Software operation abnormity analysis method and system based on AI
Through AI-based software operation anomaly analysis methods, combined with performance and text information, accurate fault detection and rapid repair are achieved in dynamic load environments, solving the problems of false alarms and missed detections in traditional methods and improving fault handling efficiency.
Patent Information
- Application Number
- CN202510821640.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-19
- Publication Date
- 2025-09-26
AI Technical Summary
Traditional anomaly detection technology has difficulty adapting to business fluctuations in dynamic load environments, resulting in false alarms or missed detections, and lacks cross-modal correlation capabilities, resulting in long fault repair times.
An AI-based software operation anomaly analysis method is adopted to collect performance information and text information, perform noise reduction and matching, calculate the judgment value P, combine the duration and fault value G to send detection instructions, locate the fault and call the repair strategy for isolation and repair.
It enables accurate judgment of software anomalies in different situations, timely detection and handling of faults, reduces fault repair time, and avoids software crashes and problem expansion.
Smart Images
Figure CN120705014A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of software operation anomaly analysis, and more specifically to an AI-based software operation anomaly analysis method and system. Background Art
[0002] As software systems expand in size and their architectures become more complex, operational anomaly detection and root cause identification have become core challenges in operations and maintenance, and are crucial steps that cannot be ignored during software development and maintenance. Whether it's a startup project or a large-scale system, software operations inevitably encounter anomalies. Effective anomaly management and handling not only improves product quality but also has a profound impact on project progress, user experience, and company reputation. Traditional anomaly detection technologies primarily rely on threshold alerts and manual rule configuration. For example, static thresholds based on metrics like CPU usage and memory usage, such as a CPU ≥ 90%, trigger an alert. However, these methods have significant drawbacks. In dynamic load environments, fixed thresholds struggle to adapt to business fluctuations. For example, short-term CPU spikes during e-commerce promotions are normal, and traditional methods are prone to triggering false alerts. Low-severity memory leaks may also be missed because they don't reach the threshold. The data generated by modern systems includes multimodal information such as logs, performance indicators, call chains (topological relationships), etc. Existing tools usually process single data types independently and lack cross-modal correlation capabilities. When anomalies occur, operations and maintenance personnel need to manually correlate logs, monitoring dashboards, and service dependency graphs. The average fault repair time is as long as several hours, so the response speed is too slow. Summary of the Invention
[0003] In order to overcome the above-mentioned defects of the prior art, the implementation regulations of the present invention provide an AI-based software operation abnormality analysis method and system to solve the technical problems raised in the background technology.
[0004] To achieve the above objectives, the present invention provides the following technical solution: an AI-based software operation anomaly analysis method, comprising the following steps: Step S1: Collecting data information, performance information and text information during software operation; Step S2: noise reduction processing is performed on the performance information, content extraction is performed on the text information, and the processed performance information and text information are matched; Step S3: Calculate a judgment value P based on the performance information. When the judgment value P is 0 or a negative number, extract the duration data SJ and the number of occurrences data CS. Send a detection instruction when the duration data SJ ≥ the duration threshold CY. Calculate a fault value G when the duration data SJ < the duration threshold CY. Step S4: When the fault value G is greater than or equal to the proportional threshold BY, a detection instruction is sent; when the fault value G is less than the proportional threshold BY, no instruction is sent; Step S5: When the judgment value P is a positive number, the detection instruction is sent if the judgment value P ≥ the fault threshold GY; otherwise, no instruction is sent; Step S6: Analyze the fault problem when sending the instruction, locate the fault location, call the repair strategy, and perform fault isolation and repair.
