Database cluster health degree detection method and device, equipment, storage medium and product

Abstract

The invention provides a database cluster health degree detection method and device, equipment, a storage medium and a product. The method comprises the following steps: acquiring actual bearing pressure, used duration and actual service pressure of each physical node in a first time period; for each physical node, performing failure analysis on the physical node according to the used duration and the service life of the physical node to obtain a bearing pressure prediction result of the physical node in the second time period; for each physical node, predicting the service pressure of the physical node in the second time period according to the actual service pressure of the physical node to obtain a service pressure prediction result of the physical node; and determining the health degree of the database cluster according to the actual bearing pressure, the actual service pressure, the bearing pressure prediction result and the service pressure prediction result of each physical node. According to the method, various pressure-related data of the database cluster are comprehensively considered, and the health degree of the database cluster can be comprehensively evaluated from multiple aspects.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a database cluster health detection method, device, equipment, storage medium and product. Background Art

[0002] A database cluster is a virtual single database logical image formed by at least two or more database servers. It can provide data services to clients just like a system with a single database server. Since there are multiple database servers in a database cluster, it has a higher fault tolerance than a centralized database environment. Failure of certain nodes in the database cluster may not affect the use of the database cluster. However, it may also lead to minor faults in the database cluster not being noticed and handled. Once a major fault that affects the use of the database occurs, the various minor faults accumulated before will make it difficult to locate the major fault, slowing down the efficiency of fault handling.

[0003] Therefore, it is necessary to monitor the database cluster in real time and evaluate the health of the database cluster so that operation and maintenance personnel can intervene in the major failures that are developing in the cluster earlier. However, the health detection methods in the existing technology are difficult to cope with the complex trend of operating status changes and cannot comprehensively evaluate the health of the cluster. Summary of the invention

[0004] The embodiments of the present application provide a database cluster health detection method, apparatus, device, storage medium and product, which can comprehensively evaluate the health of a database cluster.

[0005] In a first aspect, the present application provides a method for detecting the health of a database cluster, wherein the database cluster includes a plurality of physical nodes, and the method includes:

[0006] Obtaining the actual load bearing pressure, used time and actual business pressure of each physical node in the first time period;

[0007] For each physical node, according to the usage time and service life of the physical node, a failure analysis is performed on the physical node to obtain a prediction result of the bearing pressure of the physical node in the second time period;

[0008] For each physical node, predicting the business pressure of the physical node in the second time period according to the actual business pressure of the physical node, and obtaining a business pressure prediction result of the physical node;

[0009] The health of the database cluster is determined according to the actual load-bearing pressure of each physical node, the actual business pressure, the load-bearing pressure prediction result, and the business pressure prediction result.

[0010] In some possible implementations, predicting, for each physical node, the business pressure of the physical node in the second time period according to the actual business pressure of the physical node to obtain the business pressure prediction result of the physical node includes:

[0011] The actual business pressure is input into a preset long short-term memory network for calculation to obtain a business pressure prediction result of the physical node.

[0012] In some possible implementations, before inputting the actual service pressure into a preset long short-term memory network for calculation to obtain a service pressure prediction result of the physical node, the method further includes:

[0013] Performing Fourier transform on the actual business pressure to obtain a frequency domain representation of the actual business pressure;

[0014] Expanding the frequency range of the frequency domain representation to obtain expanded frequency domain data;

[0015] updating the actual business pressure based on an inverse Fourier transform result of the expanded frequency domain data;

[0016] The inputting the actual business pressure into a preset long short-term memory network for calculation to obtain the business pressure prediction result of the physical node includes:

[0017] The updated actual business pressure is input into a preset long short-term memory network for calculation to obtain a business pressure prediction result of the physical node.

[0018] In some possible implementations, before inputting the actual service pressure into a preset long short-term memory network for calculation to obtain a service pressure prediction result of the physical node, the method further includes:

[0019] Combining historical business pressure data of a physical node in multiple historical time periods into a sequence form in chronological order to obtain a business pressure data sequence, wherein the historical time period is before the first time period;

[0020] For the historical business pressure in each time period in the business pressure data series, the following steps are performed respectively:

[0021] Inputting the historical business pressure corresponding to the target time period into the long short-term memory network, and predicting the historical business pressure corresponding to the next time period of the target time period; the target time period is any time period among the multiple historical time periods;

[0022] Determining a loss function value of the long short-term memory network according to the historical business pressure corresponding to the next time period;

[0023] Determine whether the loss function value of the long short-term memory network meets a preset training stop condition;

[0024] If the condition is not met, the model parameters of the long short-term memory network are adjusted, and the adjusted long short-term memory network is trained using the business pressure data sequence, and the historical business pressure corresponding to the target time period is returned to the long short-term memory network to predict the historical business pressure corresponding to the next time period of the target time period until the preset training stop condition is met to obtain the trained long short-term memory network.

[0025] In some possible implementations, performing a failure analysis on the physical node according to the usage time and service life of the physical node to obtain a prediction result of the bearing pressure of the physical node in the second time period includes:

[0026] Determine the failure stage corresponding to the physical node according to the usage time and service life of the physical node;

[0027] determining a parameter value of a Weibull distribution based on the failure stage;

[0028] According to the parameter value, the failure probability corresponding to the physical node is calculated according to Weibull distribution;

[0029] According to the failure probability, a prediction result of the bearing pressure of the physical node in the second time period is calculated.

[0030] In some possible implementations, determining the failure stage corresponding to the physical node according to the usage time and service life of the physical node includes:

[0031] Based on the service life of the physical node, determine a first time threshold and a second time threshold corresponding to the physical node;

[0032] When the usage time of the physical node is less than the first time threshold, determining that the failure stage corresponding to the physical node is an early failure period;

[0033] When the usage time of the physical node is greater than the first time threshold and less than the second time threshold, determining that the failure stage corresponding to the physical node is a random failure period;

[0034] When the usage time of the physical node is greater than the second time threshold, it is determined that the failure stage corresponding to the physical node is a degradation failure period.

