Cluster interactive deployment management system and method based on cloud platform

Abstract

The present invention discloses a cluster interactive deployment management system and method based on a cloud platform, and relates to the technical field of cluster interactive deployment management. The system comprises: a cluster deployment module, a request processing module, and a cluster status monitoring module; the cluster deployment module is used to analyze the number and deployment priority of each container node in each container network in the cluster, and then select the target server where the container node is deployed according to the adaptability of each container network and a physical server and the link association between the physical servers; the request processing module allocates the requested container node according to the available load rate of the physical server deployed by each container node in the container network corresponding to the request and the resource occupancy rate of the current request allocated to each physical server; the cluster status monitoring module is used to monitor the status of each container node and the request in real time in the cluster, and to make real-time adjustments to the request allocation and process feedback of the abnormal status.

Description

Technical Field

[0001] The present invention relates to the technical field of cluster interactive deployment management, and in particular to a cluster interactive deployment management system and method based on a cloud platform. Background Art

[0002] Cluster interactive deployment management refers to the effective deployment and management of applications in a cluster environment and the allocation of resources and task loads between cluster nodes. It is the core to ensure the stable operation of large-scale distributed applications.

[0003] In today's cluster interaction deployment management, automated deployment and load balancing strategies are often used for cluster interaction deployment management, which has the following technical defects: First, when deploying cluster nodes, an overall deployment method is adopted to deploy the same functional services to all cluster nodes, which inevitably leads to functional redundancy. The actual performance of different nodes is different, and the overall deployment also lacks consideration of the resource consumption of different functional modules of the cluster; secondly, in distributed deployment, there is a lack of reasonable analysis of the communication performance and functional adaptability between different physical servers, so additional data processing is required to ensure data consistency, which increases the cost of data maintenance; in addition, when deploying container nodes, there is a lack of analysis of the performance of each physical server in processing different container network requests, and it is impossible to ensure the consistency of performance during cluster deployment, resulting in resource contention and resource redundancy in different container networks. Summary of the invention

[0004] The purpose of the present invention is to provide a cluster interactive deployment management system and method based on a cloud platform to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The cluster interactive deployment management method based on the cloud platform includes the following analysis steps:

[0007] S100: Divide the cluster network into different container networks according to service content and implementation functions, and set the priority coefficient of each container network in cluster deployment;

[0008] S200: Analyze the current saturated load of the container network and the resource consumption of each physical server in processing the current container network request, and dynamically set the number of container nodes;

[0009] S300: Analyze the node deployment priority of each container node in the cluster deployment, and determine the target server for node deployment in each container network in turn;

[0010] S400: When a request is received, a target container network is selected and a processing container node is allocated, the request and the status of the allocated processing container node are monitored in real time, and the container node allocation result is adjusted in real time.

[0011] In the above technical solution, step S100 includes the following contents:

[0012] The cluster comprehensive service content is divided into functional modules, and each functional module is processed into a separate container network, and then the cluster network is modularly divided into different container networks for separate deployment. By modularly dividing the cluster functions, the cluster network is divided into different container networks, which avoids the redundancy of container node functions caused by the overall cluster deployment during cluster deployment, and realizes the full planning and utilization of physical server resources.

[0013] Obtain the process sequence of the functional modules corresponding to each container network in the cluster service process, and comprehensively set the priority coefficient of each container network in the cluster deployment according to the sequence of the functional modules corresponding to the container network in the cluster service process and the interference of the failure of the functional modules corresponding to the container network on the overall service of the cluster;

[0014] Among them, the priority coefficient of the container network with the corresponding functional module at the front in the cluster service process is greater than that of the container network with the corresponding functional module at the back; when there is no process sequence between the corresponding functional modules of the two container networks, the interference of each container network to the overall service of the cluster is evaluated when a functional failure occurs, and the priority coefficient of the container network with greater interference caused by the functional failure is greater than that of the container network with less interference caused by the functional failure; by setting the priority coefficient of the container network, ensure that the container network with a higher order and higher importance in the cluster service process is deployed first, so as to ensure the scientificity and rationality of the overall process of cluster deployment and operation.

