to each other with the server fault repair method described above.

[0135] Please refer to Figure 5 , Figure 5 A structural block diagram of a server fault repair device provided by an embodiment of the present invention, the device may include:

[0136] The acquisition module 501 is used to acquire the device model information and the operation data to be detected from the server, and integrate the device model information and the operation data to be detected into input text;

[0137] The model processing module 502 is used to input the input text into a pre-trained language model to obtain the target fault type and target fault repair operation corresponding to the server output by the pre-trained language model; wherein the pre-trained language model is trained using a first preset correspondence between the operating data of each model of the server and the fault type, and a second preset correspondence between the fault type and the fault repair operation;

[0138] The fault repair module 503 is used to automatically repair the server according to the target fault repair operation.

[0139] Optionally, the pre-trained language model is deployed on a model deployment device;

[0140] The model processing module 502 includes:

[0141] a sending submodule, configured to send the input text to the model deployment device, so that the model deployment device inputs the input text into the pre-trained language model, and the model deployment device returns the target fault type and the target fault repair operation;

[0142] The receiving submodule is used to receive the target fault type and the target fault repair operation returned by the model deployment device.

[0143] Optionally, the model processing module 502 may include:

[0144] A prediction probability determination submodule, used for the pre-trained language model to determine the prediction probability of the server encountering various types of failures according to the input text;

[0145] A fault judgment submodule, used to judge whether the server will fail according to the predicted probability;

[0146] The fault repair operation determination submodule is used to determine the fault type with the highest predicted probability as the target fault type if it is determined that the server will fail, and determine the target fault repair operation according to the target fault type.

[0147] Optionally, the fault judgment submodule may include:

[0148] a total predicted probability calculation unit, configured to determine a total predicted probability that the server will fail using predicted probabilities of various failure types;

[0149] A judgment unit is used to judge whether the total predicted probability is greater than a preset threshold; if the total predicted probability is greater than the preset threshold, it is determined that the server will fail; if the total predicted probability is not greater than the preset threshold, it is determined that the server will not fail.

[0150] Optionally, the fault recovery module 503 may include:

[0151] An operation type determination submodule, used to determine the operation type of the target fault repair operation;

[0152] an output submodule, configured to output text information of the target fault repair operation to a display device if the target fault repair operation is a manual operation;

[0153] The fault repair submodule is used to enter the step of automatically repairing the server according to the target fault repair operation if the target fault repair operation is a code execution operation.

[0154] Optionally, the device may further include:

[0155] An output text integration module, used to integrate the execution result of the target fault repair operation and the target fault type into an output text;

[0156] The output text output module is used to output the output text to a display device.

[0157] Optionally, the device may further include:

[0158] A current operation data acquisition module is used to acquire the current operation data generated by the server after completing automatic fault repair;

[0159] A repair verification module, used for inputting the input text containing the device model information and the current operation data into the pre-trained language model, and judging whether the fault of the server is successfully repaired according to the output of the pre-trained language model;

[0160] A recording module, configured to integrate the device model information and the to-be-detected operating data into manually processed data if the fault of the server is not successfully repaired, save the manually processed data into a preset database, and output preset alarm information;

[0161] A marking judgment module, used to judge whether the manually processed data in the preset database has added a fault type and a fault repair operation;

[0162] The first training module is used to train the pre-trained language model using the artificially processed data and the fault type and fault repair operation added to the artificially processed data if the fault type and fault repair operation have been added to the artificially processed data.

[0163] Optionally, the device may further include:

[0164] The second training module is used to train the pre-trained language model using the device model information, the operating data to be detected, and the target fault type if the fault of the server has been successfully repaired.

[0165] Please refer to Figure 6 , Figure 6This is a structural block diagram of a host device provided in an embodiment of the present invention. The embodiment of the present invention provides a host device 10, including a processor 11 and a memory 12; wherein the memory 12 is used to store a computer program; and the processor 11 is used to execute the server fault repair method provided in the aforementioned embodiment when executing the computer program.

[0166] For the specific process of the above-mentioned server fault repair method, please refer to the corresponding content provided in the above-mentioned embodiment, which will not be repeated here.

[0167] Furthermore, the memory 12 as a carrier for storing resources may be a read-only memory, a random access memory, a magnetic disk or an optical disk, etc., and the storage method may be temporary storage or permanent storage.

[0168] In addition, the host device 10 also includes a power supply 13, a communication interface 14, an input / output interface 15 and a communication bus 16; wherein the power supply 13 is used to provide working voltage for each hardware device on the host device 10; the communication interface 14 can create a data transmission channel between the host device 10 and an external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present invention, and is not specifically limited here; the input / output interface 15 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0169] Optionally, the host device may be a baseboard management controller.

[0170] Please refer to Figure 7 , Figure 7 A structural block diagram of a server fault recovery system provided by an embodiment of the present invention, the system may include:

[0171] A host device 10, configured to execute the server failure repair method as described above;

[0172] A model deployment device 20, in which a pre-trained language model is deployed, is used to receive input text sent by the host device, input the input text into the pre-trained language model, and return the target fault type and target fault repair operation output by the pre-trained language model to the host device.

[0173] Optionally, the host device may be a baseboard management controller.

[0174] Since the embodiments of the server fault repair system part correspond to the embodiments of the server fault repair method part, the embodiments of this part refer to the description of the embodiments of the server fault repair method part, which will not be repeated here.

