Terminal exception handling method and terminal performance monitoring sentry system

Abstract

The invention relates to a terminal exception handling method and a terminal performance monitoring sentry system, and belongs to the technical field of terminal exception monitoring, and the exception handling method comprises the following steps: receiving and cleaning exception information to obtain effective exception prompt words, obtaining a plurality of effective keywords according to the effective exception prompt words, obtaining a target exception problem list through a similarity matching algorithm according to the plurality of effective keywords, and executing corresponding exception handling operation according to the target exception problem list; according to the method, some performance abnormity problems existing in the terminal can be found in time, early warning can be made, the abnormity problems can be quickly and effectively positioned, a solution is given, normal operation of the terminal can be ensured, the product quality of a company is gradually improved, and then the user satisfaction degree is improved.

Description

Technical Field

[0001] The present application relates to the technical field of terminal anomaly monitoring, and in particular to a terminal anomaly processing method and a terminal performance monitoring sentinel system. Background Art

[0002] In recent years, with the development of Internet TV, the number of Internet TV users in my country has exceeded 1 billion. Among them, China Mobile's smart TV OTT terminals have been installed in more than 30 million households, covering more than 100 million users. Under the business background and huge user scale of China Mobile, we serve, in order to timely discover terminal performance problems and issue warnings to avoid adverse effects on large-scale users, it is particularly important to monitor terminal performance in a timely manner and respond to risks.

[0003] The traditional way of handling terminal performance anomalies is to generate exception information when there is a performance anomaly and send the exception information directly to the exception handler for processing. The exception handler will eliminate performance problems one by one based on the data in the exception information. Because it is necessary to lock in the cause of the performance anomaly based on personal experience, the root cause tracing of the anomaly is not timely enough and the efficiency of performance anomaly handling is low. Summary of the invention

[0004] To this end, the present invention provides a terminal exception handling method and a terminal performance monitoring sentinel system to solve the problems in the prior art of insufficient timeliness in tracing the root cause of the exception and low efficiency in handling performance exceptions.

[0005] In a first aspect, a method for handling terminal exceptions is provided, the method comprising:

[0006] Receive and clean abnormal information to obtain effective abnormal prompts;

[0007] Obtaining multiple valid keywords according to the valid abnormal prompt words;

[0008] Obtaining a target abnormal problem list based on the plurality of valid keywords through a similarity matching algorithm;

[0009] Corresponding exception handling operations are performed according to the target exception problem list to restore the terminal that reports the exception to a normal state.

[0010] Furthermore, obtaining a target abnormal problem list based on the plurality of effective keywords through a similarity matching algorithm includes:

[0011] Counting the word counts of the multiple valid keywords respectively to obtain a word count array;

[0012] Obtaining a list of similar abnormal questions corresponding to the word count of each keyword through the similarity value acquisition step;

[0013] The similarity value acquisition step includes: acquiring a keyword word count from the word count array as a first word count; calculating a similarity value of each preset abnormal question in a preset abnormal question library according to the first word count; sorting the preset abnormal questions from large to small according to the similarity values ​​to obtain a list of similar abnormal questions of the first word count;

[0014] Obtaining a preset number of abnormal questions in a list of similar abnormal questions corresponding to the number of words of each keyword to obtain a plurality of target abnormal questions;

[0015] After removing duplicates from multiple target abnormal problems, the similarity values ​​are sorted from large to small to obtain a list of target similar abnormal problems.

[0016] Further, the calculating of the similarity value of each preset abnormal question in the preset abnormal question library according to the first word count includes:

[0017] Traverse the preset abnormal question library and obtain the similarity value of each preset abnormal question through the similarity formula according to the first word count; the similarity formula is:

[0018] S = i∩j / (i+ji∩j);

[0019] Among them, S is the similarity value of the preset abnormal problem, i is the number of abnormal split words obtained after splitting the effective abnormal prompt according to the first word number; j is the number of problem split words obtained after splitting the preset abnormal problem according to the first word number; i∩j is the number of identical participles in the abnormal split words and the problem split words.