[0015] In the above technical solution, step S200 includes the following contents:

[0016] Analyze the maximum number of parallel requests after the current container network deployment as the current container network saturation load N sat , test the resource occupancy rate of the current container network deployment in each physical server to process a single request through the unified cluster deployment interactive platform, set the maximum resource occupancy rate of each physical server, and use the formula: Calculate the maximum parallel load N when the current physical server is used as the target server of the container node in the current container network max ;

[0017] Among them, r th is the current maximum resource occupancy of the physical server, r used is the resource occupancy rate of the current physical server, r singleThe resource usage of the current container network deployment in the current physical server to process a single request. is the sign for rounding x down;

[0018] By setting the maximum resource occupancy rate of the physical server, we can avoid the problems of different container nodes in the same physical server competing for resources when processing user requests, resulting in reduced processing efficiency and resource utilization, which can lead to server overload and backlog of user requests, affecting the actual user experience. We can then allocate the available resources of each physical server in detail, and use the maximum parallel request volume of each container network to analyze the resource occupancy of each container network during node deployment, ensuring that the cluster network can run stably and smoothly during peak request volume.

[0019] For any container network, when allocating container node target servers, calculate the maximum parallel load N of each physical server as the container node target server in the current container network. max , each time a physical server is selected to deploy a container node, the current container network saturation load N sat Subtract N max , until the current container network saturation load is allocated, and the number of selected physical servers is recorded as the number of current container network master nodes;

[0020] Set the redundant load ratio factor k redu , calculate N redu =N max *k redu As the redundant load of the current container network, when the saturated load is allocated, the above allocation method is used to continue to select the target server as the redundant node from the remaining physical servers where the current container network node is not deployed, and the number of selected physical servers is recorded as the number of redundant nodes of the current container network; the number of container nodes in the current container network is obtained by adding the number of the current container network master node and the number of redundant nodes;

[0021] By setting the redundant load ratio coefficient to analyze the number of redundant nodes, redundant nodes are set differently for different container networks, avoiding the situation where the container network with a smaller maximum parallel request volume and the container network with a larger maximum parallel request volume have the same number of redundant nodes, which would lead to waste of server resources and increased supervision and maintenance costs.

[0022] In the above technical solution, the node deployment priority of each container node in the cluster deployment is analyzed in step S300, including the following contents:

[0023] When performing global cluster deployment, obtain the priority coefficient of the container network to which each container node belongs in the cluster deployment, and calculate the node deployment priority of each container node in the cluster deployment according to the formula:

[0024]

[0025] Among them, P is the node deployment priority of the current container node in the cluster deployment, k assi N is the priority coefficient of the container network in the cluster deployment. assi N is the allocated load in the saturated load of the container network. sat is the saturated load of the container network, τ is the attribute identifier of the current container node; τ values ​​include {a 1 ,a 2}, where a 1 >a 2 , when the current container node is the master node of the container network to which it belongs, τ=a 1 , when the current container node is a redundant node of the container network to which it belongs, τ=a 2 ;

[0026] By globally analyzing the node deployment priority of container nodes in each container network, we ensure the synchronous deployment of the scale of each functional module during cluster deployment, avoid the large differences in scale and performance between different functional modules that lead to redundant resource waste and node resource competition, and ensure the consistency of the overall service process of the cluster; at the same time, set the container node attribute identifier, and give priority to the deployment of the main node in the container network to ensure the integrity of the cluster function in actual use.

[0027] In the above technical solution, determining the target server for node deployment in each container network in step S300 includes the following contents:

[0028] For any container network, the unified cluster deployment interaction platform sends the requests processed by the current container network to each physical server, collects and tests the average time for each physical server to process the requests, and then calculates the compatibility between the current container network and each physical server according to the formula:

[0029]

[0030] Among them, F is the adaptability between the current container network and the physical server, N max N is the maximum parallel load of the current physical server as the target server of the container node in the current container network. sat is the current container network saturation load, exp() is the exponential function of the natural logarithm, t req The average time it takes for the current physical server to process requests handled by the current container network;

[0031] Select the physical server with the highest compatibility with the current container network as the target server of the first container node; select the first container node of the current container network by analyzing the compatibility between the container network and the physical server to ensure the performance requirements of the container network when it is actually used. At the same time, introduce the maximum parallel load when the current physical server is used as the target server of the container node in the current container network for analysis to avoid the situation where there are too many container nodes and the deployment is too dispersed when the container node selects the target server, thereby reducing the cost of cluster operation and maintenance;

[0032] The unified cluster deployment interaction platform is used to test the average transmission time of the remaining physical servers that have not deployed the current container network node to send data packets of the same size to the first container node target server, and the link association between the remaining physical servers that have not deployed the current container network node and the first container node target server is calculated according to the formula:

[0033]

[0034] Where L is the link affinity between the physical server and the first container node, D tran To test the transmission data packet size, t tran is the average transmission time of the test data packet, B total is the transmission link bandwidth capacity between the current physical server and the first container node, exp() is the exponential function of the natural logarithm;

[0035] According to the formula: W = F*L, calculate the remaining physical servers where the current container network node is not deployed and the screening weight W of the current container network, and select physical servers in descending order according to the screening weights as the target servers of the container nodes in the current container network;

[0036] By analyzing the link correlation between nodes, the subsequent master node screening of the container network is carried out to ensure the stability and smoothness of the communication links between container nodes within the container network during actual use, thereby ensuring the consistency of cluster data and reducing the cost of backend data maintenance.

