A monitoring and alarming system and method for a container cluster

Abstract

The application discloses a kind of monitoring and alarming system and method of container cluster, it is related to performance monitoring technical field, and the technical scheme effectively solves the problem that cross-node performance bottleneck is difficult to accurately locate, and the cross-node contention coefficient Kcz is calculated and evaluated by node performance monitoring module, and resource contention bottleneck is quickly located;System comprehensively captures the performance influence between microservices interaction, calculates microservice performance interference index Gwz using microservice interaction monitoring module, and realizes accurate evaluation of microservice call chain performance;Meanwhile, dynamic threshold self-adaptive module adjusts dynamic threshold adjustment factor Ztd by historical load and real-time data calculation, dynamically adjusts cross-node contention threshold Qc and microservice performance interference threshold Kw, combined with comprehensive performance health index Phz and health deviation value Apj grading alarm, to avoid false alarm and miss report, realize efficient monitoring and management to complex performance problems.

Description

Technical Field

[0001] This invention relates to the field of performance monitoring technology, specifically to a monitoring and alarm system and method for container clusters. Background Technology

[0002] Monitoring and alerting systems for container clusters originated from the rapid development of container technology and distributed computing. Containerization provides applications with a lightweight, isolated runtime environment, while cluster management tools enable large-scale deployment and efficient scheduling of containers. However, as the scale of container clusters expands, the complexity of system operation increases dramatically, making it difficult for traditional monitoring methods to meet real-time and granular requirements. Modern monitoring and alerting systems have gradually integrated intelligent analysis, visual monitoring, and dynamic alerting mechanisms, effectively improving the resource utilization and fault response speed of container clusters, laying the foundation for the widespread application of cloud computing and microservice architectures.

[0003] However, existing monitoring and alerting systems for container clusters often suffer from the following technical drawbacks:

[0004] 1. Difficulty in accurately locating cross-node performance bottlenecks: Performance issues in container clusters are often caused by resource contention among multiple nodes; existing monitoring systems lack effective correlation analysis between cross-node resource access latency and bottleneck location, resulting in low diagnostic efficiency. For example, in a multi-tenant environment, contention for shared network bandwidth may be mistakenly identified as a single-node resource problem.

[0005] 2. The performance impact of interactions between microservices is not fully captured: In a microservice architecture, frequent interactions between containers can create complex performance bottlenecks. However, most existing systems focus on monitoring the resources of a single container and cannot effectively assess the overall performance of the microservice call chain, such as the cumulative latency or link interruption caused by multiple calls.

[0006] 3. Static alarm thresholds lead to false alarms and missed alarms: Many monitoring systems use fixed alarm thresholds, while container resource usage fluctuates drastically with load and is difficult to adapt to dynamic changes; for example, a short-term resource peak during peak periods may trigger false alarms, while under gradually increasing load, it may miss alarms due to excessively high thresholds. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a monitoring and alarm system and method for container clusters, which solves the technical deficiencies mentioned in the background section.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solution: a monitoring and alarm system for container clusters, including a node performance monitoring module, a microservice interaction monitoring module, a dynamic threshold adaptive module, a comprehensive analysis module, and an alarm management module;

[0009] The node performance monitoring module is used to pre-acquire the electronic topology map of the node distribution in the container cluster, monitor the cross-node resource access latency in real time, and finally calculate and evaluate the cross-node contention coefficient Kcz to generate the corresponding bottleneck warning signal.

[0010] The microservice interaction monitoring module is used to perform in-depth analysis of the microservice call chain in the container cluster after receiving the bottleneck warning signal, calculate and evaluate the microservice performance interference index Gwz, and finally generate a microservice performance warning signal.

[0011] The dynamic threshold adaptive module collects and analyzes historical load change data and real-time resource usage data to construct a dynamic threshold adjustment factor Ztd; based on the dynamic threshold adjustment factor Ztd, it dynamically adjusts the cross-node contention threshold Qc and the microservice performance interference threshold Kw.

[0012] The comprehensive analysis module is used to construct and analyze the cross-node contention coefficient Kcz, microservice performance interference index Gwz, and dynamic threshold adjustment factor Ztd based on the container performance diagnostic model. Then, it calculates and evaluates the comprehensive performance health index Phz and issues a comprehensive performance health alarm signal.

[0013] The alarm management module is used to calculate and evaluate the comprehensive performance health deviation value Apj based on the evaluation content of the comprehensive performance health index Phz; finally, the comprehensive performance health alarm signals are managed in a hierarchical manner, and detailed alarm information is generated.

[0014] Preferably, the node performance monitoring module includes a node performance quantification unit and a node performance evaluation unit;

[0015] The node performance quantization unit first collects and constructs a latency dataset, and then calculates the cross-node contention coefficient Kcz based on the latency dataset. The specific calculation formula is as follows:

[0016]

[0017] In the formula, Tnw represents the network transmission delay in the delay dataset, Tio represents the I / O operation delay in the delay dataset, and Trp represents the RPC call delay in the delay dataset.