[0035] In some possible implementations, determining the health of the database cluster according to the actual load-bearing pressure of each physical node, the actual business pressure, the load-bearing pressure prediction result, and the business pressure prediction result includes:

[0036] For each physical node, determining the smaller value between the actual bearing pressure corresponding to the physical node and the bearing pressure prediction result as the bearing pressure health degree corresponding to the physical node;

[0037] For each physical node, determining the smaller value between the actual business pressure corresponding to the physical node and the business pressure prediction result as the business pressure health degree corresponding to the physical node;

[0038] Calculate the average value of the load pressure health of each physical node to obtain the load pressure health of the database cluster;

[0039] Calculate the average value of the load-bearing health of each physical node to obtain the business pressure health of the database cluster;

[0040] A smaller value between the load pressure health of the database cluster and the business pressure health of the database cluster is determined as the health of the database cluster.

[0041] In a second aspect, the present application further provides a database cluster health detection device, the device comprising:

[0042] An acquisition module, used to acquire the actual load bearing pressure, the used time and the actual business pressure of each physical node in the first time period;

[0043] An analysis module is used to perform a failure analysis on each of the physical nodes according to the usage time and service life of the physical node, so as to obtain a prediction result of the bearing pressure of the physical node in a second time period;

[0044] A prediction module, configured to predict the business pressure of each physical node in a second time period according to the actual business pressure of the physical node, and obtain a business pressure prediction result of the physical node;

[0045] A determination module is used to determine the health of the database cluster according to the actual load-bearing pressure of each physical node, the actual business pressure, the load-bearing pressure prediction result and the business pressure prediction result.

[0046] In a third aspect, the present application provides a database cluster health detection device, the device comprising: a processor, and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement the database cluster health detection method described above.

[0047] In a fourth aspect, the present application provides a computer-readable storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, the database cluster health detection method as described above is implemented.

[0048] In a fifth aspect, the present application provides a computer program product. When instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the database cluster health detection method described above.

[0049] The database cluster health detection method, device, equipment, storage medium and product provided in the embodiment of the present application obtain the actual load-bearing pressure, usage time and actual business pressure of each physical node in the database cluster in the first time period. Afterwards, for each physical node, the physical node is subjected to failure analysis according to the usage time and service life of the physical node, and the business pressure of the physical node in the second time period is predicted based on the actual business pressure, and the load-bearing pressure prediction result and the business pressure prediction result of the physical node in the second time period are obtained. Finally, the actual load-bearing pressure, the actual business pressure, the load-bearing pressure prediction result and the business pressure prediction result of each physical node are combined to determine the health of the database cluster. The embodiment of the present application takes into account the two aspects of load-bearing pressure and business pressure, and performs pressure prediction on both the load-bearing pressure and business pressure, comprehensively considers the various pressure-related data of the database cluster, and can comprehensively evaluate the health of the database cluster from multiple aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The present application can be better understood from the following description of the specific embodiments of the present application in conjunction with the accompanying drawings, in which:

[0051] Other features, objects and advantages of the present application will become more apparent by reading the following detailed description of non-limiting embodiments with reference to the accompanying drawings, in which the same or similar reference numerals represent the same or similar features.

[0052] Figure 1 This is a flowchart of a database cluster health detection method provided by an embodiment of the present application;

[0053] Figure 2 is a flowchart of a database cluster health detection method provided by another embodiment of the present application;

[0054] Figure 3 This is a flowchart of a next time period indicator prediction process provided by an embodiment of the present application;

[0055] Figure 4It is a structural diagram of a database cluster health detection device provided by an embodiment of the present application;

[0056] Figure 5 It is a hardware structure diagram of the database cluster health detection device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0057] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0058] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0059] Compared with centralized database environments, database clusters have higher fault tolerance and self-troubleshooting capabilities. Failures of certain nodes in the cluster may not affect the use of the cluster. However, correspondingly, when a major failure that affects use occurs, it may be difficult to locate the failure. Therefore, during the operation and maintenance of the database cluster, many indicators need to be monitored and judged in real time. The database cluster health judgment method in the prior art cannot take into account the specific operating conditions of each node in the cluster, is difficult to cope with situations with complex changing trends, and cannot comprehensively evaluate the real-time health of the cluster and its changing trends.

[0060] The inventors have studied the above-mentioned problems and found that if a comprehensive evaluation of the health of the database cluster operation is conducted by combining real-time monitoring indicators and predictive indicators, it can help operation and maintenance personnel filter out minor faults that do not require intervention and intervene earlier in major faults that are developing in the cluster, thereby better assessing the health of the system and improving the stability and reliability of the system.

[0061] In order to solve the problems of the prior art, the embodiments of the present application provide a database cluster health detection method, device, equipment, storage medium and product. The following first introduces the database cluster health detection method provided by the embodiments of the present application.

[0062] Figure 1 The flowchart of a database cluster health detection method provided by an embodiment of the present application is shown. In this embodiment, the database cluster includes multiple physical nodes, and the method can be executed by the above-mentioned database cluster health detection device, such as Figure 1 As shown, the method includes the following steps S101 to S104.

[0063] Step S101: Obtain the actual load-bearing pressure, used time and actual business pressure of each physical node in the first time period.

[0064] In a specific implementation, a preset monitoring tool or performance analysis software is used to collect the operating data of the physical node. The actual load-bearing pressure of the physical node is obtained based on the actual operating conditions of the physical node in the first time period, and the length of time since the physical node began to be used is obtained as the length of time the physical node has been used. The actual business pressure of the physical node is obtained based on the actual business requests processed by the physical node in the first time period.

[0065] As another example, for example, you can pre-define the node indicators that need to be monitored, such as the CPU usage, memory usage, disk input / output (I / O), etc., and then summarize various data of each physical node according to this node indicator. The actual load-bearing pressure is obtained according to the hardware-related indicator data of the node, the actual business pressure is obtained according to the indicator data at the application level of the node, and the use time of the physical node is determined according to the startup time and current time of the physical node in the record.

[0066] Step S102: For each physical node, a failure analysis is performed on the physical node according to the usage time and service life of the physical node to obtain a bearing pressure prediction result of the physical node in a second time period, where the second time period is after the first time period.

[0067] According to the usage time and service life of the physical node, a failure analysis is performed on the physical node to obtain a bearing pressure prediction result of the physical node in the second time period.