[0020] Furthermore, the number of abnormal split words obtained by splitting the effective abnormal prompt word according to the first word count includes:

[0021] Storing each character in the effective abnormal prompt into an array in sequence to obtain a first character array;

[0022] Extract m abnormal split words from the first character array according to the extraction formula, the extraction formula is:

[0023] C n =a[k]a[k+1]…a[k+q-1];

[0024] Among them, C n is the nth abnormal split word; a is the first character array; a[k] is the character with index k in the first character array, k=n-1; q is the first word number;

[0025] The calculation formula for the number of abnormal split words is: m=s-(q-2), where m is the number of abnormal split words, s is the maximum index value of the first character array, and q is the first word number.

[0026] Furthermore, the number of question split words obtained after splitting the preset abnormal question according to the first word count includes:

[0027] Storing each character in the preset abnormal problem into an array in sequence to obtain a second character array;

[0028] Extract t question split words from the second character array according to the extraction formula, the extraction formula is:

[0029] W n = b[x]b[x+1]…b[x+q-1];

[0030] Among them, W n Split the word for the nth question; b is the second character array; b[x] is the character with index x in the second character array, x=n-1; where q is the first word number;

[0031] The formula for calculating the number of problem split words is: t=r-(q-2), where t is the number of abnormal split words, r is the maximum index value of the second character array, and q is the number of the first word.

[0032] Furthermore, the step of removing duplicates from multiple target abnormal problems includes:

[0033] Remove duplicate target abnormal questions with lower similarity values, or remove duplicate target abnormal questions with equal similarity values.

[0034] Furthermore, the multiple valid keywords obtained according to the valid abnormal prompt include:

[0035] Identify the effective abnormal prompt words through the existing technology to obtain multiple problem words;

[0036] The question word is searched in a preset keyword library, and if it can be found, the question word is a valid keyword.

[0037] Further, the receiving and cleaning of the abnormal information to obtain a valid abnormal prompt includes:

[0038] Extracting information of a preset field in the abnormal information as an abnormal prompt;

[0039] The characters in the abnormal prompt are retained to obtain a valid abnormal prompt.

[0040] Furthermore, the performing of corresponding exception handling operations according to the target exception problem list includes:

[0041] Obtain all the target abnormal problems in the target abnormal problem list;

[0042] Acquire multiple target solutions from a preset solution library according to the target abnormal problem;

[0043] The corresponding target solutions are sent to the exception handling personnel one by one in descending order of the similarity values ​​of the target exception problems.

[0044] In a second aspect, a terminal performance monitoring sentinel system is provided, wherein the sentinel system applies any of the above-mentioned terminal exception handling methods.

[0045] The present invention adopts the above technical solution and has at least the following beneficial effects:

[0046] A terminal exception handling method and a terminal performance monitoring sentinel system are provided, which receive and clean exception information to obtain effective exception prompts, obtain multiple effective keywords based on the effective exception prompts, obtain a target exception problem list through a similarity matching algorithm based on the multiple effective keywords, and perform corresponding exception handling operations based on the target exception problem list; this method can timely discover some performance abnormalities in the terminal and issue early warnings, and quickly and effectively locate the abnormal problems and provide solutions to ensure the normal operation of the terminal, gradually improve the company's product quality, and thus improve user satisfaction.

[0047] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 The figure is a flow chart of a method for handling terminal exceptions according to an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0051] In the existing technology: the method of discovering terminal performance anomalies faces several serious problems, such as: the number of terminals is too large to be managed, the anomaly discovery is not proactive enough and relies on user complaints, the anomaly discovery is not timely enough, resulting in the spread of the problem, the inability to quickly determine the scope of the affected terminals is prone to omissions, and the difficulty in tracing the root cause leads to a long solution time.