[0018] Preferably, the node performance evaluation unit is used to preset the cross-node contention threshold Qc, evaluate the cross-node contention coefficient Kcz, determine the bottleneck strength of resource contention in the current container cluster, and finally generate a corresponding bottleneck warning signal; the specific evaluation content is as follows:

[0019] If the cross-node contention coefficient Kcz ≤ cross-node contention threshold Qc, then the current resource contention intensity is considered normal, and no bottleneck warning signal is generated.

[0020] If the cross-node contention coefficient Kcz > the cross-node contention threshold Qc, then the current resource contention intensity is considered abnormal, indicating a bottleneck problem, and an abnormal bottleneck warning signal is generated.

[0021] Preferably, the microservice interaction monitoring module includes a link performance calculation unit and an interference assessment unit;

[0022] The link performance calculation unit collects data related to the interaction latency of the microservice call link, the cumulative latency of multiple calls, and the probability of link interruption, and then constructs a microservice link performance index dataset. Based on the microservice link performance index dataset, it extracts the single request response time Tre, the average call link latency Tav, the number of call retries Rre, and the link interruption rate Pbr from the dataset. After dimensionless processing, it calculates the microservice performance interference index Gwz using the following formula:

[0023]

[0024] Preferably, the interference assessment unit is used to preset a microservice performance interference threshold Kw, and compare and evaluate the microservice performance interference index Gwz to determine the health status of the microservice call chain performance and generate a microservice performance warning signal; the specific assessment content is as follows:

[0025] If the microservice performance interference index Gwz ≤ microservice performance interference threshold Kw, it means that the call chain performance is in a healthy state and no warning is needed.

[0026] If the microservice performance interference index Gwz > the microservice performance interference threshold Kw, it indicates that the call chain performance is being interfered with and there are potential performance problems. At this time, a microservice performance warning signal will be issued.

[0027] Preferably, the dynamic threshold adaptive module analyzes historical load change data and real-time resource usage data to obtain the short-term fluctuation value Sbd, the long-term trend value Cqx, the peak resource utilization rate Rup, and the current stability coefficient Stb. After performing dimensionless processing, it calculates the dynamic threshold adjustment factor Ztd using the following formula:

[0028]

[0029] Based on the dynamic threshold adjustment factor Ztd, the cross-node contention threshold Qc and the microservice performance interference threshold Kw are linearly adjusted, and the specific adjustment formula is as follows:

[0030] Qc' = Qc × (1 + Ztd);

[0031] Kw' = Kw × (1 + Ztd);

[0032] In the formula, Qc' represents the adjusted cross-node contention threshold, and Kw' represents the adjusted microservice performance interference threshold;

[0033] When the dynamic threshold adjustment factor Ztd > 0, the cross-node contention threshold Qc and the microservice performance interference threshold Kw will be increased by 10%.

[0034] When the dynamic threshold adjustment factor Ztd≤0, the cross-node contention threshold Qc and the microservice performance interference threshold Kw will be reduced by 10%;

[0035] Finally, the adjusted cross-node contention threshold Qc and microservice performance interference threshold Kw are applied to the node performance monitoring module and the microservice interaction monitoring module.

[0036] Preferably, the comprehensive analysis module includes a performance index calculation unit and a performance alarm evaluation unit:

[0037] The performance index calculation unit, based on the container performance diagnostic model, jointly analyzes the cross-node contention coefficient Kcz, the microservice performance interference index Gwz, and the dynamic threshold adjustment factor Ztd, and comprehensively extracts performance characteristic parameters, including the average resource contention intensity Rci, the microservice call chain stability factor Msf, and the dynamic threshold adjustment sensitivity Zsa. Finally, it calculates the comprehensive performance health index Phz using the following formula:

[0038]

[0039] Preferably, the performance alarm evaluation unit is used to preset a comprehensive performance health threshold Zj, and then evaluate the comprehensive performance health index Phz. The specific evaluation content is as follows:

[0040] If the overall performance health index Phz ≥ the overall performance health threshold Zj, it means that the container cluster performance is in a normal state and no warning is needed.

[0041] If the overall performance health index Phz < the overall performance health threshold Zj, it indicates that the container cluster performance is in an abnormal state, with performance problems or bottlenecks, and an overall performance health alarm signal will be issued.