[0068] In the specific implementation, a failure analysis model is pre-established. Based on this failure analysis model and the usage time and historical performance data of the physical node, the expected performance degradation rate, resource utilization rate growth rate, etc. are calculated to estimate the future reliability and performance status of the node, and predict the load bearing pressure of the physical node in the second time period. The second time period is later than the above-mentioned first time period.

[0069] As another example, the failure risk of the physical node can be predicted based on the usage time, historical performance data, and possible failure modes of the physical node, using corresponding statistical methods or machine learning models as failure analysis models.

[0070] Step S103: for each physical node, predict the business pressure of the physical node in the second time period according to the actual business pressure of the physical node, and obtain a business pressure prediction result of the physical node.

[0071] Based on the actual business pressure, the business pressure of the physical node in the second time period is predicted to obtain a prediction result of the business pressure of the physical node.

[0072] In the specific implementation, business-related indicators that reflect the ability of physical nodes to handle actual business loads can be collected, such as business request volume, response time, error rate, etc. Then, according to the business pressure prediction model, combined with the current business indicator data, the business pressure faced by the physical node in the second time period is predicted.

[0073] As another example, a specific business pressure prediction model may use time series analysis or a machine learning model to analyze historical data of business indicators to predict future business load trends.

[0074] Step S104: determining the health of the database cluster according to the actual load-bearing pressure of each physical node, the actual business pressure, the load-bearing pressure prediction result and the business pressure prediction result.

[0075] In the specific implementation, the health assessment algorithm is executed in combination with the actual load pressure data, the load pressure prediction results, the actual business pressure data and the business pressure prediction results to comprehensively consider the actual load status and prediction of each node, determine the health of the above-mentioned database cluster, and score or mark the health of the database cluster.

[0076] As another example, the specific health assessment algorithm may be a threshold detection rule or a classifier based on machine learning.

[0077] The database cluster health detection method provided in the embodiment of the present application obtains the actual load-bearing pressure, usage time and actual business pressure of each physical node in the database cluster in the first time period. Afterwards, for each physical node, a failure analysis is performed on the above-mentioned physical node according to the usage time and service life of the physical node, and based on the above-mentioned actual business pressure, the business pressure of the physical node in the second time period is predicted to obtain the load-bearing pressure prediction result and the business pressure prediction result of the physical node in the second time period, and finally the above-mentioned actual load-bearing pressure, the above-mentioned actual business pressure, the above-mentioned load-bearing pressure prediction result and the above-mentioned business pressure prediction result of each physical node are combined to determine the health of the above-mentioned database cluster. Taking into account the various pressure-related data of the database cluster, the health of the database cluster can be comprehensively evaluated from multiple aspects.

[0078] In order to obtain a more accurate prediction result of the service pressure of the physical node, in some implementations, the above step 103 may include:

[0079] The actual business pressure is input into a preset long short-term memory network for calculation to obtain a business pressure prediction result of the physical node.

[0080] In a specific implementation, the actual business pressure is converted into a format suitable for a long short-term memory network, and then input into a trained preset long short-term memory network, which generates a final business pressure prediction result of the physical node.

[0081] The above-mentioned method of the embodiment of the present application inputs the actual business pressure into a preset long short-term memory network for calculation, and the long short-term memory network is used to predict the business pressure prediction result of the physical node, so as to obtain a more accurate business pressure prediction result of the physical node.

[0082] In order to improve the prediction capability of the long short-term memory network for high-frequency changes in business pressure, in some embodiments, before inputting the actual business pressure into the preset long short-term memory network for calculation to obtain the business pressure prediction result of the physical node, the method may further include steps A1 to C1:

[0083] Step A1: Perform Fourier transform on the actual business pressure to obtain a frequency domain representation of the actual business pressure.

[0084] In a specific implementation, the data of the actual business pressure is organized into a time series form, and the actual business pressure is converted using Fourier transform to obtain a frequency domain representation of the actual business pressure.

[0085] As another example, the frequency domain representation of the actual business pressure may include the amplitude of each frequency component, the phase information of each frequency component and the actual frequency value of each frequency component.

[0086] Step B1: Expand the frequency range of the above frequency domain representation to obtain expanded frequency domain data.

[0087] In a specific implementation, zero padding may be performed on the low-frequency or high-frequency part of the above frequency domain representation to obtain extended frequency domain data.

[0088] Step C1: based on the inverse Fourier transform result of the expanded frequency domain data, the actual business pressure is updated.

[0089] In a specific implementation, the expanded frequency domain data is converted back to the time domain to obtain new business pressure data after frequency domain expansion and processing. The above actual business pressure is updated based on the business pressure data.

[0090] The actual business pressure is input into the preset long short-term memory network for calculation to obtain the business pressure prediction result of the physical node, including:

[0091] The updated actual business pressure is input into the preset long short-term memory network for calculation to obtain the business pressure prediction result of the above physical node.

[0092] The above method of the embodiment of the present application converts the actual business pressure into a frequency domain representation by using Fourier transform, then expands the frequency range of the above frequency domain representation to obtain the expanded frequency domain data, and then performs an inverse Fourier transform based on the above expanded frequency domain data, and updates the above actual business pressure according to the inverse Fourier transform result. Thus, the updated actual business pressure is used for prediction, and the business pressure prediction result of the physical node is obtained, which can improve the prediction ability of the long short-term memory network for high-frequency changes in business pressure.

[0093] In order to quickly complete the training of the long short-term memory network, in some embodiments, before the actual business pressure is input into the preset long short-term memory network for calculation to obtain the business pressure prediction result of the physical node, the method may further include steps A2 to E2:

[0094] Step A2: combining historical business pressure data of a physical node in multiple historical time periods into a sequence in chronological order to obtain a business pressure data sequence, wherein the historical time period is before the first time period.

[0095] In a specific implementation, actual business pressure data in multiple time periods are obtained from a physical node. The actual business pressure data in multiple time periods are sorted in chronological order. After the chronological order is completed, it can be serialized into a time series form to obtain a business pressure data sequence. The data of multiple time periods constitute a business pressure data sequence in a serial form.

[0096] As another example, after obtaining actual business pressure data in multiple time periods from a physical node, the obtained actual business pressure data can be cleaned to remove outliers or incomplete data, and the data can be aligned through timestamps or time series indexes to ensure that data in different time periods can be aligned on the time axis.