[0005] An AI-based software operation anomaly analysis system includes an acquisition unit, a processing unit, a fusion unit, a central unit, a detection unit, a positioning unit, a repair strategy library, and a fault unit. The acquisition unit is used to collect data information during software operation. The processing unit processes the data information collected by the acquisition unit. The fusion unit fuses and matches the data processed by the processing unit. The central unit receives data sent by the processing unit and sends instructions. The detection unit retrieves information and sends it to the fault unit. The positioning unit is used to locate the location of the software operation anomaly. The repair strategy library is used to provide fault processing measurements. The fault unit is used to perform fault repair. The acquisition unit includes a performance module and a text module. The performance module acquires performance information when the software is running. The performance information includes CPU occupancy data ZY, API interface delay data YC and garbage collection frequency data HS. The text module acquires application logs corresponding to the software running. The application logs are text information. The performance module and the text module send the collected data to the processing unit respectively. The processing unit receives the performance information sent by the performance module. The processing unit performs noise reduction processing on the performance information in the performance module, and the processing unit uses discrete wavelet transform for denoising to remove high-frequency noise and retain low-frequency trend components.
[0006] In a preferred embodiment, the processing unit receives the data sent by the text module and performs content extraction, the processing unit extracts the timestamp, log level, service name and thread ID data in the application log, and the processing unit uses the timestamp to correspond the log level, service name and thread ID data to the same time window, and the processing unit matches the processed performance module with the information in the text module and sends it to the detection unit.
[0007] In a preferred embodiment, the processing unit sends the data processed in the performance module to the central unit, and the central unit receives the data sent by the processing unit and calculates the judgment value P. The calculation formula of the judgment value P in the central unit is: , where k1, k2, and k3 are weights, sgn is the rounding function, PL is the standard garbage collection frequency, and BZ is the CPU usage threshold.
[0008] In a preferred embodiment, when the judgment value P calculated by the central unit is 0 or a negative number, the central unit extracts the duration data SJ and the number of occurrences data CS when the judgment value P is 0 or a negative number, and the central unit compares the duration data SJ with its internal duration threshold CY. When the duration data SJ ≥ the duration threshold CY, the central unit sends a detection instruction to the detection unit. When the duration data SJ < the duration threshold CY, the central unit calculates the fault value G.
[0009] In a preferred embodiment, the calculation formula of the fault value G in the central unit is: , where i is the current i-th judgment value P is 0 or negative, SJ i is the duration of the i-th judgment value P being 0 and a negative number, ZS is the total time when the judgment value P is 0 and a negative number, the central unit compares the fault value G with its internal proportional threshold BY, when the fault value G ≥ the proportional threshold BY, the central unit sends a detection instruction to the detection unit, when the fault value G < the proportional threshold BY, the central unit does not send an instruction.
[0010] In a preferred embodiment, when the judgment value P calculated by the central unit is a positive number, the central unit compares the detected judgment value P with its internal fault threshold GY. When the judgment value P≥fault threshold GY, the central unit sends a detection instruction to the detection unit. When the judgment value P<fault threshold G, the central unit does not send an instruction. After the detection unit receives the detection instruction, the detection unit retrieves the information sent by the processing unit when the judgment value P≥fault threshold GY, and the detection unit extracts the information in the text module and sends it to the fault unit.
[0011] In a preferred embodiment, the fault unit receives the information sent by the detection unit and analyzes the fault problem, and the fault unit sends the analyzed fault problem to the positioning unit. The positioning unit locates the location of the abnormal software operation fault, and the fault unit calls the repair strategy from the repair strategy library. The fault unit uses the called repair strategy to repair the fault.
[0012] In a preferred embodiment, the fault unit includes an isolation module and a temporary module. The isolation module directs the traffic of the abnormal software to the backup node, and the isolation module points the domain name resolution of the fault area to the healthy area, marks the faulty Pod as unschedulable, and the temporary module restarts the software node where the fault problem occurs and rolls back the code data.