[0042] Preferably, the alarm management module is used to calculate and evaluate the overall performance health deviation value Apj; when the cluster performance is in an abnormal state, the overall performance health deviation value Apj is calculated using the following formula:

[0043]

[0044] By presetting the first performance health deviation threshold W1 and the second performance health deviation threshold W2, the comprehensive performance health deviation value Apj is evaluated, and at the same time, the comprehensive performance health warning signal is hierarchically managed, and finally detailed warning information is generated; and the first performance health deviation threshold W1 is greater than the second performance health deviation threshold W2, and the specific content is as follows:

[0045] Level 1 warning signal: When Apj ≤ W2, it indicates that the performance fluctuation is slight and no emergency intervention is required;

[0046] Level 2 warning signal: When W2 < Apj ≤ W1, it indicates that the performance fluctuation is medium, and it is prompted to optimize the performance management strategy;

[0047] Level 3 warning signal: When Apj > W1, it indicates that the performance fluctuation seriously deviates from the healthy state. At this time, immediate intervention measures are taken, including adjusting the resource allocation strategy, adjusting the microservice call chain, and adjusting the dynamic threshold parameters.

[0048] A monitoring and warning method for a container cluster includes the following steps:

[0049] Step 1: Obtain the electronic topology map of the node distribution in the container cluster in advance, and monitor the cross-node resource access delay in real time. Finally, calculate and evaluate the cross-node contention coefficient Kcz, and generate the corresponding bottleneck warning signal;

[0050] Step 2: After receiving the bottleneck warning signal, deeply analyze the microservice call link in the container cluster, calculate and evaluate the microservice performance interference index Gwz, and finally generate the microservice performance warning signal;

[0051] Step 3: Collect historical load change data and real-time resource usage data and analyze them to construct the dynamic threshold adjustment factor Ztd; based on the dynamic threshold adjustment factor Ztd, dynamically adjust the cross-node contention threshold Qc and the microservice performance interference threshold Kw;

[0052] Step 4: Construct and analyze the cross-node contention coefficient Kcz, the microservice performance interference index Gwz, and the dynamic threshold adjustment factor Ztd based on the container performance diagnosis model, then calculate and evaluate the comprehensive performance health index Phz, and issue the comprehensive performance health warning signal;

[0053] Step 5: Calculate and evaluate the comprehensive performance health deviation value Apj according to the evaluation content of the comprehensive performance health index Phz; finally, hierarchically manage the comprehensive performance health warning signal, and finally generate detailed warning information.

[0054] The present invention provides a monitoring and warning system and method for a container cluster. It has the following beneficial effects:

[0055] (1) This container cluster monitoring and alarm system and method effectively solves the problem of difficult accurate location of cross-node performance bottlenecks. Through the node performance monitoring module, the network transmission latency Tnw, I / O operation latency Tio and RPC call latency Trp in the container cluster are monitored and collected in real time to construct a latency dataset and calculate the cross-node contention coefficient Kcz. Using the node performance evaluation unit, the cross-node contention coefficient Kcz is compared and analyzed with the preset cross-node contention threshold Qc, and a bottleneck warning signal is generated in a timely manner. Through the performance index calculation unit in the comprehensive analysis module, the average resource contention intensity Rci is used as one of the performance characteristic parameters. The comprehensive performance health index Phz is calculated by combining the microservice performance interference index Gwz and the dynamic threshold adjustment factor Ztd, so as to achieve accurate location and rapid response of cross-node resource contention.

[0056] (2) This container cluster monitoring and alarm system and method comprehensively captures the performance impact of interactions between microservices. Through the link performance calculation unit of the microservice interaction monitoring module, it extracts the single request response time Tre, average call link latency Tav, call retries Rre, and link interruption rate Pbr from the microservice link performance index dataset, constructs a microservice performance interference index Gwz, and compares it with the preset microservice performance interference threshold Kw. When the microservice performance interference index Gwz is greater than the microservice performance interference threshold Kw, a microservice performance warning signal is generated, indicating that there may be performance problems in the call chain. Combined with the microservice call chain stability factor Msf in the comprehensive analysis module, the microservice interaction performance is quantitatively analyzed to ensure that the overall performance of the call chain is comprehensively evaluated and to avoid omissions caused by the limitations of single container monitoring.

[0057] (3) This container cluster monitoring and alarm system and method effectively solves the problem of false alarms and missed alarms caused by static alarm thresholds. Through the dynamic threshold adaptive module, it analyzes historical load change data and real-time resource usage data, extracts short-term fluctuation value Sbd, long-term trend value Cqx, peak resource utilization Rup and current stability coefficient Stb, calculates dynamic threshold adjustment factor Ztd, and dynamically adjusts cross-node contention threshold Qc and microservice performance interference threshold Kw. The adjusted thresholds are applied to contention assessment and interference assessment processes to improve the monitoring system's adaptability to load fluctuations. Combined with the alarm management module, alarm events are classified and managed through the comprehensive performance health index Phz and comprehensive performance health deviation value Apj, generating mild, moderate and severe alarm signals to ensure that the monitoring system has higher accuracy and sensitivity. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the framework structure of a container cluster monitoring and alarm system according to the present invention;

[0059] Figure 2 This is a flowchart illustrating the steps of a container cluster monitoring and alarm method according to the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Example 1

[0062] Please see Figure 1 A monitoring and alarm system for container clusters includes a node performance monitoring module, a microservice interaction monitoring module, a dynamic threshold adaptive module, a comprehensive analysis module, and an alarm management module.