[0037] In the above technical solution, step S400 includes the following contents:

[0038] When the cluster receives a user request through the unified cluster deployment interaction platform, it determines the container network to which the request belongs and allocates container nodes;

[0039] Obtain the load request data of the physical servers deployed on each master node in the current container network through the unified cluster deployment interactive platform, and calculate the available load rate of the current physical server according to the formula:

[0040]

[0041] Among them, r load is the available load rate of the current physical server, r th is the maximum resource occupancy of the current physical server, n is the number of the container node deployed in the current physical server, N node is the number of container nodes deployed in the current physical server, The resource usage of container node n deployed in the current physical server to process a single request. The number of requests processed in parallel by the container node n deployed in the current physical server;

[0042] Sort the container nodes in the current container network from large to small according to the available load, and select the container node with the largest available load to process the current request;

[0043] Analyze the resource usage when the current request is allocated to each physical server single,req , if r single,req >r load , determine that all the main nodes in the current container network are saturated, use redundant nodes to calculate and sort the available load rate, and determine the container node that processes the current request; if the available load rate of all redundant nodes is less than r single,req , a container network load saturation signal is sent to postpone the current request. When any container node is processing a request, the container node is allocated again. The container nodes are sorted according to the available load. When the main node has no available resources, redundant nodes are enabled to process requests to ensure the timeliness of user request processing under high concurrency.

[0044] The unified cluster deployment interactive platform is used to monitor the status of container nodes in each container network in real time. When any container node is assigned a request and the container node is in an idle state, the container node is judged to be in an abnormal state and abnormal feedback of the container node is sent. When abnormal feedback of the container node is received, the request assigned to the abnormal container node is reallocated to the container node for processing, and the physical server information deployed by the abnormal container node is fed back to the supervisor.

[0045] When any request can be allocated a container node and the request is in a pending state, the request is judged to be in an abnormal state and a request abnormality feedback is sent; when a request abnormality feedback is received, the container node allocation of the abnormal request is stopped, and the source user information of the abnormal request is fed back to the supervisor.

[0046] A cluster interactive deployment management system based on a cloud platform that applies the cluster interactive deployment management method based on a cloud platform in the above technical solution includes a cluster deployment module, a request processing module, and a cluster status monitoring module;

[0047] The cluster deployment module is used to analyze the number and deployment priority of each container node in each container network in the cluster, and then select the target server for deploying the container node according to the adaptability of each container network and physical server and the link correlation between physical servers; the request processing module allocates the requested container node according to the available load rate of the physical server deployed by each container node in the container network corresponding to the request and the resource occupancy rate of the current request allocated to each physical server; the cluster status monitoring module is used to monitor the status of each container node and request in the cluster in real time, and make real-time adjustments to the request allocation and process feedback of abnormal status.

[0048] In the above technical solution, the cluster deployment module includes: a container network division unit, a deployment priority analysis unit, and a target server selection unit;

[0049] The container network division unit divides each part of the functions in the cluster comprehensive service into separate container networks; the deployment priority analysis unit is used to set the priority coefficient of each container network in the cluster deployment, and analyze the node deployment priority of each container node in each container network in the cluster deployment; the target server selection unit is used to select the target server of each container node for container deployment.

[0050] In the above technical solution, the request processing module includes: a load monitoring unit and a request allocation unit;

[0051] The load monitoring unit is used to monitor the requests allocated to each container node and calculate the available load rate of each container node; the request allocation unit is used to analyze the resource occupancy rate of user requests received by the unified cluster deployment interaction platform and processed in each physical server, and allocate container nodes to the received user requests.