[0013] The technical effects and advantages of the present invention are as follows: By collecting performance information and text information during software operation, the present invention can understand the fault location when a problem occurs during software operation. In addition, instead of using only the CPU occupancy threshold for judgment, the present invention performs abnormal operation fault analysis when the CPU occupancy data ZY exceeds the threshold, and performs analysis when the CPU occupancy ZY does not reach the threshold. This allows for timely detection and processing of faults in the early stages of the fault, thus avoiding direct abnormal software crashes. The present invention calculates a judgment value P by collecting CPU occupancy data ZY, API interface delay data YC, and garbage collection frequency data HS. When the judgment value P is 0 or a negative number, a judgment is made directly. The duration data SJ is compared with the duration threshold CY. When the duration data SJ ≥ the duration threshold CY, a long-term CPU occupancy abnormality occurs and fault processing is performed. When the duration data SJ < the duration threshold CY, the calculated fault value G is used for judgment. When the fault value G ≥ the ratio threshold BY, although the duration of the CPU occupancy abnormality is short, it will recur, so fault processing is also performed in a timely manner. When the fault value G < the ratio threshold BY, it is a short-term peak abnormality, not a fault. Therefore, the present application can make judgments in different situations to ensure the accuracy of the judgment. When the present invention performs fault processing, the positioning unit locates the location of the software operation abnormality fault based on the collected performance information and text information, and calls the repair strategy from the repair strategy library, so that subsequent processing work can be carried out quickly. When performing fault processing, the isolation module isolates the location of the software operation abnormality to avoid damage to the remaining software and the remaining operating links of the software, and then temporarily processes the location of the software operation abnormality fault through the temporary module. Minor problems can be repaired directly, and serious problems can also be temporarily processed to avoid the problem from expanding, thereby reducing the time for the final complete repair of the fault. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 Schematic diagram of the steps of the software operation abnormality analysis method of the present invention.
[0015] Figure 2 It is a schematic diagram of the overall system structure of the present invention. DETAILED DESCRIPTION
[0016] The technical solutions of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the present invention. In addition, the forms of the various structures described in the following embodiments are merely examples. The AI-based software operation anomaly analysis method and system involved in the present invention are not limited to the various structures described in the following embodiments. All other implementations obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0017] Reference Figure 1 and Figure 2 The present invention provides an AI-based software operation abnormality analysis method, comprising the following steps: Step S1: Collecting data information, performance information and text information during software operation; Step S2: noise reduction processing is performed on the performance information, content extraction is performed on the text information, and the processed performance information and text information are matched; Step S3: Calculate a judgment value P based on the performance information. When the judgment value P is 0 or a negative number, extract the duration data SJ and the number of occurrences data CS. Send a detection instruction when the duration data SJ ≥ the duration threshold CY. Calculate a fault value G when the duration data SJ < the duration threshold CY. Step S4: When the fault value G is greater than or equal to the proportional threshold BY, a detection instruction is sent; when the fault value G is less than the proportional threshold BY, no instruction is sent; Step S5: When the judgment value P is a positive number, the detection instruction is sent if the judgment value P ≥ the fault threshold GY; otherwise, no instruction is sent; Step S6: Analyze the fault problem when sending the instruction, locate the fault location, call the repair strategy and perform fault isolation and repair; An AI-based software operation anomaly analysis system includes an acquisition unit, a processing unit, a fusion unit, a central unit, a detection unit, a positioning unit, a repair strategy library and a fault unit. The acquisition unit is used to collect data information during software operation, the processing unit processes the data information collected by the acquisition unit, the fusion unit fuses and matches the data processed by the processing unit, the central unit receives the data sent by the processing unit and sends instructions, the detection unit retrieves information and sends it to the fault unit, the positioning unit is used to locate the location of the software operation anomaly, the repair strategy library is used to provide fault processing measurements, and the fault unit is used to perform fault repair.
[0018] In an embodiment of the present application, when performing software operation abnormality analysis, the present application collects performance information and text information during software operation during the data collection stage. Therefore, when a problem occurs in the software operation, the content of the text information can be parsed to understand the fault location, and the present application calculates a judgment value P based on the performance information. When the judgment value P is 0 or a negative number, the CPU occupancy data ZY has exceeded the threshold, and fault judgment and analysis are directly performed. At this time, the duration data SJ ≥ the continuous threshold CY, which has exceeded the threshold, and the fault is directly processed. When the duration data SJ is less than the continuous threshold CY, it is separated separately to avoid a single CPU occupancy ZY being too high and directly performing a fault judgment, and analysis is performed when the CPU occupancy ZY does not reach the threshold, so that the fault is discovered in time at the early stage and processed in time to avoid the situation where the software directly collapses abnormally.