[0063] The node performance monitoring module is used to pre-acquire the electronic topology map of the node distribution in the container cluster, monitor the cross-node resource access latency in real time, and finally calculate and evaluate the cross-node contention coefficient Kcz to generate the corresponding bottleneck warning signal.

[0064] The microservice interaction monitoring module is used to perform in-depth analysis of the microservice call chain in the container cluster after receiving the bottleneck warning signal, calculate and evaluate the microservice performance interference index Gwz, and finally generate a microservice performance warning signal.

[0065] The dynamic threshold adaptive module collects and analyzes historical load change data and real-time resource usage data to construct a dynamic threshold adjustment factor Ztd; based on the dynamic threshold adjustment factor Ztd, it dynamically adjusts the cross-node contention threshold Qc and the microservice performance interference threshold Kw.

[0066] The comprehensive analysis module is used to construct and analyze the cross-node contention coefficient Kcz, microservice performance interference index Gwz, and dynamic threshold adjustment factor Ztd based on the container performance diagnostic model. Then, it calculates and evaluates the comprehensive performance health index Phz and issues a comprehensive performance health alarm signal.

[0067] The alarm management module is used to calculate and evaluate the comprehensive performance health deviation value Apj based on the evaluation content of the comprehensive performance health index Phz; finally, the comprehensive performance health alarm signals are managed in a hierarchical manner, and detailed alarm information is generated.

[0068] In this embodiment, the node performance monitoring module can collect network transmission latency Tnw, I / O operation latency Tio, and RPC call latency Trp, calculate the cross-node contention coefficient Kcz, and combine it with the cross-node contention threshold Qc to achieve accurate location and bottleneck warning of cross-node resource contention; the microservice interaction monitoring module can extract single request response time Tre, average call chain latency Tav, call retries Rre, and link interruption rate Pbr to construct a microservice link performance index dataset, calculate the microservice performance interference index Gwz, and compare it with the microservice performance interference threshold Kw for evaluation, thereby achieving accurate monitoring of microservice call chain performance; the dynamic threshold adaptive module analyzes historical load change data and real-time resource usage data to extract short-term fluctuation value Sbd and long-term trend value C. The system uses qx, peak resource utilization Rup, and current stability coefficient Stb to calculate a dynamic threshold adjustment factor Ztd, dynamically adjusting the cross-node contention threshold Qc and microservice performance interference threshold Kw to improve the system's adaptability to load fluctuations and the accuracy of alarms. The comprehensive analysis module calculates a comprehensive performance health index Phz by combining the average resource contention intensity Rci, the microservice call chain stability factor Msf, and the dynamic threshold adjustment sensitivity Zsa, achieving an accurate assessment of the overall performance status of the container cluster. The alarm management module calculates a comprehensive performance health deviation value Apj based on the assessment results of the comprehensive performance health index Phz, and combines the first performance health deviation threshold W1 and the second performance health deviation threshold W2 to perform hierarchical management of alarm events, generating detailed alarm information, effectively improving the traceability and processing efficiency of performance issues.

[0069] Example 2

[0070] The node performance monitoring module includes a node performance quantification unit and a node performance evaluation unit;

[0071] The node performance quantization unit first collects and constructs a latency dataset, and then calculates the cross-node contention coefficient Kcz based on the latency dataset. The specific calculation formula is as follows:

[0072]

[0073] In the formula, Tnw represents the network transmission delay in the delay dataset, Tio represents the I / O operation delay in the delay dataset, and Trp represents the RPC call delay in the delay dataset.

[0074] The node performance evaluation unit is used to preset the cross-node contention threshold Qc, evaluate the cross-node contention coefficient Kcz, determine the bottleneck strength of resource contention in the current container cluster, and finally generate the corresponding bottleneck warning signal; the specific evaluation content is as follows:

[0075] If the cross-node contention coefficient Kcz ≤ cross-node contention threshold Qc, then the current resource contention intensity is considered normal, and no bottleneck warning signal is generated.

[0076] If the cross-node contention coefficient Kcz > the cross-node contention threshold Qc, then the current resource contention intensity is considered abnormal, indicating a bottleneck problem, and an abnormal bottleneck warning signal is generated.