[0052] The embodiments of the present application provide a method for handling terminal exceptions and a terminal performance monitoring sentinel system, which can promptly detect some performance abnormalities in the terminal and issue early warnings, and quickly and effectively locate the abnormal problems and provide solutions to ensure that the terminal can operate normally, gradually improve the company's product quality, and thus improve user satisfaction.

[0053] The testing method and device in this application are described below through specific examples.

[0054] See also Figure 1 , Figure 1 is a flowchart of a method for handling terminal exceptions according to an exemplary embodiment of the present invention. Figure 1 , the method comprising:

[0055] Step S11, receiving and cleaning abnormal information to obtain effective abnormal prompts;

[0056] Step S12, obtaining multiple valid keywords according to the valid abnormal prompts;

[0057] Step S13, obtaining a target abnormal problem list through a similarity matching algorithm based on multiple valid keywords;

[0058] Step S14: Execute corresponding exception handling operations according to the target exception problem list to restore the terminal that reports the exception to a normal state.

[0059] It should be noted that the technical solution provided in this embodiment can be applied to scenarios including, but not limited to: abnormal terminal performance monitoring.

[0060] It can be understood that the testing method provided in this embodiment can timely discover some performance abnormalities in the terminal and issue early warnings, and quickly and effectively locate the abnormal problems and provide solutions to ensure that the terminal can operate normally, gradually improve the company's product quality, and thus improve user satisfaction.

[0061] In specific practice, step S11 "receiving and cleaning the abnormal information to obtain a valid abnormal prompt" includes: extracting information of a preset field in the abnormal information as the abnormal prompt; retaining characters in the abnormal prompt to obtain a valid abnormal prompt.

[0062] It should be noted that the exception information is log information, which includes information from multiple fields, such as a recording time field, an exception occurrence duration field, etc.; the preset field may be an exception information field, which mainly records a description of the performance exception.

[0063] In specific practice, step S12 "obtaining multiple valid keywords based on valid abnormal prompts" is specifically: identifying valid abnormal prompts through existing technologies to obtain multiple problem words; searching for the problem words in a preset keyword library, and if the problem words can be found, the corresponding problem words are valid keywords.

[0064] It should be noted that the preset keyword library is obtained by extracting keywords from common performance abnormality problems, and the extracted keywords are accumulated and stored to form the preset keyword library.

[0065] In specific practice, step S13 "obtaining a target abnormal problem list through a similarity matching algorithm based on multiple valid keywords" includes: respectively counting the keyword word counts of multiple valid keywords to obtain a word count array; obtaining a similar abnormal problem list corresponding to each keyword word count through a similarity value acquisition step; similarity value acquisition step: obtaining a keyword word count from the word count array as the first word count; calculating the similarity value of each preset abnormal problem in the preset abnormal problem library based on the first word count; sorting the preset abnormal problems from large to small according to the similarity value to obtain a similar abnormal problem list of the first word count; obtaining a preset number of preset abnormal problems before the similar abnormal problem list corresponding to each keyword word count to obtain multiple target abnormal problems; after removing duplicates from the multiple target abnormal problems, sorting them from large to small according to the similarity value to obtain a target similar abnormal problem list.

[0066] It should be noted that the preset abnormal problem library is a collection of preset abnormal problems; each time, some typical performance abnormal problems are selected from the reported performance abnormal problems as preset abnormal problems; the preset number is set according to the similarity accuracy requirement, and the higher the accuracy requirement, the more preset numbers are set.

[0067] Specifically, calculating the similarity value of each preset abnormal question in the preset abnormal question library according to the first word count includes: traversing the preset abnormal question library and obtaining the similarity value of each preset abnormal question according to the first word count by a similarity formula; the similarity formula is:

[0068] S=i∩j / (i+ji∩j); wherein S is the similarity value of the preset abnormal problem, i is the number of abnormal split words obtained after splitting the effective abnormal prompt according to the first number of words; j is the number of problem split words obtained after splitting the preset abnormal problem according to the first number of words; i∩j is the number of identical participles in the abnormal split words and the problem split words.