[0019] Reference Figure 2 The collection unit includes a performance module and a text module. The performance module collects performance information when the software is running. The performance information includes CPU occupancy data ZY, API interface delay data YC and garbage collection frequency data HS. The text module collects the application log corresponding to the software when it is running. The application log is text information. The performance module and the text module send the collected data to the processing unit respectively. The processing unit receives the performance information sent by the performance module. The processing unit performs noise reduction processing on the performance information in the performance module, and the processing unit uses discrete wavelet transform for noise reduction processing to remove high-frequency noise and retain low-frequency trend components. The processing unit receives the data sent by the text module and performs content extraction. The processing unit extracts the timestamp, log level, service name and thread ID data in the application log, and the processing unit uses the timestamp to correspond the log level, service name and thread ID data to the same time window, and the processing unit matches the information in the processed performance module and text module and sends them to the detection unit.
[0020] In an embodiment of the present application, the collection unit collects data information through its internal performance module and text module. The performance module collects CPU occupancy data ZY, API interface delay data YC, and garbage collection frequency data HS. The above three types can accurately express the health information of the software during operation, while the text module collects the application log corresponding to the software during operation. The text module records the specific text information of the software operation, which is convenient for locating the location of software operation failures and judging the type of failure, and then facilitates fault repair. The new energy information is denoised to ensure the accuracy of the collected data, and the timestamp, log level, service name and thread ID data in the application log are extracted. The timestamp is used to correspond the log level, service name and thread ID data to the same time window, so that the text information within the same time is bound. At this time, the information in the performance module and the text module are matched. The performance information can be matched with the information in the application log according to time. When an operation abnormality occurs, the time when the abnormality occurred can be understood, and all information within the time can be mastered.
[0021] Reference Figure 2 The processing unit sends the data processed in the performance module to the central unit, and the central unit receives the data sent by the processing unit and calculates the judgment value P. The calculation formula of the judgment value P in the central unit is: , where k1, k2 and k3 are weights, sgn is a rounding function, PL is the standard garbage collection frequency, BZ is the CPU occupancy threshold, when the judgment value P calculated by the central unit is 0 or a negative number, the central unit extracts the duration data SJ and the number of occurrences data CS when the judgment value P is 0 or a negative number, and the central unit compares the duration data SJ with its internal duration threshold CY. When the duration data SJ ≥ the duration threshold CY, the central unit sends a detection instruction to the detection unit. When the duration data SJ < the duration threshold CY, the central unit calculates the fault value G. The calculation formula of the fault value G in the central unit is: , where i is the current i-th judgment value P is 0 or negative, SJ i is the duration of the i-th judgment value P being 0 and a negative number, ZS is the total time when the judgment value P is 0 and a negative number, the central unit compares the fault value G with its internal proportional threshold BY, when the fault value G ≥ the proportional threshold BY, the central unit sends a detection instruction to the detection unit, when the fault value G < the proportional threshold BY, the central unit does not send an instruction.
[0022] In the embodiment of the present application, sgn is a rounding function, so it outputs 1 when a positive number is input, 0 when 0 is input, and -1 when a negative number is input. First, a judgment value P is calculated. When the judgment value P is 0 or a negative number, the CPU usage data ZY has reached or exceeded the CPU usage threshold. Therefore, in this case, a judgment is made directly. When the judgment value P is 0 or a negative number, the duration data SJ is compared with the duration threshold CY. When the duration data SJ ≥ the duration threshold CY, a long-term CPU usage anomaly occurs. Therefore, a detection instruction is directly sent, and the fault is subsequently processed. When the duration data SJ is less than the duration threshold CY, the calculated fault value G is used for judgment. When the fault value G ≥ the proportional threshold BY, although the duration of the CPU usage anomaly is short, it will recur. Therefore, the fault is handled in a timely manner. When the fault value G is less than the proportional threshold BY, it is a short-term peak anomaly, not a fault. Therefore, the present application can make judgments in different situations to ensure the accuracy of the judgment.