[0052] In the above technical solution, the cluster status monitoring module includes: a container node abnormality monitoring unit, a request abnormality monitoring unit, and an abnormality handling unit;

[0053] The container node abnormality monitoring unit is used to monitor whether there is a container node allocated to the request and in an idle state; the request abnormality monitoring unit is used to monitor whether there is a request that can be allocated to a container node and is in a pending state; the abnormality processing unit is used to process and provide abnormality feedback for container nodes and user requests with container node abnormalities and request abnormalities.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention performs detailed analysis and management on cluster deployment and user request processing through a unified cluster deployment interactive platform; on the one hand, by dividing the cluster service functions into modules, the cluster network is divided into separate container networks and then each container network is deployed, thereby reducing functional redundancy and resource waste caused by the overall deployment of the cluster network; on the other hand, the performance of each physical server is analyzed, and the scale of container network deployment in different physical servers is dynamically set to make full use of the performance resources of each physical server. At the same time, the maximum resource occupancy rate of the physical server is set to ensure that each physical server is in the performance linear area when in use, avoiding excessive load leading to a decrease in overall performance and resource utilization, and ensuring the overall stability of the cluster service; in addition, when selecting the target server for the container node, the adaptability of each physical server to the current container network function requirements and the performance of the communication link between servers are fully analyzed to ensure the performance and data consistency of each container network in actual use, and at the same time, the priority of each container node in the global deployment of the cluster is calculated and analyzed to ensure the consistency of the actual performance of each functional module in the cluster service, and avoid the problems of resource redundancy and resource competition in different container networks at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a flow chart of the cluster interactive deployment management method based on the cloud platform of the present invention;

[0057] Figure 2 This is an organizational chart of the cluster interactive deployment management system based on the cloud platform of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] Example: See Figure 1-Figure 2 , the present invention provides a technical solution,

[0060] like Figure 1 As shown, the present invention provides a cluster interactive deployment management method based on a cloud platform, comprising the following analysis steps:

[0061] S100: Divide the cluster network into different container networks according to service content and implementation functions, and set the priority coefficient of each container network in cluster deployment;

[0062] S200: Analyze the current saturated load of the container network and the resource consumption of each physical server in processing the current container network request, and dynamically set the number of container nodes;

[0063] S300: Analyze the node deployment priority of each container node in the cluster deployment, and determine the target server for node deployment in each container network in turn;

[0064] S400: When a request is received, a target container network is selected and a processing container node is allocated, the request and the status of the allocated processing container node are monitored in real time, and the container node allocation result is adjusted in real time.

[0065] Step S100 includes the following contents:

[0066] Divide the cluster's comprehensive service content into functional modules, and divide the processing of each functional module into a separate container network, and then modularize the cluster network into different container networks for separate deployment;

[0067] Obtain the process sequence of the functional modules corresponding to each container network in the cluster service process, and comprehensively set the priority coefficient of each container network in the cluster deployment according to the sequence of the functional modules corresponding to the container network in the cluster service process and the interference of the failure of the functional modules corresponding to the container network on the overall service of the cluster;

[0068] Among them, the priority coefficient of the container network with the corresponding functional module in the front order in the cluster service process is greater than that of the container network with the corresponding functional module in the back order; when there is no process sequence between the corresponding functional modules of the two container networks, the interference to the overall service of the cluster when each container network has a functional failure is evaluated, and the priority coefficient of the container network with greater interference caused by the functional failure is greater than that of the container network with less interference caused by the functional failure;

[0069] In specific implementation, due to the scale and complexity of cluster services, modular programming is generally used when writing applications. Therefore, modular programs are divided, and data is shared between program modules through data communication, so that the overall cluster service is divided into different container networks; when setting priorities, the actual usage frequency or processing request concurrency of the corresponding functional modules of different container networks can be counted, and the container network with a higher usage frequency or a higher processing request concurrency is more important.

[0070] Step S200 includes the following contents:

[0071] Analyze the maximum number of parallel requests after the current container network deployment as the current container network saturation load N sat , test the resource occupancy rate of the current container network deployment in each physical server to process a single request through the unified cluster deployment interactive platform, set the maximum resource occupancy rate of each physical server, and use the formula: Calculate the maximum parallel load N when the current physical server is used as the target server of the container node in the current container networkmax ;

[0072] Among them, r th is the current maximum resource occupancy of the physical server, r used is the resource occupancy rate of the current physical server, r single The resource usage of the current container network deployment in the current physical server to process a single request. is the sign for rounding x down;

[0073] In the specific implementation, the maximum resource occupancy rate of the physical server is set to the linear and nonlinear interval points of the physical server resource usage, and the linear and nonlinear are the relationship between the server processing request concurrency and the resource occupancy rate; for the above calculation formula, the following specific implementation example is given: Assuming that in the current physical server, r th =80%,r used =50%,r single =5%, the maximum parallel load N is calculated max =6;

[0074] In addition, in the specific implementation, due to the different functional modules and processing requests corresponding to different container networks, the occupancy of different resources is also different. When selecting the resources to be considered, such as CPU occupancy, I / O occupancy, bandwidth occupancy, etc., r is calculated for different resources. single , and select r single The resource with the largest value is used as the consideration resource to avoid performance shortcomings that may cause the overall performance of the cluster to deteriorate during actual use.