[0069] Specifically, the number of abnormal split words obtained after splitting the effective abnormal prompt according to the first word number includes: storing each character in the effective abnormal prompt into an array in sequence to obtain a first character array; extracting m abnormal split words from the first character array according to an extraction formula, and the extraction formula is: C n =a[k]a[k+1]…a[k+q-1]; among them, C n is the nth abnormal split word; a is the first character array; a[k] is the character with index k in the first character array, k=n-1; q is the first word number; the calculation formula for the number of abnormal split words is: m=s-(q-2), where m is the number of abnormal split words, s is the maximum index value of the first character array, and q is the first word number.

[0070] Specifically, the number of problem split words obtained after splitting the preset abnormal question according to the first word count includes: storing each character in the preset abnormal question into an array in sequence to obtain a second character array; extracting t problem split words from the second character array according to an extraction formula, and the extraction formula is: W n =b[x]b[x+1]…b[x+q-1]; where W n is the nth problem split word; b is the second character array; b[x] is the character with index x in the second character array, x=n-1; q is the first word number; the formula for calculating the number of problem split words is: t=r-(q-2), where t is the number of abnormal split words, r is the maximum index value of the second character array, and q is the first word number.

[0071] In specific practice, multiple target anomaly issues are deduplicated, including: removing duplicate target anomaly issues with low similarity values, or removing any duplicate target anomaly issues with equal similarity values.

[0072] In specific practice, step S14 "performing corresponding exception handling operations based on the target exception problem list" includes: obtaining all target exception problems in the target exception problem list; obtaining multiple target solutions from a preset solution library based on the target exception problems; and sending the corresponding target solutions to the exception handling personnel one by one in descending order of similarity values ​​of the target exception problems.

[0073] It should be noted that the preset solution library is associated with the preset abnormal problem library, and the preset solution to each preset abnormal problem in the preset abnormal problem library exists in the preset solution library; each time some typical performance abnormal problems are selected from the reported performance abnormal problems as preset abnormal problems, and the solution to solve the performance abnormal problem is used as the preset solution.

[0074] In a specific embodiment, the abnormal information extracted from the log is cleaned, filtered and intelligently segmented, and the effective abnormal prompt is: "Startup abnormality, source configuration missing."

[0075] After cleaning the special symbols in the effective abnormal prompt, the preset keyword library is matched. After matching, the effective keywords are obtained as "start, abnormal, startup abnormal, source configuration". The keyword word count of the effective keyword is obtained as the word count array [2, 3, 4], and then the three different word counts in the word count array are used to perform similarity matching with the preset abnormal problems in the preset abnormal problem library:

[0076] For example, there are problems in the preset abnormal problem library; Q1: Solutions to common abnormal startup problems of set-top boxes; Q2...Qz;

[0077] Decompose the effective abnormal prompt words using 2 in the word count array, and get the abnormal split words "start", "dynamic difference", "abnormal", "constant source", "source matching", "configuration", "absent", "missing", and the number is 8; then decompose the question Q1 according to 2 in the word count array, and get the question split words "set-top", "top box"... "solution", the number is 14, and the number of the same words in the abnormal split words and the question split words is 3, which are "start", "dynamic difference", and "abnormal"; calculate the similarity value:

[0078] Q1 obtains similarity value: 3 / (14+8-3)=3 / 19=0.157;

[0079] Similarly, calculate the similarity values ​​of Q2...Qz.

[0080] Then use 3 and 4 in the word count array to calculate the similarity values ​​of Q1...Qz again, and arrange Q1...Qz corresponding to the keyword word counts 2, 3, and 4 in order of their respective similarity values ​​from large to small to obtain the corresponding three sortings; the first 5 preset abnormal problems in each sorting are used as target abnormal problems; remove repeated target abnormal problems with low similarity values, or remove any repeated target abnormal problems with equal similarity values ​​to obtain a list of target similar abnormal problems, obtain all target abnormal problems in the target abnormal problem list, obtain multiple target solutions from the preset solution library based on the target abnormal problem, and send the corresponding target solutions to the abnormality handlers one by one in order of the similarity values ​​of the target abnormal problems from large to small.