[0023] Reference Figure 2 When the judgment value P calculated by the central unit is a positive number, the central unit compares the detected judgment value P with its internal fault threshold GY. When the judgment value P≥fault threshold GY, the central unit sends a detection instruction to the detection unit. When the judgment value P<fault threshold G, the central unit does not send an instruction. After the detection unit receives the detection instruction, the detection unit retrieves the information sent by the processing unit when the judgment value P≥fault threshold GY, and the detection unit extracts the information in the text module and sends it to the fault unit.
[0024] In an embodiment of the present application, when the judgment value P is a positive number, the CPU occupancy rate ZY has not reached the threshold, but problems may occur during its operation. At this time, the CPU occupancy rate data ZY, API interface delay data YC and garbage collection frequency data HS are used for comprehensive judgment. The larger the values of the CPU occupancy rate data ZY, API interface delay data YC and garbage collection frequency data HS, the higher the possibility of an abnormality. When the judgment value P ≥ fault threshold GY, an abnormality has occurred during the software operation and is processed in a timely manner, so that the present application can perform detection and analysis under different circumstances and ensure the accuracy of detection and analysis.
[0025] Reference Figure 2The fault unit receives the information sent by the detection unit and parses the fault problem, and the fault unit sends the parsed fault problem to the positioning unit. The positioning unit locates the location of the abnormal software operation fault, and the fault unit calls the repair strategy from the repair strategy library. The fault unit uses the called repair strategy to repair the fault. The fault unit includes an isolation module and a temporary module. The isolation module directs the traffic of the abnormal software to the backup node, and the isolation module points the domain name resolution of the fault area to the healthy area, and marks the faulty Pod as unschedulable. The temporary module restarts the software node where the fault problem occurs and rolls back the code data.
[0026] In an embodiment of the present application, when performing fault handling, the positioning unit locates the location of the software operation abnormality based on the collected performance information and text information, and calls the repair strategy from the repair strategy library, so that subsequent processing work can be carried out quickly. When performing fault handling, the isolation module isolates the location of the software operation abnormality to avoid damage to the remaining software and the remaining operating links of the software, and then temporarily handles the location of the software operation abnormality through the temporary module. Minor problems can be repaired directly, and serious problems can also be temporarily handled to avoid the problem from expanding, thereby reducing the time for the final complete repair of the fault.
[0027] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The units and algorithm steps of each example described in the embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can 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 this application.
[0028] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0029] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.
[0030] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. An AI-based software operation anomaly analysis method, characterized by: The following steps are involved: Step S1: Collecting data information, performance information and text information during software operation; Step S2: noise reduction processing is performed on the performance information, content extraction is performed on the text information, and the processed performance information and text information are matched; Step S3: Calculate a judgment value P based on the performance information. When the judgment value P is 0 or a negative number, extract the duration data SJ and the number of occurrences data CS. Send a detection instruction when the duration data SJ ≥ the duration threshold CY. Calculate a fault value G when the duration data SJ < the duration threshold CY. Step S4: When the fault value G is greater than or equal to the proportional threshold BY, a detection instruction is sent; when the fault value G is less than the proportional threshold BY, no instruction is sent; Step S5: When the judgment value P is a positive number, the detection instruction is sent if the judgment value P ≥ the fault threshold GY; otherwise, no instruction is sent; Step S6: Analyze the fault problem when sending the instruction, locate the fault location, call the repair strategy, and perform fault isolation and repair.