[0075] For any container network, when allocating container node target servers, calculate the maximum parallel load N of each physical server as the container node target server in the current container network. max , each time a physical server is selected to deploy a container node, the current container network saturation load N sat Subtract N max , until the current container network saturation load is allocated, and the number of selected physical servers is recorded as the number of current container network master nodes;

[0076] In specific implementation, the minimum number of master nodes can be set. When allocating target servers, if a single physical server can bear all the load requirements of the container network, the load requirements will be divided and a container node will be deployed in the current physical server to ensure the stability of the container network and avoid the downtime of the entire container network due to the failure of the current physical server.

[0077] Set the redundant load ratio factor k redu , calculate N redu =N max *k reduAs the redundant load of the current container network, when the saturated load is allocated, the above allocation method is used to continue to select the target server as the redundant node from the remaining physical servers where the current container network node is not deployed, and the number of selected physical servers is recorded as the number of redundant nodes of the current container network;

[0078] Add the number of primary nodes and redundant nodes in the current container network to get the number of container nodes in the current container network.

[0079] The node deployment priority of each container node in the cluster deployment is analyzed in step S300, including the following contents:

[0080] When performing global cluster deployment, obtain the priority coefficient of the container network to which each container node belongs in the cluster deployment, and calculate the node deployment priority of each container node in the cluster deployment according to the formula:

[0081]

[0082] Among them, P is the node deployment priority of the current container node in the cluster deployment, k assi N is the priority coefficient of the container network in the cluster deployment. assi N is the allocated load in the saturated load of the container network. sat is the saturated load of the container network, τ is the attribute identifier of the current container node; τ values ​​include {a 1 ,a 2}, where a 1 >a 2 , when the current container node is the master node of the container network to which it belongs, τ=a 1 , when the current container node is a redundant node of the container network to which it belongs, τ=a 2 ;

[0083] For the above calculation formula, the following specific implementation example is given: Assuming that the current container node k assi =2,N assi =50,N sat =100,a 1 =0.8,a 2 =0.2. If the current container node is the master node of the container network to which it belongs, the node deployment priority of the current container node in the cluster deployment is calculated to be P=0.8. If the current container node is a redundant node of the container network to which it belongs, the node deployment priority of the current container node in the cluster deployment is calculated to be P=0.2.

[0084] Determining the target server for node deployment in each container network in step S300 includes the following:

[0085] For any container network, the unified cluster deployment interaction platform sends the requests processed by the current container network to each physical server, collects and tests the average time for each physical server to process the requests, and then calculates the compatibility between the current container network and each physical server according to the formula:

[0086]

[0087] Among them, F is the adaptability between the current container network and the physical server, N max N is the maximum parallel load of the current physical server as the target server of the container node in the current container network. sat is the current container network saturation load, exp() is the exponential function of the natural logarithm, t req The average time it takes for the current physical server to process requests handled by the current container network;

[0088] Select the physical server with the highest adaptability to the current container network as the target server of the first container node. Test the average transmission time of the remaining physical servers that have not deployed the current container network node to send the same size data packets to the target server of the first container node through the unified cluster deployment interaction platform. Calculate the link association between the remaining physical servers that have not deployed the current container network node and the first container node according to the formula:

[0089]

[0090] Where L is the link affinity between the physical server and the first container node, D tran To test the transmission data packet size, t tran is the average transmission time of the test data packet, B total is the transmission link bandwidth capacity between the current physical server and the first container node, exp() is the exponential function of the natural logarithm;

[0091] According to the formula: W = F*L, the remaining physical servers where the current container network node is not deployed and the screening weight W of the current container network are calculated, and the physical servers are selected in descending order according to the screening weights as the target servers of the container nodes in the current container network.