[0081] In specific practice, a terminal performance monitoring sentinel system is also provided, and the sentinel system applies any of the above-mentioned terminal exception handling methods.

[0082] In a specific embodiment, in order to solve the pressure on the server caused by large-scale data access, the terminal performance monitoring sentinel system 1. Use Nginx load balancing to filter illegal requests while evenly distributing the pressure of data access to the portal access end, and then upload the data to the Kafka message queue after the access end processes the data to achieve buffering of large amounts of data access. 2. Use Flink streaming real-time computing technology to extract Kafka data for cleaning, screening and data calculation, and then write the valid data back to Kafka again. 3. Use Clickhouse cluster to consume Kafka data through multi-node engine tables and write it to the database, spreading the data pressure to all nodes.

[0083] In a specific embodiment, in order to solve the statistical query efficiency problem of billions of data, the terminal performance monitoring sentinel system pre-processes the data according to the time granularity according to the actual application scenario, and performs regular statistics of historical data at the hourly granularity and daily granularity respectively: a. Every morning, through the scheduled task, extract the historical data of the previous day, group them by time granularity, count the data required for the query and store them in the hourly table and daily table, saving the time consumed by the query. b. Perform time segmentation according to the page time range, judge and query the historical hourly table, historical daily table and real-time table according to the segmented time, and then combine the data to complete the return of the final data.

[0084] In a specific embodiment, the terminal performance monitoring sentinel system also provides multi-tenant multi-application, realizes the separation of business data, reduces business coupling, and facilitates better expansion in the future. According to the company's business needs, multi-tenants are configured by channel, each channel can be managed independently, and then multiple applications are managed under the tenant, isolating the differences between different application businesses, leaving enough space for the subsequent personalized expansion of each application.

[0085] In a specific embodiment, the terminal performance monitoring sentinel system has unified configuration management, 1. Configure the switches of each module of the SDK application. According to business needs, we divide the modules into user experience, network, exception, log, heartbeat and other modules. The switch control and sampling rate configuration can be realized for each module, and the data uploaded by the SDK can be flexibly switched on and off. 2. Configure the black and white lists, which can accurately operate VIP users, and can also intercept invalid reported data and dirty data from entering the system to avoid wasting system resources. 3. Custom parameter configuration can control the reporting cycle and reporting items of events such as pages, fluency, heartbeat, freeze, log, etc., and can also control the size of uploaded logs, log capture time, fluency threshold, startup delay, etc.

[0086] In a specific embodiment, the terminal performance monitoring sentinel system has real-time alarms. 1. In order to promptly detect large-scale failures and abnormal problems in the terminal, the sentinel platform provides a real-time alarm configuration module. The scheduled task is based on the configured system type, device type, application and other business parameters, at a time granularity of 1 minute, 5 minutes, etc., and multiple alarm dimensions for statistics and alarms.

[0087] In a specific embodiment, the terminal performance monitoring sentinel system has user tracking, 1. It provides a hot key trigger log reporting function. According to the combination of keys, users can trigger the reporting of the captured log through the hot key when encountering problems, and then upload it to the platform. Through the self-service root cause analysis function, the algorithm is called to automatically retrieve the problem and return the processing method. 2. The sentinel platform provides the ability to accurately track specific users. For user complaints or VIP users of precision marketing, various event data and traceability abnormal causes reported by the user terminal can be viewed, and the online viewing and downloading functions of the log are provided.

[0088] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

[0089] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0090] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0091] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for handling terminal exceptions, characterized in that: The method comprises: Receive and clean abnormal information to obtain effective abnormal prompts; Obtaining multiple valid keywords according to the valid abnormal prompt words; Obtaining a target abnormal problem list based on the plurality of valid keywords through a similarity matching algorithm; Corresponding exception handling operations are performed according to the target exception problem list to restore the terminal that reports the exception to a normal state.