[0097] For the historical business pressure in each time period in the business pressure data sequence, the following steps B2 to E2 are performed respectively:

[0098] Step B2: inputting the historical business pressure corresponding to the target time period into the long short-term memory network to predict the historical business pressure corresponding to the next time period of the target time period; the target time period is any time period among the multiple historical time periods.

[0099] In a specific implementation, the actual business pressure data of the target time period is input into the long short-term memory network that needs to be trained, and the long short-term memory network predicts the actual business pressure corresponding to the next time period of the target time period based on the input data.

[0100] As another example, the actual business pressure data may be normalized to a suitable range, and the data may be divided into sequences according to time steps.

[0101] Step C2: Determine the loss function value of the long short-term memory network according to the historical business pressure corresponding to the next time period.

[0102] In a specific implementation, the actual business pressure data corresponding to the next time period of the target time period is obtained from the business pressure data sequence, and the value of the loss function is calculated using the difference between the predicted value and the actual value.

[0103] As another example, mean square error or mean absolute error, etc., may be used as the loss function to calculate the loss function value.

[0104] Step D2: Determine whether the loss function value of the long short-term memory network meets the preset training stop condition.

[0105] In the specific implementation, a training stop condition is preset, such as reaching the maximum training round, the loss function value drops to a certain threshold, etc. After calculating the loss function value of the long short-term memory network, check whether the current loss function value meets the preset training stop condition.

[0106] Step E2: If the conditions are not met, the model parameters of the long short-term memory network are adjusted, and the adjusted long short-term memory network is trained using the business pressure data sequence, and the historical business pressure corresponding to the target time period is input into the long short-term memory network to predict the historical business pressure corresponding to the next time period of the target time period, until the preset training stop condition is met, and the trained long short-term memory network is obtained.

[0107] In the specific implementation, if the current loss function value does not meet the stopping condition, adjust the parameters in the long short-term memory network. Use the entire business pressure data sequence to continue training the adjusted long short-term memory network. Use the adjusted model again for prediction in step B2 to obtain a new prediction result and loss function value. Repeat steps C2 and D2 until the loss function value meets the preset training stop condition.

[0108] The above-mentioned method of the embodiment of the present application obtains a business pressure data sequence by combining the data of actual business pressure of the physical node in multiple time periods into a sequence form in chronological order, and then trains the long short-term memory network according to this business pressure data sequence, so that the long short-term memory network predicts the actual business pressure corresponding to the next time period based on the actual business pressure corresponding to each time period in the pressure data sequence, and can quickly complete the training of the long short-term memory network.

[0109] In order to accurately obtain the bearing pressure prediction result, in some embodiments, the above step 102 may include steps A3 to D3:

[0110] Step A3: Determine the failure stage corresponding to the physical node according to the usage time and service life of the physical node.

[0111] In a specific implementation, the usage time and service life data of each physical node are obtained, and the current failure stage of each physical node is determined according to the ratio of the usage time of the physical node to the service life of the physical node.

[0112] Step B3: Determine the parameter value of the Weibull distribution based on the above failure stages.

[0113] In the specific implementation, each failure stage corresponds to a specific Weibull distribution parameter. For example, different failure stages may have different shape parameters and scale parameters. Through the pre-defined mapping rules, the corresponding Weibull distribution parameter value is determined according to the current failure stage of the physical node.

[0114] Step C3: According to the above parameter values, the failure probability corresponding to the above physical node is calculated according to Weibull distribution.

[0115] In a specific implementation, the Weibull distribution parameter value determined in step B3 is used to determine the corresponding Weibull distribution probability density function, thereby calculating the probability value of the physical node failing in the next time period, that is, the failure probability corresponding to the physical node.

[0116] Step D3: According to the above failure probability, the bearing pressure prediction result of the above physical node in the second time period is calculated.

[0117] In a specific implementation, a system load analysis is performed based on the failure probability calculated in step C3 to determine the workload and efficiency expectations of the node, and to evaluate the carrying capacity and expected performance of the physical node in the second time period. Combining the failure probability and the predicted load demand, a prediction result of the bearing pressure of the physical node in the second time period is generated.

[0118] The above method of the embodiment of the present application determines the failure stage corresponding to the above physical node through the usage time and service life of the physical node, and then determines the parameter value of the Weibull distribution corresponding to the failure stage, and calculates the failure probability corresponding to the physical node according to the parameter value and the Weibull distribution, and then obtains the load-bearing pressure prediction result of the physical node in the second time period. Using the Weibull distribution, the load-bearing pressure prediction result can be accurately obtained.

[0119] In order to accurately determine the failure stage corresponding to the physical node, in some implementations, the above step A3 may include steps A4 to D4:

[0120] Step A4: Based on the service life of the physical node, determine a first time threshold and a second time threshold corresponding to the physical node.

[0121] In a specific implementation, based on the service life of the physical node and according to a set service life rule, the first time threshold and the second time threshold of each node are determined. The service life rule can be determined based on the node type.

[0122] Step B4: when the usage time of the physical node is less than or equal to the first time threshold, determine that the failure stage corresponding to the physical node is an early failure period.

[0123] In a specific implementation, for each physical node, check whether its current used time is less than or equal to the first time threshold. If the used time of the node is less than or equal to the first time threshold, the failure stage of the node is determined to be an early failure period. This state usually means that the node is still in the early warning stage of possible production failure or in the early stage of its service life.

[0124] Step C4: when the usage time of the physical node is greater than the first time threshold and less than the second time threshold, determine that the failure stage corresponding to the physical node is a random failure period.

[0125] In a specific implementation, for each physical node, check whether its current usage time is greater than or equal to the first time threshold and less than the second time threshold. If the usage time of the node meets the above conditions, the failure stage of the node is determined as a random failure period. This means that the failure probability of the node is in the middle stage, and there is a high possibility of random failure.

[0126] Step D4: when the usage time of the physical node is greater than or equal to the second time threshold, determine that the failure stage corresponding to the physical node is a degradation failure period.