[0092] Step S400 includes the following contents:

[0093] When the cluster receives a user request through the unified cluster deployment interaction platform, it determines the container network to which the request belongs and allocates container nodes;

[0094] Obtain the load request data of the physical servers deployed on each master node in the current container network through the unified cluster deployment interactive platform, and calculate the available load rate of the current physical server according to the formula:

[0095]

[0096] Among them, r load is the available load rate of the current physical server, r th is the maximum resource occupancy of the current physical server, n is the number of the container node deployed in the current physical server, N node is the number of container nodes deployed in the current physical server, The resource usage of container node n deployed in the current physical server to process a single request. The number of requests processed in parallel by the container node n deployed in the current physical server;

[0097] For the above calculation formula, the following specific implementation example is given: Assuming that the maximum resource occupancy rate r in the current physical server th =90%, number of nodes N node =3; among them, node 1: Node 2: Node 3: Calculate the current available load rate r of the physical server load =16%;

[0098] Sort the container nodes in the current container network from large to small according to the available load, and select the container node with the largest available load to process the current request;

[0099] Analyze the resource usage when the current request is allocated to each physical server single,req , if r single,req >r load , determine that all the main nodes in the current container network are saturated, use redundant nodes to calculate and sort the available load rate, and determine the container node that processes the current request; if the available load rate of all redundant nodes is less than r single,req , a container network load saturation signal is sent to postpone the current request, and the container node is allocated again when the request being processed by any container node is completed;

[0100] The unified cluster deployment interactive platform is used to monitor the status of container nodes in each container network in real time. When any container node is assigned a request and the container node is in an idle state, the container node is judged to be in an abnormal state and abnormal feedback of the container node is sent. When abnormal feedback of the container node is received, the request assigned to the abnormal container node is reallocated to the container node for processing, and the physical server information deployed by the abnormal container node is fed back to the supervisor.

[0101] When any request can be allocated a container node and the request is in a pending state, the request is judged to be in an abnormal state and a request abnormality feedback is sent; when a request abnormality feedback is received, the container node allocation of the abnormal request is stopped, and the source user information of the abnormal request is fed back to the supervisor;

[0102] During the specific implementation, if a container node abnormality occurs, a node maintenance signal can be sent to the supervisor. If a request abnormality occurs, the request needs to be isolated and the source of the request needs to be checked to avoid risks such as data leakage within the cluster.

[0103] like Figure 2 As shown, the present invention also provides a cluster interactive deployment management system based on a cloud platform, including a cluster deployment module, a request processing module, and a cluster status monitoring module;

[0104] The cluster deployment module is used to analyze the number and deployment priority of each container node in each container network in the cluster, and then select the target server for deploying the container node according to the adaptability of each container network and physical server and the link correlation between physical servers; the request processing module allocates the requested container node according to the available load rate of the physical server deployed by each container node in the container network corresponding to the request and the resource occupancy rate of the current request allocated to each physical server; the cluster status monitoring module is used to monitor the status of each container node and request in the cluster in real time, and make real-time adjustments to the request allocation and process feedback of abnormal status.

[0105] The cluster deployment module includes: container network division unit, deployment priority analysis unit, and target server selection unit;

[0106] The container network division unit divides each part of the functions in the cluster comprehensive service into separate container networks; the deployment priority analysis unit is used to set the priority coefficient of each container network in the cluster deployment, and analyze the node deployment priority of each container node in each container network in the cluster deployment; the target server selection unit is used to select the target server of each container node for container deployment.

[0107] The request processing module includes: a load monitoring unit and a request allocation unit;

[0108] The load monitoring unit is used to monitor the requests allocated to each container node and calculate the available load rate of each container node; the request allocation unit is used to analyze the resource occupancy rate of user requests received by the unified cluster deployment interaction platform and processed in each physical server, and allocate container nodes to the received user requests.

[0109] The cluster status monitoring module includes: a container node abnormality monitoring unit, a request abnormality monitoring unit, and an abnormality handling unit;

[0110] The container node abnormality monitoring unit is used to monitor whether there is a container node allocated to the request and in an idle state; the request abnormality monitoring unit is used to monitor whether there is a request that can be allocated to a container node and is in a pending state; the abnormality processing unit is used to process and provide abnormality feedback for container nodes and user requests with container node abnormalities and request abnormalities.

[0111] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A cluster interactive deployment management method based on a cloud platform, characterized in that , the method comprises the following analysis steps: S100: Divide the cluster network into different container networks according to service content and implementation functions, and set the priority coefficient of each container network in cluster deployment; S200: Analyze the current saturated load of the container network and the resource consumption of each physical server in processing the current container network request, and dynamically set the number of container nodes; S300: Analyze the node deployment priority of each container node in the cluster deployment, and determine the target server for node deployment in each container network in turn; S400: When a request is received, a target container network is selected and a processing container node is allocated, the request and the status of the allocated processing container node are monitored in real time, and the container node allocation result is adjusted in real time.