[0077] The microservice interaction monitoring module includes a link performance calculation unit and an interference assessment unit;

[0078] The link performance calculation unit collects data related to the interaction latency of the microservice call link, the cumulative latency of multiple calls, and the probability of link interruption, and then constructs a microservice link performance index dataset. Based on the microservice link performance index dataset, it extracts the single request response time Tre, the average call link latency Tav, the number of call retries Rre, and the link interruption rate Pbr from the dataset. After dimensionless processing, it calculates the microservice performance interference index Gwz using the following formula:

[0079]

[0080] The interference assessment unit is used to preset the microservice performance interference threshold Kw, and compare and evaluate the microservice performance interference index Gwz to determine the health status of the microservice call chain performance and generate a microservice performance warning signal; the specific assessment content is as follows:

[0081] If the microservice performance interference index Gwz ≤ microservice performance interference threshold Kw, it means that the call chain performance is in a healthy state and no warning is needed.

[0082] If the microservice performance interference index Gwz > the microservice performance interference threshold Kw, it indicates that the call chain performance is being interfered with and there are potential performance problems. At this time, a microservice performance warning signal will be issued.

[0083] The dynamic threshold adaptive module analyzes historical load change data and real-time resource usage data to obtain the short-term fluctuation value Sbd, long-term trend value Cqx, peak resource utilization rate Rup, and current stability coefficient Stb. After performing dimensionless processing, it calculates the dynamic threshold adjustment factor Ztd using the following formula:

[0084]

[0085] Based on the dynamic threshold adjustment factor Ztd, the cross-node contention threshold Qc and the microservice performance interference threshold Kw are linearly adjusted, and the specific adjustment formula is as follows:

[0086] Qc' = Qc × (1 + Ztd);

[0087] Kw' = Kw × (1 + Ztd);

[0088] In the formula, Qc' represents the adjusted cross-node contention threshold, and Kw' represents the adjusted microservice performance interference threshold;

[0089] When the dynamic threshold adjustment factor Ztd > 0, the cross-node contention threshold Qc and the microservice performance interference threshold Kw will be increased by 10%.

[0090] When the dynamic threshold adjustment factor Ztd≤0, the cross-node contention threshold Qc and the microservice performance interference threshold Kw will be reduced by 10%;

[0091] Finally, the adjusted cross-node contention threshold Qc and microservice performance interference threshold Kw are applied to the node performance monitoring module and the microservice interaction monitoring module.

[0092] In this embodiment, multiple modules work together to achieve comprehensive monitoring and accurate alerts for container cluster performance. The node performance monitoring module generates bottleneck warning signals to ensure that resource contention issues can be accurately located. Specifically, network transmission latency Tnw represents the delay time of data transmission between container nodes, reflecting the stability and transmission efficiency of the network link; I / O operation latency Tio refers to the input / output operation latency when accessing shared storage or a distributed file system across nodes, reflecting disk read / write speed and the responsiveness of the storage system; RPC call latency Trp represents the response latency when microservices communicate via Remote Procedure Call (RPC), reflecting the efficiency and interaction performance of inter-service calls; and the cross-node contention coefficient Kcz is used to quantify the intensity of resource contention between nodes; a higher value indicates more severe cross-node resource contention.

[0093] The microservice interaction monitoring module generates microservice performance warning signals to achieve dynamic performance monitoring of the call chain. Among these, the single request response time (Tre) represents the total response time from initiation to completion of a single microservice request, reflecting the microservice's real-time processing capability; the average call chain latency (Tav) represents the average response latency of each node in the microservice call chain, reflecting the overall efficiency of the call chain; the number of call retries (Rre) refers to the number of service retries caused by request failures or timeouts, reflecting the service's reliability and stability; the link interruption rate (Pbr) represents the probability that a request in the microservice call chain is interrupted or fails, directly reflecting the reliability of service interaction; and the microservice performance interference index (Gwz) is used to assess the overall health of the microservice call chain, with a higher Gwz value indicating severe performance interference in the call chain.

[0094] The dynamic threshold adaptive module analyzes historical load change data and real-time resource usage data to calculate the dynamic threshold adjustment factor Ztd, and dynamically adjusts the cross-node contention threshold Qc and the microservice performance interference threshold Kw to improve the system's adaptability to dynamic load changes. Specifically, the short-term fluctuation value Sbd represents the fluctuation range of system resource usage within a short period, reflecting the system's sensitivity to sudden load changes; the long-term trend value Cqx describes the trend of system resource usage over time, reflecting the long-term load change trend; the peak resource utilization Rup represents the highest utilization rate of resources (such as CPU, memory, network, etc.) within the monitoring period, reflecting the load limit state; and the current stability coefficient Stb measures the system's operational stability under the current load, considering indicators such as system failure rate and error rate.

[0095] The acquisition of each sub-parameter can quantify different performance dimensions of the cluster operation. The calculation and evaluation of each parameter further classifies and quantifies performance issues, ultimately achieving real-time monitoring, alarm classification, and precise optimization of system performance.