2. An AI-based software operation anomaly analysis system, characterized by: It includes a collection unit, a processing unit, a fusion unit, a central unit, a detection unit, a positioning unit, a repair strategy library and a fault unit. The collection unit is used to collect data information when the software is running. The processing unit processes the data information collected by the collection unit. The fusion unit fuses and matches the data processed by the processing unit. The central unit receives the data sent by the processing unit and sends instructions. The detection unit retrieves information and sends it to the fault unit. The positioning unit is used to locate the location where the software is running abnormally. The repair strategy library is used to provide fault processing measurements. The fault unit is used to perform fault repair. The acquisition unit includes a performance module and a text module. The performance module acquires performance information when the software is running. The performance information includes CPU occupancy data ZY, API interface delay data YC and garbage collection frequency data HS. The text module acquires application logs corresponding to the software running. The application logs are text information. The performance module and the text module send the collected data to the processing unit respectively. The processing unit receives the performance information sent by the performance module. The processing unit performs noise reduction processing on the performance information in the performance module, and the processing unit uses discrete wavelet transform for denoising to remove high-frequency noise and retain low-frequency trend components.
3. The AI-based software operation anomaly analysis system according to claim 2, characterized in that: The processing unit receives the data sent by the text module and performs content extraction. The processing unit extracts the timestamp, log level, service name and thread ID data in the application log, and uses the timestamp to correspond the log level, service name and thread ID data to the same time window. The processing unit matches the processed performance module with the information in the text module and sends it to the detection unit.
4. The AI-based software operation anomaly analysis system according to claim 3, characterized in that: The processing unit sends the data processed in the performance module to the central unit, and the central unit receives the data sent by the processing unit and calculates the judgment value P. The calculation formula of the judgment value P in the central unit is: , where k1, k2, and k3 are weights, sgn is the rounding function, PL is the standard garbage collection frequency, and BZ is the CPU usage threshold.
5. The AI-based software operation anomaly analysis system according to claim 4, characterized in that: When the judgment value P calculated by the central unit is 0 or a negative number, the central unit extracts the duration data SJ and the number of occurrences data CS when the judgment value P is 0 or a negative number. The central unit compares the duration data SJ with its internal duration threshold CY. When the duration data SJ ≥ the duration threshold CY, the central unit sends a detection instruction to the detection unit. When the duration data SJ < the duration threshold CY, the central unit calculates the fault value G.
6. The AI-based software operation anomaly analysis system according to claim 5, characterized in that: The calculation formula of the fault value G in the central unit is: , where i is the current i-th judgment value P is 0 or negative, SJ i is the duration of the i-th judgment value P being 0 and a negative number, ZS is the total time when the judgment value P is 0 and a negative number, the central unit compares the fault value G with its internal proportional threshold BY, when the fault value G ≥ the proportional threshold BY, the central unit sends a detection instruction to the detection unit, when the fault value G < the proportional threshold BY, the central unit does not send an instruction.
7. The AI-based software operation anomaly analysis system according to claim 6, characterized in that: When the judgment value P calculated by the central unit is a positive number, the central unit compares the detected judgment value P with its internal fault threshold GY. When the judgment value P≥fault threshold GY, the central unit sends a detection instruction to the detection unit. When the judgment value P<fault threshold G, the central unit does not send an instruction. After the detection unit receives the detection instruction, the detection unit retrieves the information sent by the processing unit when the judgment value P≥fault threshold GY, and the detection unit extracts the information in the text module and sends it to the fault unit.
8. The AI-based software operation anomaly analysis system according to claim 7, characterized in that: The fault unit receives the information sent by the detection unit and analyzes the fault problem, and the fault unit sends the analyzed fault problem to the positioning unit. The positioning unit locates the location of the abnormal software operation fault, and the fault unit calls the repair strategy from the repair strategy library. The fault unit uses the called repair strategy to repair the fault.
9. The AI-based software operation anomaly analysis system according to claim 2, characterized in that: The fault unit includes an isolation module and a temporary module. The isolation module directs the traffic of the abnormal software to the backup node, and the isolation module points the domain name resolution of the fault area to the healthy area, marks the faulty Pod as unschedulable, and the temporary module restarts the software node where the fault problem occurs and rolls back the code data.