[0127] In a specific implementation, for each physical node, check whether its current usage time is greater than or equal to the second time threshold. If the usage time of the node is greater than or equal to the second time threshold, the failure stage of the node is determined as the degradation failure period. This indicates that the node is approaching or has exceeded the end stage of its design life and is more likely to fail due to reaching its service life.

[0128] The above-mentioned method of the embodiment of the present application determines the first time threshold and the second time threshold corresponding to the physical node based on the service life of the physical node, and then compares the usage time of the physical node with the first time threshold and the second time threshold to determine the failure stage corresponding to the physical node, so as to accurately determine the failure stage corresponding to the physical node.

[0129] In order to accurately determine the health of the database cluster, in some implementations, the above S104 may include steps A5 to E5:

[0130] Step A5: For each physical node, the smaller value between the actual bearing pressure corresponding to the physical node and the bearing pressure prediction result is determined as the bearing pressure health degree corresponding to the physical node.

[0131] In a specific implementation, for each physical node, the smaller value between the actual bearing pressure and the bearing pressure prediction result is determined as the bearing pressure health of the physical node.

[0132] Step B5: For each physical node, a smaller value between the actual business pressure corresponding to the physical node and the business pressure prediction result is determined as the business pressure health level corresponding to the physical node.

[0133] In a specific implementation, for each physical node, the smaller value between the actual business pressure and the business pressure prediction result is determined as the business pressure health of the physical node.

[0134] Step C5: Calculate the average value of the load-bearing health of each physical node to obtain the load-bearing health of the database cluster.

[0135] In a specific implementation, the load-bearing health of all physical nodes is added up and then divided by the total number of all physical nodes, and the average is calculated to obtain the load-bearing health of the database cluster.

[0136] Step D5: Calculate the average value of the load-bearing health of each physical node to obtain the business pressure health of the database cluster.

[0137] In a specific implementation, the business pressure health of all physical nodes is added up and then divided by the total number of all physical nodes, and the average is calculated to obtain the business pressure health of the database cluster.

[0138] Step E5: The smaller value between the load pressure health of the database cluster and the business pressure health of the database cluster is determined as the health of the database cluster.

[0139] In a specific implementation, the load pressure health and business pressure health of the database cluster are compared, and the smaller value between the load pressure health and business pressure health of the database cluster is determined as the health of the database cluster.

[0140] The above-mentioned method of the embodiment of the present application determines, for each physical node, the smaller value between the above-mentioned actual bearing pressure corresponding to the physical node and the above-mentioned bearing pressure prediction result as the bearing pressure health degree corresponding to the physical node; and determines the smaller value between the above-mentioned actual business pressure corresponding to the physical node and the above-mentioned business pressure prediction result as the business pressure health degree corresponding to the physical node. Then, the average value of the bearing pressure health degree in the database cluster is taken as the bearing pressure health degree of the database cluster, and the average value of the business pressure health degree in the database cluster is taken as the business pressure health degree of the database cluster. Finally, the smaller value between the bearing pressure health degree of the above-mentioned database cluster and the business pressure health degree of the above-mentioned database cluster is determined as the health degree of the above-mentioned database cluster, which can comprehensively consider the situation of each physical node in the database cluster to accurately determine the health degree of the database cluster.

[0141] In one embodiment of the present application, a method for detecting the health of a database cluster is executed in a database cluster health detection device, which is composed of a load pressure real-time monitoring module, a load pressure prediction module, a load pressure calculation module, a business pressure real-time monitoring module, a business pressure prediction module, a load pressure calculation module and a database cluster health calculation module.

[0142] The load pressure real-time monitoring module is used to detect whether the read and write functions of the physical nodes in the cluster are normal. The load pressure prediction module predicts the probability of physical nodes failing in a certain period of time in the future. The load pressure health calculation module calculates the health of the load pressure based on the detection value of the load pressure real-time monitoring module and the prediction value of the load pressure prediction module. The business pressure real-time monitoring module is used to detect various business pressure indicators of the physical nodes in the cluster, such as CPU usage, memory usage, etc. The business pressure prediction module predicts the business pressure in the next period based on the historical data of the business pressure real-time monitoring module. The business pressure health calculation module calculates the health of the business pressure based on the detection value of the business pressure real-time monitoring module and the prediction value of the business pressure prediction module. Finally, the database cluster health calculation module conducts a comprehensive evaluation of the business pressure health and the load pressure health to obtain the overall health score of the database cluster.

[0143] You can refer to Figure 2 In the method for detecting the health of a database cluster implemented in this embodiment, step S201: obtaining the load pressure of the current cluster from the database cluster real-time monitoring module. The measured load pressure of a single physical node is Indicates. i represents the physical node number, ranging from 1 to N, where N is the total number of physical nodes. Check the read and write functions of the physical node. If the read and write functions of the physical node are normal, then otherwise,

[0144] Step S202: Predict the load capacity in the future based on the physical node usage and Weibull distribution. The predicted load capacity of a single physical node is Represents. According to the use time of the physical node, it is divided into three stages: early failure period, random failure period, and degradation failure period. The early failure period is defined as 0~T1, and T1 can be set as the first warranty period of the physical node, such as 12 months, 36 months, etc. The causes of early failure include imperfect production, unsuitable use environment, etc. In this stage, as the normal use time increases, the failure probability gradually decreases. The random failure period is defined as T1~T2, and T2 can be set as the general scrapping time limit of the physical node, such as 60 months. The causes of random failure are usually misuse, fluctuations in the use environment, etc. In this stage, as the use time increases, the failure probability remains basically unchanged. The degradation failure period is defined as T2~∞, and the causes of degradation failure usually include aging of components, performance degradation, etc. In this stage, as the use time increases, the failure probability increases. Then use the Weibull distribution to calculate the failure probability of the i-th physical node in a certain time period (t, t+Δt) in the future. The calculation formula is:

[0145]

[0146] Among them, β is called the shape parameter, η is called the scaling parameter, and γ is called the location parameter. The early failure period is 0<β<1, the random failure period β is 1, and the degradation failure period β>1. γ can be set to 0. η can be adjusted according to the definition of the three failure stages of the physical node.