2. The cluster interactive deployment management method based on the cloud platform according to claim 1 is characterized in that ,Step S100 includes the following contents: Divide the cluster's comprehensive service content into functional modules, and divide the processing of each functional module into a separate container network, and then modularize the cluster network into different container networks for separate deployment; Obtain the process sequence of the functional modules corresponding to each container network in the cluster service process, and comprehensively set the priority coefficient of each container network in the cluster deployment according to the sequence of the functional modules corresponding to the container network in the cluster service process and the interference of the failure of the functional modules corresponding to the container network on the overall service of the cluster; Among them, the priority coefficient of the container network with the corresponding functional module at the front in the cluster service process is greater than that of the container network with the corresponding functional module at the back; when there is no process sequence between the corresponding functional modules of the two container networks, the interference to the overall service of the cluster when each container network has a functional failure is evaluated, and the priority coefficient of the container network with greater interference caused by the functional failure is greater than that of the container network with less interference caused by the functional failure.

3. The cluster interactive deployment management method based on the cloud platform according to claim 1 is characterized in that ,Step S200 includes the following contents: Analyze the maximum number of parallel requests after the current container network deployment as the current container network saturation load N sat , test the resource occupancy rate of the current container network deployment in each physical server to process a single request through the unified cluster deployment interactive platform, set the maximum resource occupancy rate of each physical server, and use the formula: Calculate the maximum parallel load N when the current physical server is used as the target server of the container node in the current container network max ; Among them, r th is the current maximum resource occupancy of the physical server, r used is the resource occupancy rate of the current physical server, r single The resource usage of the current container network deployment in the current physical server to process a single request. is the sign for rounding x down; For any container network, when allocating container node target servers, calculate the maximum parallel load N of each physical server as the container node target server in the current container network. max , each time a physical server is selected to deploy a container node, the current container network saturation load N sat Subtract N max , until the current container network saturation load is allocated, and the number of selected physical servers is recorded as the number of current container network master nodes; Set the redundant load ratio factor k redu , calculate N redu =N max *k redu As the redundant load of the current container network, when the saturated load is allocated, the above allocation method is used to continue to select the target server as the redundant node from the remaining physical servers where the current container network node is not deployed, and the number of selected physical servers is recorded as the number of redundant nodes of the current container network; Add the number of primary nodes and redundant nodes in the current container network to get the number of container nodes in the current container network.

4. The cluster interactive deployment management method based on the cloud platform according to claim 3 is characterized in that ,In step S300, analyzing the node deployment priority of each container node in the cluster deployment includes the following contents: When performing global cluster deployment, obtain the priority coefficient of the container network to which each container node belongs in the cluster deployment, and calculate the node deployment priority of each container node in the cluster deployment according to the formula: Among them, P is the node deployment priority of the current container node in the cluster deployment, k assi N is the priority coefficient of the container network in the cluster deployment. assi N is the allocated load in the saturated load of the container network. sat is the saturated load of the container network to which it belongs, and τ is the attribute identifier of the current container node; the value of τ includes {a1, a2}, where a1>a2. When the current container node is the primary node of the container network to which it belongs, τ=a1; when the current container node is the redundant node of the container network to which it belongs, τ=a2.

5. The cluster interactive deployment management method based on the cloud platform according to claim 3 is characterized in that: Determining the target server for node deployment in each container network in step S300 includes the following: For any container network, the unified cluster deployment interaction platform sends the requests processed by the current container network to each physical server, collects and tests the average time for each physical server to process the requests, and then calculates the compatibility between the current container network and each physical server according to the formula: Among them, F is the adaptability between the current container network and the physical server, N max N is the maximum parallel load of the current physical server as the target server of the container node in the current container network. sat is the current container network saturation load, exp() is the exponential function of the natural logarithm, t req The average time it takes for the current physical server to process requests handled by the current container network; Select the physical server with the highest adaptability to the current container network as the target server of the first container node. Test the average transmission time of the remaining physical servers that have not deployed the current container network node to send the same size data packets to the target server of the first container node through the unified cluster deployment interaction platform. Calculate the link association between the remaining physical servers that have not deployed the current container network node and the first container node according to the formula: Where L is the link affinity between the physical server and the first container node, D tran To test the transmission data packet size, t tran is the average transmission time of the test data packet, B total is the transmission link bandwidth capacity between the current physical server and the first container node, exp() is the exponential function of the natural logarithm; According to the formula: W = F*L, the remaining physical servers where the current container network node is not deployed and the screening weight W of the current container network are calculated, and the physical servers are selected in descending order according to the screening weights as the target servers of the container nodes in the current container network.