[0096] Example 3

[0097] The comprehensive analysis module includes a performance index calculation unit and a performance alarm evaluation unit:

[0098] The performance index calculation unit, based on the container performance diagnostic model, jointly analyzes the cross-node contention coefficient Kcz, the microservice performance interference index Gwz, and the dynamic threshold adjustment factor Ztd, and comprehensively extracts performance characteristic parameters, including the average resource contention intensity Rci, the microservice call chain stability factor Msf, and the dynamic threshold adjustment sensitivity Zsa. Finally, it calculates the comprehensive performance health index Phz using the following formula:

[0099]

[0100] The performance alarm evaluation unit is used to preset the comprehensive performance health threshold Zj, and then evaluate the comprehensive performance health index Phz. The specific evaluation content is as follows:

[0101] If the overall performance health index Phz ≥ the overall performance health threshold Zj, it means that the container cluster performance is in a normal state and no warning is needed.

[0102] If the overall performance health index Phz < the overall performance health threshold Zj, it indicates that the container cluster performance is in an abnormal state, with performance problems or bottlenecks, and an overall performance health alarm signal will be issued.

[0103] The alarm management module is used to calculate and evaluate the overall performance health deviation value Apj. When the cluster performance is in an abnormal state, the overall performance health deviation value Apj is calculated using the following formula:

[0104]

[0105] By presetting the first performance health deviation threshold W1 and the second performance health deviation threshold W2, the comprehensive performance health deviation value Apj is evaluated, and at the same time, hierarchical management is carried out on the comprehensive performance health warning signal, and finally detailed warning information is generated; and the first performance health deviation threshold W1 is greater than the second performance health deviation threshold W2, and the specific content is as follows:

[0106] Level 1 warning signal: When Apj ≤ W2, it indicates that the performance fluctuation is slight and no emergency intervention is required;

[0107] Level 2 warning signal: When W2 < Apj ≤ W1, it indicates that the performance fluctuation is medium, and it is prompted to optimize the performance management strategy;

[0108] Level 3 warning signal: When Apj > W1, it indicates that the performance fluctuation seriously deviates from the healthy state. At this time, immediate intervention measures are taken, including adjusting the resource allocation strategy, adjusting the microservice call chain, and adjusting the dynamic threshold parameters.

[0109] In this embodiment, through the collaborative work of the comprehensive analysis module and the warning management module, the system performance monitoring and warning capabilities are comprehensively improved; in the comprehensive analysis module, the comprehensive performance health index Phz is calculated through formulas to achieve multi-dimensional quantification of the overall performance of the cluster; the performance warning evaluation unit judges whether there are performance problems or bottlenecks in the system by evaluating the health state of the comprehensive performance health index Phz, and generates warning signals in a timely manner; among them, the average resource contention intensity Rci represents the average level of cross-node resource contention within a period of time and is used for long-term monitoring of the resource competition situation; the microservice call chain stability factor Msf synthesizes various performance indicators of the microservice call chain and quantifies its stability; the dynamic threshold adjustment sensitivity Zsa describes the response speed and amplitude of the system to dynamic threshold adjustment and measures the system's ability to adapt to performance fluctuations;

[0110] The warning management module calculates the comprehensive performance health deviation value Apj to quantify the deviation degree between the current performance state and the health threshold, and reflects the severity of the system performance deviating from the normal state; combined with the first performance health deviation threshold W1 and the second performance health deviation threshold W2, hierarchical management is carried out on warning events. The level 1 warning indicates that the performance fluctuation is slight and no emergency intervention is required, the level 2 warning indicates that the performance fluctuation is medium and the performance management strategy needs to be optimized, and the level 3 warning indicates that the performance seriously deviates from the healthy state and measures such as adjusting the resource allocation strategy, optimizing the microservice call chain, or adjusting the dynamic threshold parameters need to be taken immediately; the collection and calculation of each lower-level parameter respectively quantify the key dimensions of resource contention, call chain performance, and system adaptability, providing a scientific basis for the accurate positioning and efficient optimization of performance problems.

[0111] Example 4

[0112] Please see Figure 2 A monitoring and alerting method for container clusters includes the following steps:

[0113] Step 1: First, obtain the electronic topology map of the node distribution in the container cluster, monitor the cross-node resource access latency in real time, and finally calculate and evaluate the cross-node contention coefficient Kcz to generate the corresponding bottleneck warning signal.

[0114] Step 2: After receiving the bottleneck warning signal, perform in-depth analysis of the microservice call chain in the container cluster, calculate and evaluate the microservice performance interference index Gwz, and finally generate a microservice performance warning signal.

[0115] Step 3: Collect and analyze historical load change data and real-time resource usage data to construct a dynamic threshold adjustment factor Ztd; based on the dynamic threshold adjustment factor Ztd, dynamically adjust the cross-node contention threshold Qc and the microservice performance interference threshold Kw.