2. The method according to claim 1, characterized in that The method of obtaining a target abnormal problem list based on the plurality of valid keywords through a similarity matching algorithm includes: Counting the word counts of the multiple valid keywords respectively to obtain a word count array; Obtaining a list of similar abnormal questions corresponding to the word count of each keyword through the similarity value acquisition step; The similarity value acquisition step includes: acquiring a keyword word count from the word count array as a first word count; calculating a similarity value of each preset abnormal question in a preset abnormal question library according to the first word count; sorting the preset abnormal questions from large to small according to the similarity values ​​to obtain a list of similar abnormal questions of the first word count; Obtaining a preset number of abnormal questions in a list of similar abnormal questions corresponding to the number of words of each keyword to obtain a plurality of target abnormal questions; After removing duplicates from multiple target abnormal problems, the similarity values ​​are sorted from large to small to obtain a list of target similar abnormal problems.

3. The method according to claim 2, characterized in that The calculating, based on the first number of words, a similarity value of each preset abnormal question in a preset abnormal question library comprises: Traverse the preset abnormal question library and obtain the similarity value of each preset abnormal question through the similarity formula according to the first word count; the similarity formula is: S = i∩j / (i+ji∩j); Among them, S is the similarity value of the preset abnormal problem, i is the number of abnormal split words obtained after splitting the effective abnormal prompt according to the first word number; j is the number of problem split words obtained after splitting the preset abnormal problem according to the first word number; i∩j is the number of identical participles in the abnormal split words and the problem split words.

4. The method according to claim 3, characterized in that The number of abnormal split words obtained by splitting the effective abnormal prompt words according to the first word count includes: Storing each character in the effective abnormal prompt into an array in sequence to obtain a first character array; Extract m abnormal split words from the first character array according to the extraction formula, the extraction formula is: C n =a[k]a[k+1]…a[k+q-1]; Among them, C n is the nth abnormal split word; a is the first character array; a[k] is the character with index k in the first character array, k=n-1; q is the first word number; The calculation formula for the number of abnormal split words is: m=s-(q-2), where m is the number of abnormal split words, s is the maximum index value of the first character array, and q is the first word number.

5. The method according to claim 3, characterized in that: The number of question split words obtained by splitting the preset abnormal question according to the first word count includes: Storing each character in the preset abnormal problem into an array in sequence to obtain a second character array; Extract t question split words from the second character array according to the extraction formula, the extraction formula is: W n =b[x]b[x+1]…b[x+q-1]; Among them, W n Split the word for the nth question; b is the second character array; b[x] is the character with index x in the second character array, x=n-1; where q is the first word number; The formula for calculating the number of problem split words is: t=r-(q-2), where t is the number of abnormal split words, r is the maximum index value of the second character array, and q is the number of the first word.

6. The method according to claim 2, characterized in that Deduplication of multiple target abnormal issues includes: Remove duplicate target abnormal questions with lower similarity values, or remove duplicate target abnormal questions with equal similarity values.

7. The method according to claim 1, characterized in that The multiple valid keywords obtained according to the valid abnormal prompt include: Identify the effective abnormal prompt words through the existing technology to obtain multiple problem words; The problem word is searched in a preset keyword library, and if it can be found, the problem word is a valid keyword.

8. The method according to claim 1, characterized in that The receiving and cleaning of abnormal information to obtain effective abnormal prompts includes: Extracting information of a preset field in the abnormal information as an abnormal prompt; The characters in the abnormal prompt are retained to obtain a valid abnormal prompt.

9. The method according to claim 1, characterized in that: The performing corresponding exception handling operations according to the target exception problem list includes: Obtain all the target abnormal problems in the target abnormal problem list; Acquire multiple target solutions from a preset solution library according to the target abnormal problem; The corresponding target solutions are sent to the exception handling personnel one by one in descending order of the similarity values ​​of the target exception problems.

10. A terminal performance monitoring sentinel system, characterized in that: The sentinel system applies the terminal exception handling method described in any one of claims 1-9.

Images