6. The cluster interactive deployment management method based on the cloud platform according to claim 1 is characterized in that ,Step S400 includes the following contents: When the cluster receives a user request through the unified cluster deployment interaction platform, it determines the container network to which the request belongs and allocates container nodes; Obtain the load request data of the physical servers deployed on each master node in the current container network through the unified cluster deployment interactive platform, and calculate the available load rate of the current physical server according to the formula: Among them, r load is the available load rate of the current physical server, r th is the maximum resource occupancy of the current physical server, n is the number of the container node deployed in the current physical server, N node is the number of container nodes deployed in the current physical server, The resource usage of container node n deployed in the current physical server to process a single request. The number of requests processed in parallel by the container node n deployed in the current physical server; Sort the container nodes in the current container network from large to small according to the available load, and select the container node with the largest available load to process the current request; Analyze the resource usage when the current request is allocated to each physical server single,req , if r single,req >r load , determine that all the main nodes in the current container network are saturated, use redundant nodes to calculate and sort the available load rate, and determine the container node that processes the current request; if the available load rate of all redundant nodes is less than r single,req , a container network load saturation signal is sent to postpone the current request, and the container node is allocated again when the request being processed by any container node is completed; The unified cluster deployment interactive platform is used to monitor the status of container nodes in each container network in real time. When any container node is assigned a request and the container node is in an idle state, the container node is judged to be in an abnormal state and abnormal feedback of the container node is sent. When abnormal feedback of the container node is received, the request assigned to the abnormal container node is reallocated to the container node for processing, and the physical server information deployed by the abnormal container node is fed back to the supervisor. When any request can be allocated a container node and the request is in a pending state, the request is judged to be in an abnormal state and a request abnormality feedback is sent; when a request abnormality feedback is received, the container node allocation of the abnormal request is stopped, and the source user information of the abnormal request is fed back to the supervisor.

7. A cluster interactive deployment management system based on a cloud platform using the cluster interactive deployment management method based on a cloud platform according to any one of claims 1 to 6, characterized in that ,The system includes: a cluster deployment module, a request processing module, and a cluster ,status monitoring module; The cluster deployment module is used to analyze the number and deployment priority of each container node in each container network in the cluster, and then select the target server for deploying the container node according to the adaptability of each container network and physical server and the link correlation between physical servers; the request processing module allocates the requested container node according to the available load rate of the physical server deployed by each container node in the container network corresponding to the request and the resource occupancy rate of the current request allocated to each physical server; the cluster status monitoring module is used to monitor the status of each container node and request in the cluster in real time, and make real-time adjustments to the request allocation and process feedback of abnormal status.

8. The cloud platform-based cluster interactive deployment management system according to claim 7, characterized in that ,The cluster deployment module includes: a container network partitioning unit, a deployment priority analysis unit, and a target server selection unit; The container network division unit divides each part of the functions in the cluster comprehensive service into separate container networks; the deployment priority analysis unit is used to set the priority coefficient of each container network in the cluster deployment, and analyze the node deployment priority of each container node in each container network in the cluster deployment; the target server selection unit is used to select the target server of each container node for container deployment.

9. The cloud platform-based cluster interactive deployment management system according to claim 7, characterized in that ,The request processing module includes: a load monitoring unit and a request allocation unit; The load monitoring unit is used to monitor the requests allocated to each container node and calculate the available load rate of each container node; the request allocation unit is used to analyze the resource occupancy rate of user requests received by the unified cluster deployment interaction platform and processed in each physical server, and allocate container nodes to the received user requests.

10. The cloud platform-based cluster interactive deployment management system according to claim 7, characterized in that ,The cluster status monitoring module includes: a container node abnormality monitoring unit, a request abnormality monitoring unit, and an abnormality handling unit; The container node abnormality monitoring unit is used to monitor whether there is a container node allocated to the request and in an idle state; the request abnormality monitoring unit is used to monitor whether there is a request that can be allocated to a container node and is in a pending state; the abnormality processing unit is used to process and provide abnormality feedback for container nodes and user requests with container node abnormalities and request abnormalities.

Images