[0116] Step 4: Build and analyze the cross-node contention coefficient Kcz, microservice performance interference index Gwz, and dynamic threshold adjustment factor Ztd based on the container performance diagnostic model. Then calculate and evaluate the comprehensive performance health index Phz and issue a comprehensive performance health alarm signal.

[0117] Step 5: Based on the evaluation content of the comprehensive performance health index Phz, calculate and evaluate the comprehensive performance health deviation value Apj; finally, manage the comprehensive performance health alarm signals in a hierarchical manner and generate detailed alarm information.

[0118] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A monitoring and alarm system for a container cluster, characterized in that: It includes a node performance monitoring module, a microservice interaction monitoring module, a dynamic threshold adaptive module, a comprehensive analysis module, and an alarm management module; The node performance monitoring module is used to pre-acquire the electronic topology map of the node distribution in the container cluster, monitor the cross-node resource access latency in real time, and finally calculate and evaluate the cross-node contention coefficient Kcz to generate the corresponding bottleneck warning signal. The microservice interaction monitoring module is used to perform in-depth analysis of the microservice call chain in the container cluster after receiving the bottleneck warning signal, calculate and evaluate the microservice performance interference index Gwz, and finally generate a microservice performance warning signal. The dynamic threshold adaptive module collects and analyzes historical load change data and real-time resource usage data to construct a dynamic threshold adjustment factor Ztd. Based on the dynamic threshold adjustment factor Ztd, the cross-node contention threshold Qc and the microservice performance interference threshold Kw are dynamically adjusted. The comprehensive analysis module is used to construct and analyze the cross-node contention coefficient Kcz, microservice performance interference index Gwz, and dynamic threshold adjustment factor Ztd based on the container performance diagnostic model. Then, it calculates and evaluates the comprehensive performance health index Phz and issues a comprehensive performance health alarm signal. The alarm management module is used to calculate and evaluate the comprehensive performance health deviation value Apj based on the evaluation content of the comprehensive performance health index Phz. Finally, the overall performance health alarm signals are managed hierarchically, and detailed alarm information is generated. The node performance monitoring module includes a node performance quantification unit and a node performance evaluation unit; The node performance quantization unit first collects and constructs a latency dataset, and then calculates the cross-node contention coefficient Kcz based on the latency dataset. The specific calculation formula is as follows: ； In the formula, Tnw represents the network transmission delay in the delay dataset, Tio represents the I / O operation delay in the delay dataset, and Trp represents the RPC call delay in the delay dataset; The node performance evaluation unit is used to preset the cross-node contention threshold Qc, evaluate the cross-node contention coefficient Kcz, determine the bottleneck strength of resource contention in the current container cluster, and finally generate the corresponding bottleneck warning signal; the specific evaluation content is as follows: If the cross-node contention coefficient Kcz ≤ cross-node contention threshold Qc, then the current resource contention intensity is considered normal, and no bottleneck warning signal is generated. If the cross-node contention coefficient Kcz > the cross-node contention threshold Qc, then the current resource contention intensity is considered abnormal and there is a bottleneck problem. At this time, an abnormal bottleneck warning signal is generated. The microservice interaction monitoring module includes a link performance calculation unit and an interference assessment unit; The link performance calculation unit collects data related to the interaction latency of the microservice call link, the cumulative latency of multiple calls, and the probability of link interruption, and then constructs a microservice link performance index dataset. Based on the microservice link performance index dataset, it extracts the single request response time Tre, the average call link latency Tav, the number of call retries Rre, and the link interruption rate Pbr from the dataset. After dimensionless processing, it calculates the microservice performance interference index Gwz using the following formula: ； The interference assessment unit is used to preset the microservice performance interference threshold Kw, compare and evaluate the microservice performance interference index Gwz, determine the health status of the microservice call link performance, and generate a microservice performance warning signal. The specific assessment content is as follows: If the microservice performance interference index Gwz ≤ microservice performance interference threshold Kw, it means that the call chain performance is in a healthy state and no warning is needed. If the microservice performance interference index Gwz > the microservice performance interference threshold Kw, it indicates that the call chain performance is interfered with and there is a potential performance problem. At this time, a microservice performance warning signal is issued. The dynamic threshold adaptive module analyzes historical load change data and real-time resource usage data to obtain the short-term fluctuation value Sbd, long-term trend value Cqx, peak resource utilization rate Rup, and current stability coefficient Stb. After performing dimensionless processing, it calculates the dynamic threshold adjustment factor Ztd using the following formula: ； Based on the dynamic threshold adjustment factor Ztd, the cross-node contention threshold Qc and the microservice performance interference threshold Kw are linearly adjusted, and the specific adjustment formula is as follows: ； ； In the formula, This represents the adjusted cross-node contention threshold. This indicates the adjusted microservice performance interference threshold; When the dynamic threshold adjustment factor Ztd > 0, the cross-node contention threshold Qc and the microservice performance interference threshold Kw will be increased by 10%; When the dynamic threshold adjustment factor Ztd≤0, the cross-node contention threshold Qc and the microservice performance interference threshold Kw will be reduced by 10%; Finally, the adjusted cross-node contention threshold Qc and microservice performance interference threshold Kw are applied to the node performance monitoring module and the microservice interaction monitoring module.