[0175] An embodiment of the present invention further provides a computer program product, including a computer program / instruction, which implements the server fault repair method described in the above embodiment when executed by a processor.

[0176] Since the embodiments of the computer program product part correspond to the embodiments of the server fault repair method part, please refer to the description of the embodiments of the server fault repair method part for the embodiments of the computer program product part, which will not be repeated here.

[0177] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the server fault repair method described in the above embodiment is implemented.

[0178] Since the embodiments of the computer-readable storage medium part correspond to the embodiments of the server fault repair method part, please refer to the description of the embodiments of the server fault repair method part for the embodiments of the storage medium part, which will not be repeated here.

[0179] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0180] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0181] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0182] The above is a detailed introduction to a server fault repair method, device, host device, system and medium provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, the present invention can also be improved and modified, and these improvements and modifications also fall within the scope of protection of the present invention.

Claims

1. A method for repairing a server failure. It is characterized in that include: Obtaining device model information and the operation data to be detected from the server, and integrating the device model information and the operation data to be detected into input text; Inputting the input text into a pre-trained language model to obtain a target fault type and a target fault repair operation corresponding to the server output by the pre-trained language model; wherein the pre-trained language model is trained using a first preset correspondence between operation data of each model of server and a fault type, and a second preset correspondence between a fault type and a fault repair operation; Automatically perform fault repair on the server according to the target fault repair operation.

2. The server failure repair method according to claim 1, It is characterized in that The pre-trained language model is deployed on a model deployment device; The step of inputting the input text into a pre-trained language model to obtain a target fault type and a target fault repair operation corresponding to the server output by the pre-trained language model includes: Sending the input text to the model deployment device so that the model deployment device inputs the input text into the pre-trained language model and the model deployment device returns the target fault type and the target fault repair operation; Receive the target fault type and the target fault repair operation returned by the model deployment device.

3. The server failure repair method according to claim 1, It is characterized in that The process of determining the target fault type and the target fault repair operation by the pre-trained language model includes: The pre-trained language model determines the predicted probability of the server encountering various types of failures based on the input text; Determining whether the server will fail according to the predicted probability; If it is determined that the server will fail, the failure type with the highest predicted probability is determined as the target failure type, and the target failure repair operation is determined according to the target failure type.

4. The server failure repair method according to claim 3, It is characterized in that The determining, according to the predicted probability, whether the server will fail includes: Determining an overall predicted probability that the server will fail using the predicted probabilities of various failure types; Determining whether the total predicted probability is greater than a preset threshold; If the total predicted probability is greater than the preset threshold, it is determined that the server will fail; If the total predicted probability is not greater than the preset threshold, it is determined that the server will not fail.

5. The server failure repair method according to claim 1, It is characterized in that Before automatically repairing the server according to the target fault repair operation, the method includes: Determining an operation type of the target fault repair operation; If the target fault repair operation is a manual operation, outputting text information of the target fault repair operation to a display device; If the target fault repair operation is a code execution operation, the step of automatically repairing the server according to the target fault repair operation is entered.

6. The server failure repair method according to claim 1, It is characterized in that After automatically repairing the server according to the target fault repair operation, the method further includes: Integrate the execution result of the target fault repair operation and the target fault type into an output text; The output text is output to a display device.

7. The server fault repair method according to any one of claims 1 to 6, It is characterized in that After automatically repairing the server according to the target fault repair operation, the method further includes: Acquire current operation data generated by the server after completing automatic fault repair; Inputting the input text including the device model information and the current operation data into a pre-trained language model, and judging whether the fault of the server is successfully repaired according to the output of the pre-trained language model; If the fault of the server is not successfully repaired, the device model information and the to-be-detected operation data are integrated into manually processed data, the manually processed data is saved in a preset database, and preset alarm information is output; Determining whether the manually processed data in the preset database adds a fault type and a fault repair operation; If the manually processed data has been added with fault types and fault repair operations, the pre-trained language model is trained using the manually processed data and the fault types and fault repair operations added to the manually processed data.

8. A server fault repair device, It is characterized in that include: An acquisition module, used for acquiring device model information and the operation data to be detected from the server, and integrating the device model information and the operation data to be detected into input text; A model processing module, used for inputting the input text into a pre-trained language model to obtain the target fault type and target fault repair operation corresponding to the server output by the pre-trained language model; wherein the pre-trained language model is trained using a first preset correspondence between the operating data of each model of server and the fault type, and a second preset correspondence between the fault type and the fault repair operation; A fault repair module is used to automatically repair the server according to the target fault repair operation.

9. A host device, It is characterized in that include: Memory for storing computer programs; A processor, configured to implement the server fault repair method according to any one of claims 1 to 7 when executing the computer program.

10. The host device according to claim 9, It is characterized in that The host device is a baseboard management controller.

11. A server fault repair system, It is characterized in that include: A host device, configured to execute the server fault repair method according to any one of claims 1 to 7; A model deployment device, in which a pre-trained language model is deployed, is used to receive input text sent by the host device, input the input text into the pre-trained language model, and return the target fault type and target fault repair operation output by the pre-trained language model to the host device.

12. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are loaded and executed by the processor, the server fault repair method according to any one of claims 1 to 7 is implemented.