[0147] The predicted load bearing capacity of the ith physical node is defined as:

[0148]

[0149] Step S203: Calculate the bearing pressure health according to the real-time monitored bearing pressure and the predicted bearing pressure. The bearing pressure health of the i-th physical node is defined as:

[0150] Step S204: Obtain the business pressure of the current cluster from the database cluster real-time monitoring module. The business pressure index of the ith physical node is composed of multiple indicators, such as CPU usage, I / O port usage, network interface usage, etc. Assume that there are M indicators in total, the database cluster monitoring cycle is a unit time of 1, and each t detection cycles is an evaluation time period. Calculate the average value of the jth indicator (such as the average CPU usage) within t detection cycles. The average value of the measured value of the jth indicator in the kth evaluation time period is expressed as Where i = 1, 2…i; j = 1, 2…M. The measured service pressure of the i-th node is expressed as express, Defined as:

[0151] Step S205: predict the business pressure in the future based on the historical business pressure data. Specifically, the index prediction value at time k+1 is expressed as The predicted business pressure of the i-th node is express, Defined as: Specifically, the above The calculation process can be referred to Figure 3 First, execute step S301 to form a sequence with the average values ​​of the indicators in the K evaluation time periods. The sequence is in the following form:

[0152]

[0153] Afterwards, the parameters of the long short-term memory network (LSTM) module are randomly initialized.

[0154] Step S302: frequency domain expansion. In order to improve the LSTM module’s ability to predict high-frequency changes in business pressure, Perform frequency domain expansion to obtain the expanded for:

[0155]

[0156] Where L is a positive integer and can be set to 5.

[0157] Step S303: As the input of LSTM, calculate the state of LSTM (h k ) and output (y k ), and get the average value of the jth indicator in the next period. The specific calculation formula is:

[0158]

[0159] Where σ is the Sigmoid function. Using the gradient descent method, minimize the output and The mean square error between:

[0160]

[0161] Thereby optimizing the parameters W of the LSTM module f, W g, W q, W y, b f ,b g ,b q ,b y .in ⊙ represents the multiplication of elements at corresponding positions in the matrix, and [·;·] represents the concatenation of two column vectors.

[0162] Step S206: Calculate the business pressure health according to the real-time detected business pressure and the predicted business pressure. Specifically, the business pressure health of the i-th physical node is defined as:

[0163] Step S207: Calculate the health of the cluster based on the load pressure health and the business pressure health. Specifically, the load pressure health of the entire cluster is defined as:

[0164]

[0165] Where N is the number of nodes in the cluster, c i Indicates the bearing pressure health of the i-th node.

[0166] The business pressure health of the entire cluster is defined as:

[0167]

[0168] Cluster health is defined as:

[0169] H=min(C,O)

[0170] The minimum value of the load bearing health and the business capacity health is used as the cluster health here because if either health is 0, the cluster will be unusable. Therefore, the cluster health should be the lower limit of the two.

[0171] Based on the database cluster health detection method provided in the above embodiment, the present application also provides a specific implementation of a database cluster health detection device. Please refer to the following embodiment.

[0172] See first Figure 4 The database cluster health detection device 400 provided in the embodiment of the present application includes the following modules:

[0173] The acquisition module 401 is used to acquire the actual load bearing pressure, the used time and the actual business pressure of each physical node in the first time period.

[0174] The analysis module 402 is used to perform failure analysis on each physical node according to the usage time and service life of the physical node, and obtain the load-bearing pressure prediction result of the physical node in a second time period, and the second time period is after the first time period.

[0175] The prediction module 403 is used to predict the business pressure of each physical node in the second time period according to the actual business pressure of the physical node, and obtain the business pressure prediction result of the physical node.

[0176] The determination module 404 is used to determine the health of the database cluster according to the actual load-bearing pressure of each physical node, the actual business pressure, the load-bearing pressure prediction result and the business pressure prediction result.

[0177] The database cluster health detection device provided in the embodiment of the present application obtains the actual load-bearing pressure, usage time and actual business pressure of each physical node in the database cluster in the first time period. Afterwards, for each physical node, a failure analysis is performed on the above-mentioned physical node according to the usage time and service life of the physical node, and based on the above-mentioned actual business pressure, the business pressure of the physical node in the second time period is predicted to obtain the load-bearing pressure prediction result and the business pressure prediction result of the physical node in the second time period, and finally the above-mentioned actual load-bearing pressure, the above-mentioned actual business pressure, the above-mentioned load-bearing pressure prediction result and the above-mentioned business pressure prediction result of each physical node are combined to determine the health of the above-mentioned database cluster. Taking into account the various pressure-related data of the database cluster, the health of the database cluster can be comprehensively evaluated from multiple aspects.

[0178] As an implementation of the present application, the database cluster health detection device 400 may further include:

[0179] The calculation module is used to input the above actual business pressure into a preset long short-term memory network for calculation to obtain the business pressure prediction result of the above physical node.

[0180] The above method of the embodiment of the present application inputs the above actual business pressure into a preset long short-term memory network for calculation, and the long short-term memory network is used to predict the business pressure prediction result of the above physical node, so as to obtain a more accurate business pressure prediction result of the physical node.

[0181] As an implementation of the present application, the database cluster health detection device 400 may further include:

[0182] The transformation module is used to perform Fourier transformation on the actual business pressure to obtain a frequency domain representation of the actual business pressure.

[0183] The expansion module is used to expand the frequency range of the above frequency domain representation to obtain expanded frequency domain data.

[0184] The transformation module is also used to update the actual business pressure based on the inverse Fourier transform result of the expanded frequency domain data.

[0185] The calculation module is used to input the updated actual business pressure into a preset long short-term memory network for calculation to obtain the business pressure prediction result of the above physical node.

[0186] The above method of the embodiment of the present application converts the actual business pressure into a frequency domain representation by using Fourier transform, then expands the frequency range of the above frequency domain representation to obtain the expanded frequency domain data, and then performs inverse Fourier transform based on the above expanded frequency domain data, and updates the above actual business pressure according to the inverse Fourier transform result, thereby using the updated actual business pressure for prediction, and obtaining the business pressure prediction result of the physical node. It can improve the prediction ability of the long short-term memory network for high-frequency changes in business pressure.

[0187] As an implementation of the present application, the database cluster health detection device 400 may further include:

[0188] The combination module is used to combine the historical business pressure data of the physical node in multiple historical time periods into a sequence form in chronological order to obtain a business pressure data sequence.