2. The monitoring and alarm system for a container cluster according to claim 1, characterized in that: The comprehensive analysis module includes a performance index calculation unit and a performance alarm evaluation unit: The performance index calculation unit, based on the container performance diagnostic model, jointly analyzes the cross-node contention coefficient Kcz, the microservice performance interference index Gwz, and the dynamic threshold adjustment factor Ztd, and comprehensively extracts performance characteristic parameters, including the average resource contention intensity Rci, the microservice call chain stability factor Msf, and the dynamic threshold adjustment sensitivity Zsa. Finally, it calculates the comprehensive performance health index Phz using the following formula: ； The average resource contention intensity Rci represents the average level of cross-node resource contention over a period of time and is used for long-term monitoring of resource contention. The microservice call chain stability factor Msf integrates various performance indicators of the microservice call chain to quantify its stability. The dynamic threshold adjustment sensitivity Zsa describes the system's response speed and magnitude to dynamic threshold adjustments, and measures the system's ability to adapt to performance fluctuations.

3. The monitoring and alarm system for a container cluster according to claim 2, characterized in that: The performance warning evaluation unit is used to preset the comprehensive performance health threshold Zj, and then evaluate the comprehensive performance health index Phz. The specific evaluation content is as follows: If the comprehensive performance health index Phz ≥ the comprehensive performance health threshold Zj, it means that the performance of the container cluster is in a normal state and no warning is required; If the comprehensive performance health index Phz < the comprehensive performance health threshold Zj, it means that the performance of the container cluster is in an abnormal state, there are performance problems or bottlenecks, and at this time, a comprehensive performance health warning signal is issued.

4. The monitoring and alarm system for a container cluster according to claim 3, characterized in that: The warning management module is used to calculate and evaluate the comprehensive performance health deviation value Apj; when the cluster performance is in an abnormal state, the comprehensive performance health deviation value Apj is calculated and obtained through the following formula: ； By presetting the first performance health deviation threshold W1 and the second performance health deviation threshold W2, the comprehensive performance health deviation value Apj is evaluated, and at the same time, the comprehensive performance health warning signal is managed hierarchically, and finally detailed warning information is generated; And the first performance health deviation threshold W1 is greater than the second performance health deviation threshold W2, and the specific content is as follows: Level 1 warning signal: When Apj ≤ W2, it means that the performance fluctuation is slight and no emergency intervention is required; Level 2 warning signal: When W2 < Apj ≤ W1, it means that the performance fluctuation is medium, and it is prompted to optimize the performance management strategy; Level 3 warning signal: When Apj > W1, it means that the performance fluctuation seriously deviates from the healthy state, and at this time, intervention measures are immediately taken, including adjusting the resource allocation strategy, adjusting the microservice call chain, and adjusting the dynamic threshold parameters.

5. A monitoring and alarm method for a container cluster, applied to a monitoring and alarm system for a container cluster as described in any one of claims 1 to 4, characterized in that: It includes the following steps: Step 1: Obtain the electronic topology map of the node distribution in the container cluster in advance, and monitor the cross-node resource access delay in real time. Finally, calculate and evaluate the cross-node contention coefficient Kcz, and generate the corresponding bottleneck warning signal; Step 2: After receiving the bottleneck warning signal, deeply analyze the microservice call link in the container cluster, calculate and evaluate the microservice performance interference index Gwz, and finally generate the microservice performance warning signal; Step 3: Collect historical load change data and real-time resource usage data and analyze them to construct the dynamic threshold adjustment factor Ztd; Based on the dynamic threshold adjustment factor Ztd, dynamically adjust the cross-node contention threshold Qc and the microservice performance interference threshold Kw; Step 4: Construct and based on the container performance diagnosis model, analyze the cross-node contention coefficient Kcz, the microservice performance interference index Gwz, and the dynamic threshold adjustment factor Ztd, and then calculate and evaluate the comprehensive performance health index Phz, and issue the comprehensive performance health warning signal; Step 5: Calculate and evaluate the comprehensive performance health deviation value Apj according to the evaluation content of the comprehensive performance health index Phz; Finally, the comprehensive performance health warning signal is managed hierarchically, and finally detailed warning information is generated.

Images