[0189] The input module is used to input the historical business pressure corresponding to the target time period into the long short-term memory network, and predict the historical business pressure corresponding to the next time period of the target time period.

[0190] A determination module is used to determine the loss function value of the long short-term memory network according to the historical business pressure corresponding to the next time period.

[0191] The judgment module is used to judge whether the loss function value of the long short-term memory network meets the preset training stop condition.

[0192] The adjustment module is used to adjust the model parameters of the long short-term memory network when the conditions are not met, and use the business pressure data sequence to train the adjusted long short-term memory network, return the historical business pressure corresponding to the target time period to the long short-term memory network, and predict the historical business pressure corresponding to the next time period of the target time period until the preset training stop condition is met to obtain the trained long short-term memory network.

[0193] The above-mentioned method of the embodiment of the present application obtains a business pressure data sequence by combining the data of actual business pressure of the physical node in multiple time periods into a sequence form in chronological order, and then trains the long short-term memory network according to this business pressure data sequence, so that the long short-term memory network predicts the actual business pressure corresponding to the next time period based on the actual business pressure corresponding to each time period in the pressure data sequence, and can quickly complete the training of the long short-term memory network.

[0194] As an implementation of the present application, the database cluster health detection device 400 may further include:

[0195] The determination module is used to determine the failure stage corresponding to the physical node according to the usage time and service life of the physical node.

[0196] The determination module is also used to determine the parameter value of the Weibull distribution based on the above failure stage.

[0197] The calculation module is used to calculate the failure probability corresponding to the above physical node according to the above parameter values ​​and Weibull distribution.

[0198] The calculation module is also used to calculate the bearing pressure prediction result of the above-mentioned physical node in the second time period according to the above-mentioned failure probability.

[0199] The above method of the embodiment of the present application determines the failure stage corresponding to the above physical node through the usage time and service life of the physical node, and then determines the parameter value of the Weibull distribution corresponding to the failure stage, and calculates the failure probability corresponding to the physical node according to the parameter value and the Weibull distribution, and then obtains the load-bearing pressure prediction result of the physical node in the second time period. Using the Weibull distribution, the load-bearing pressure prediction result can be accurately obtained.

[0200] As an implementation of the present application, the database cluster health detection device 400 may further include:

[0201] The determination module is used to determine a first time threshold and a second time threshold corresponding to the physical node based on the service life of the physical node.

[0202] The determination module is further used to determine that the failure stage corresponding to the physical node is an early failure period when the usage time of the physical node is less than or equal to the first time threshold.

[0203] The determination module is also used to determine that the failure stage corresponding to the above physical node is a random failure period when the usage time of the physical node is greater than the first time threshold and less than the second time threshold.

[0204] The determination module is further used to determine that the failure stage corresponding to the above-mentioned physical node is a degradation failure period when the usage time of the physical node is greater than or equal to the second time threshold.

[0205] The above-mentioned method of the embodiment of the present application determines the first time threshold and the second time threshold corresponding to the physical node based on the service life of the physical node, and then compares the usage time of the physical node with the first time threshold and the second time threshold to determine the failure stage corresponding to the physical node, so as to accurately determine the failure stage corresponding to the physical node.

[0206] As an implementation of the present application, the database cluster health detection device 400 may further include:

[0207] The determination module is used to determine, for each physical node, the smaller value between the actual bearing pressure corresponding to the physical node and the bearing pressure prediction result as the bearing pressure health degree corresponding to the physical node.

[0208] The determination module is further used to determine, for each physical node, a smaller value between the actual business pressure corresponding to the physical node and the business pressure prediction result as the business pressure health level corresponding to the physical node.

[0209] The calculation module is used to calculate the average value of the load-bearing pressure health corresponding to each physical node to obtain the load-bearing pressure health of the database cluster.

[0210] The determination module is further configured to determine the smaller value between the load bearing pressure health of the database cluster and the business pressure health of the database cluster as the health of the database cluster.

[0211] The above-mentioned method of the embodiment of the present application determines, for each physical node, the smaller value between the above-mentioned actual bearing pressure corresponding to the physical node and the above-mentioned bearing pressure prediction result as the bearing pressure health degree corresponding to the physical node; and determines the smaller value between the above-mentioned actual business pressure corresponding to the physical node and the above-mentioned business pressure prediction result as the business pressure health degree corresponding to the physical node. Then, the average value of the bearing pressure health degree in the database cluster is taken as the bearing pressure health degree of the database cluster, and the average value of the business pressure health degree in the database cluster is taken as the business pressure health degree of the database cluster. Finally, the smaller value between the bearing pressure health degree of the above-mentioned database cluster and the business pressure health degree of the above-mentioned database cluster is determined as the health degree of the above-mentioned database cluster, which can comprehensively consider the situation of each physical node in the database cluster to accurately determine the health degree of the database cluster.

[0212] Each module in the database cluster health detection device provided in the embodiment of the present application can implement each step in the above-mentioned database cluster health detection method and achieve corresponding effects, which will not be repeated here for the sake of concise description.

[0213] Figure 5 A schematic diagram of the structure of the database cluster health detection hardware provided in an embodiment of the present application is shown.

[0214] The database cluster health detection device may include a processor 501 and a memory 502 storing computer program instructions.

[0215] Specifically, the processor 501 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0216] The memory 502 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 502 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 502 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 502 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 502 is a non-volatile solid-state memory.

[0217] The memory may include a read-only memory (ROM), a random access memory (RAM), a disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Therefore, generally, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the database cluster health detection method according to any one embodiment of the present disclosure.

[0218] The processor 501 reads and executes the computer program instructions stored in the memory 502 to implement any one of the database cluster health detection methods in the above embodiments.

[0219] In one example, the database cluster health detection device may further include a communication interface 503 and a bus 510. Figure 5 As shown, the processor 501, the memory 502, and the communication interface 503 are connected via a bus 510 and communicate with each other.

[0220] The communication interface 503 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0221] Bus 510 includes hardware, software or both, and the parts of online data flow billing equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front-end bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 510 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the present application considers any suitable bus or interconnection.

[0222] In addition, in combination with the method for detecting the health of a database cluster in the above embodiment, the embodiment of the present application can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by a processor, any one of the methods for detecting the health of a database cluster in the above embodiment is implemented.

[0223] An embodiment of the present application further provides a computer program product, including a computer program, which, when executed, implements any one of the database cluster health detection methods in the above embodiments.

[0224] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0225] The functional blocks shown in the above-mentioned block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0226] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0227] Aspects of the present disclosure are described above with reference to the flowchart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the function / action specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0228] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A database cluster health detection method, characterized in that: The database cluster includes multiple physical nodes, and the method includes: Obtaining the actual load bearing pressure, used time and actual business pressure of each physical node in the first time period; For each physical node, according to the usage time and service life of the physical node, a failure analysis is performed on the physical node to obtain a bearing pressure prediction result of the physical node in a second time period, where the second time period is after the first time period; For each physical node, predicting the business pressure of the physical node in the second time period according to the actual business pressure of the physical node, and obtaining a business pressure prediction result of the physical node; The health of the database cluster is determined according to the actual load-bearing pressure of each physical node, the actual business pressure, the load-bearing pressure prediction result, and the business pressure prediction result.

2. The database cluster health detection method according to claim 1, characterized in that: The predicting, for each physical node, the business pressure of the physical node in the second time period according to the actual business pressure of the physical node, to obtain the business pressure prediction result of the physical node includes: The actual business pressure is input into a preset long short-term memory network for calculation to obtain a business pressure prediction result of the physical node.

3. The database cluster health detection method according to claim 2, characterized in that: Before inputting the actual business pressure into a preset long short-term memory network for calculation to obtain a business pressure prediction result of the physical node, the method further includes: Performing Fourier transform on the actual business pressure to obtain a frequency domain representation of the actual business pressure; Expanding the frequency range of the frequency domain representation to obtain expanded frequency domain data; updating the actual business pressure based on an inverse Fourier transform result of the expanded frequency domain data; The inputting the actual business pressure into a preset long short-term memory network for calculation to obtain the business pressure prediction result of the physical node includes: The updated actual business pressure is input into a preset long short-term memory network for calculation to obtain a business pressure prediction result of the physical node.

4. The database cluster health detection method according to claim 2, characterized in that: Before inputting the actual business pressure into a preset long short-term memory network for calculation to obtain a business pressure prediction result of the physical node, the method further includes: Combining historical business pressure data of a physical node in multiple historical time periods into a sequence form in chronological order to obtain a business pressure data sequence, wherein the historical time period is before the first time period; For the historical business pressure in each time period in the business pressure data series, the following steps are performed respectively: Inputting the historical business pressure corresponding to the target time period into the long short-term memory network, and predicting the historical business pressure corresponding to the next time period of the target time period; the target time period is any time period among the multiple historical time periods; Determining a loss function value of the long short-term memory network according to the historical business pressure corresponding to the next time period; Determine whether the loss function value of the long short-term memory network meets a preset training stop condition; If the condition is not met, the model parameters of the long short-term memory network are adjusted, and the adjusted long short-term memory network is trained using the business pressure data sequence, and the historical business pressure corresponding to the target time period is returned to the long short-term memory network to predict the historical business pressure corresponding to the next time period of the target time period until the preset training stop condition is met to obtain the trained long short-term memory network.

5. The database cluster health detection method according to claim 1, characterized in that: The performing failure analysis on the physical node according to the usage time and service life of the physical node to obtain a prediction result of the bearing pressure of the physical node in the second time period includes: Determine the failure stage corresponding to the physical node according to the usage time and service life of the physical node; determining a parameter value of a Weibull distribution based on the failure stage; According to the parameter value, the failure probability corresponding to the physical node is calculated according to Weibull distribution; According to the failure probability, a prediction result of the bearing pressure of the physical node in the second time period is calculated.

6. The database cluster health detection method according to claim 5, characterized in that: The determining, according to the usage time and service life of the physical node, the failure stage corresponding to the physical node includes: Based on the service life of the physical node, determine a first time threshold and a second time threshold corresponding to the physical node; When the usage time of the physical node is less than or equal to the first time threshold, determining that the failure stage corresponding to the physical node is an early failure period; When the usage time of the physical node is greater than the first time threshold and less than the second time threshold, determining that the failure stage corresponding to the physical node is a random failure period; When the usage time of the physical node is greater than or equal to the second time threshold, it is determined that the failure stage corresponding to the physical node is a degradation failure period.

7. The database cluster health detection method according to claim 1, characterized in that: The determining the health of the database cluster according to the actual load-bearing pressure of each physical node, the actual business pressure, the load-bearing pressure prediction result, and the business pressure prediction result includes: For each physical node, determining the smaller value between the actual bearing pressure corresponding to the physical node and the bearing pressure prediction result as the bearing pressure health degree corresponding to the physical node; For each physical node, determining the smaller value between the actual business pressure corresponding to the physical node and the business pressure prediction result as the business pressure health degree corresponding to the physical node; Calculate the average value of the load pressure health of each physical node to obtain the load pressure health of the database cluster; Calculate the average value of the business pressure health of each physical node to obtain the business pressure health of the database cluster; A smaller value between the load pressure health of the database cluster and the business pressure health of the database cluster is determined as the health of the database cluster.

8. A database cluster health detection device, characterized in that: The device comprises: An acquisition module is used to acquire the actual load bearing pressure, used time and actual business pressure of each physical node in the first time period; An analysis module is used to perform a failure analysis on each of the physical nodes according to the usage time and service life of the physical node, so as to obtain a prediction result of the bearing pressure of the physical node in a second time period; A prediction module, configured to predict the business pressure of each physical node in a second time period according to the actual business pressure of the physical node, and obtain a business pressure prediction result of the physical node; A determination module is used to determine the health of the database cluster according to the actual load-bearing pressure of each physical node, the actual business pressure, the load-bearing pressure prediction result and the business pressure prediction result.

9. A database cluster health detection device, characterized in that: The device comprises: a processor and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement the database cluster health detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the database cluster health detection method according to any one of claims 1 to 7 is implemented.

11. A computer program product, characterized in that When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the database cluster health detection method according to any one of claims 1 to 7.

Images